Indie Kings

ShadowMountPlus: A PS5 Auto-Mounter That Warns of Data Corruption

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:55 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: ShadowMountPlus is a background auto-mounter for jailbroken PlayStation 5 consoles. It finds game dumps on internal or external storage, mounts them, and registers them in the system library without opening a separate payload for each title. Its own README opens with a warning that mounting images can cause shutdown problems and data corruption on internal drives.

That warning is the story. ShadowMountPlus is not a game loader with a friendly front end. It is a low-level storage tool that attaches disk images to a console the user has already modified, and it changes what the system sees at boot. The project is popular, with 1,066 stars and 92 forks at the time of check, but it is still publishing beta builds. The latest tag, 1.7beta7, arrived on October 21, 2026.

Image: GitHub social preview for the ShadowMountPlus repository. Image credit: drakmor/ShadowMountPlus, via GitHub.

What ShadowMountPlus does

The project describes itself as a fully automated background auto-mounter payload for jailbroken PS5 consoles. It detects, mounts, and installs game dumps from internal and external storage. The point is to remove manual steps that otherwise require tools such as DumpRunner or Itemzflow for every title.

It works by watching configured scan paths. When it finds a supported image or a folder game, it attaches the image through the console's storage layers, reads the title metadata, and registers the game so it appears in the library. The README says the first scan can take a few seconds on a large library, and games of 100 GB or more may need extra time for the system to verify integrity before the installed notification appears.

The tool also handles backport overlays, per-game fakelib overlays, emulator updates, and a small web interface with an HTTP and JSON API. Those features make it more than a mount script. It is closer to a storage service that runs alongside the console shell and reacts to games starting and stopping.

GitHub records the repository as a fork of RDX-Sci01's ShadowMount1.3GBT, which in turn descends from ShadowMount. The README's credits section names Drakmor for the evolution of ShadowMount into ShadowMountPlus, VoidWhisper for ShadowMount itself, and BestPig for BackPort.

Item	Detail from the repository
Project	ShadowMountPlus for jailbroken PlayStation 5
Language	C
Licence	GPL-3.0
Stars and forks	1,066 stars, 92 forks, 15 open issues at check
Repository created	February 13, 2026
Latest tag	1.7beta7 on October 21, 2026
Compatibility	Jailbroken PS5 with kstuff-lite v1.07 or newer, including firmware 2.00
Default API bind	127.0.0.1 on port 10101

The data corruption warning

The README's first paragraph after the repository links is a warning, not a feature list: mounting images can cause shutdown problems and data corruption on internal drives. It says the risk depends on many factors and is more common with older firmware versions.

That is not a generic disclaimer. Attaching disk images to a modified console asks the storage stack to do something it was not designed to do in that configuration. A badly formed image, an interrupted write, a power loss, or a firmware mismatch can leave the backing file or the console's internal storage in a bad state. The warning is not scoped to one mount mode. The project defaults to read-only mounts, which reduces one class of risk, but it does not remove the warning.

The README also documents a recovery trap. If a mounted game does not start, one troubleshooting step is to remove the game data from system settings and retry. The README notes that this will delete game saves. That is the kind of instruction that should be read before touching a library, not after.

Two defaults are worth knowing. Read-only mode is on by default, and automatic removal of missing games is off by default. If a user turns on auto_remove_missing_games, the tool will remove a game from the system library after its source has been unavailable for a configured delay, 300 seconds by default. The delay exists to avoid removing a game during a temporary USB disconnect, but it is still an automated delete from the library.

Supported image formats

ShadowMountPlus supports four image families. The README recommends UFS for normal use and treats the other formats as compatibility or experimental paths. The distinction matters because the mount backend and the filesystem are not the same thing.

Extension	Filesystem	Status in the README
`.ffpkg`	UFS	Recommended for normal use
`.exfat`	exFAT	Compatibility path, mainly for titles that need external-drive-style handling
`.ffpfs`	PFS	Experimental
`.ffpfsc`	PFS container	Experimental container for nested images

For manually built exFAT images, the README asks for a 64 KB cluster size and warns that smaller clusters can reduce performance. It also documents a fast path for images stored on the internal BFS volume, with a 512-byte logical sector and a 64 KiB secondary unit. Those are storage-tuning details, and getting them wrong is one reason the project ships builder scripts for Linux, macOS, and Windows.

Folder games are also supported. The layout rule is strict: game files must sit at the image root, with sce_sys/param.json directly inside. An extra top-level folder inside the image produces a missing or invalid parameter file error. That single rule explains many of the troubleshooting entries in the README.

Installation and autoload

There are two installation paths. The manual path sends shadowmountplus.elf to port 9021 with a payload sender. The recommended path adds the ELF to autoload.txt for plk-autoloader so it starts on every boot. The sample order loads ShadowMountPlus, waits 3,000 milliseconds, then loads kstuff.

The project depends on kstuff-lite v1.07 or newer. Without that layer, the payload has nothing to hook. The README lists firmware 2.00 as supported, but it does not claim support for every firmware. Users should check their own version against the compatibility note before installing anything.

Configuration lives in /data/shadowmount/config.ini. If the file is missing, the payload creates it from a bundled example and uses built-in defaults. The web interface is optional, and the API binds to localhost by default. Changing api_bind_address to 0.0.0.0 exposes it to the network, which the README presents as an option rather than a recommendation.

Debug logging is on by default and writes to both the console and /data/shadowmount/debug.log. That is helpful during setup, but it also means a log file grows until the user sets debug=0 or cleans it up.

The configuration surface

The configuration file is large because the tool touches several layers at once. The most important settings are the ones that change what happens to a library without a user watching.

mount_read_only defaults to 1, so images are mounted read-only unless changed.
scan_interval_seconds defaults to 15, so the tool keeps looking for new sources.
stability_wait_seconds defaults to 10, so a file being copied is not processed immediately.
auto_remove_missing_games defaults to 0, but can be enabled with a delay.
persistent_image_mounts defaults to 0, so images are released when a game exits.
fan_target_temperature defaults to system control, but can be set to a target in degrees Celsius.

The fan setting is a reminder that this is not only a file manager. It can apply a fan target when a game starts, which changes console behavior beyond mounting. The README documents it, but a user should not enable it without understanding the thermal tradeoff.

The web interface adds another surface. It can show game cards, change fakelib modes, and trigger background work. The API is documented in docs/api.md and an OpenAPI manifest. Binding it to localhost is the safe default. Exposing it on a LAN without authentication would be a bad idea, and the README does not describe an authentication layer for the API.

Jailbreak, backups, and legal risk

ShadowMountPlus requires a jailbroken console. That already places it outside the normal warranty and support path. The README's warning about data corruption is the most concrete risk, but it is not the only one. A jailbroken console may be unstable after firmware changes, and a payload that hooks the shell can fail in ways that are hard to diagnose.

The legal side is separate. The tool mounts game dumps. It does not include a legal opinion or a licence to any game. In most jurisdictions, making a backup of a disc you own may be permitted, but downloading a copy you do not own is not. Users should treat this as a tool for their own backups and check their local law. Nothing in the repository suggests Sony has approved it.

The security posture is also worth a second look. The project is open source under GPL-3.0, which helps inspection, but the release assets are binaries. The latest release ships shadowmountplus.elf and a zip. There is no code-signing chain for a jailbroken console, so users have to trust the repository and the release. That is normal for this scene, but it is still trust.

FAQ

What is ShadowMountPlus?
It is a background payload for jailbroken PS5 consoles that finds, mounts, and registers game dumps from internal or external storage. It can also apply backports, fakelib overlays, and emulator updates.

Does ShadowMountPlus work on an unmodified PS5?
No. The README requires a jailbroken console running kstuff-lite v1.07 or newer. It lists firmware 2.00 as supported but does not claim support for every firmware.

Is it safe to use?
The project's own README warns that mounting images can cause shutdown problems and data corruption on internal drives, especially on older firmware. Read-only mode is the default, but the warning still applies.

Which image format should I use?
The README recommends .ffpkg with UFS for normal use. Use .exfat only for titles that need external-drive-style compatibility, and treat .ffpfs and .ffpfsc as experimental.

Does it delete games automatically?
Not by default. auto_remove_missing_games is off unless the user enables it. If enabled, a game is removed from the system library only after its source stays unavailable for the configured delay.

Is ShadowMountPlus official or endorsed by Sony?
No. It is an independent homebrew project for jailbroken consoles. The repository does not claim Sony approval, and using it can affect warranty, stability, and data.

Our take: ShadowMountPlus is a serious piece of storage engineering, and its README is more honest than most homebrew projects about the risk. The corruption warning, read-only default, and delayed auto-removal are the right instincts. The caution is that the project is still shipping beta tags and the failure mode is data loss, not a crashed app. Anyone testing it should use backups, avoid writable mounts unless necessary, and keep the console's important saves somewhere else.

Bottom Line

ShadowMountPlus is a GPL-3.0 auto-mounter for jailbroken PS5 consoles. It scans internal and external storage, mounts UFS, exFAT, and experimental PFS images, registers games, and can apply backports and fakelib overlays. It had 1,066 stars and 92 forks at the time of check, with 1.7beta7 released on October 21, 2026. The project's own warning about shutdown problems and internal-drive data corruption is the first thing to read. It requires a jailbroken console, works on game dumps the user supplies, and is not endorsed by Sony.

Sources

drakmor/ShadowMountPlus repository README, including the corruption warning, compatibility note, image formats, configuration keys, scan behavior, installation methods, and troubleshooting, verified October 21, 2026.

drakmor/ShadowMountPlus releases, including 1.7beta7 published October 21, 2026 and 1.7beta6-fix2 published the same day.

GitHub repository metadata for drakmor/ShadowMountPlus, verified October 21, 2026: 1,066 stars, 92 forks, 15 open issues, C, GPL-3.0, created February 13, 2026, last push October 21, 2026.

CleanYourTV: An Apple TV-Style Home for Android TV Bloat

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:52 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: CleanYourTV is an open-source tool that strips advertising and preloaded apps from Android TV and Google TV sets over Wi-Fi, then installs a small replacement home screen modeled on Apple TV. It does not root the TV and does not uninstall anything. The debloat changes are package disables that a restore script can undo.

The project went from a Saturday experiment to 15,898 stars in four days. Its README says the author gave Claude Code a TV model number, let it find the set on the local network, removed 25 Xiaomi packages, and wrote a launcher. The repository was created on October 17, 2026. Version 2.5.0 landed on October 21 with 147 TV models in its release notes, and the README's contributed list is already larger.

Image: CleanYourTV banner. Image credit: Gathin23/cleanyourtv, via the repository README.

What CleanYourTV does

Android TV and Google TV sets ship with a home screen that mixes installed apps with sponsored rows, recommendation feeds, and vendor services. CleanYourTV attacks that layer in two ways. It disables packages the owner does not want, and it installs a replacement launcher that shows only the apps the owner chooses.

The project's promise is narrow and specific: no root, no cable, nothing uninstalled, and a one-line restore. The TV keeps its factory firmware. The debug bridge is used to disable packages in the user profile, the same mechanism an Android phone uses when an app is disabled rather than removed. The original packages remain on the device and can be enabled again.

The replacement launcher is a native Android app written in Java. It is deliberately small, around 405 KB in the v2.5.0 release notes, and built for sets with 1 GB of RAM. The README describes it as about 429 KB installed, and the v2.5.0 notes add a CI size gate of 420 KB. It has no network access, no background services, and no video. The README says it uses AndroidX RecyclerView and nothing heavier.

Item	Detail from the repository
Project	CleanYourTV
Purpose	Debloat Android TV and Google TV, then replace the home screen
Language	Java
Licence	MIT per the README, though GitHub reports no recognised licence
Stars and forks	15,898 stars, 1,742 forks, 135 open issues at check
Repository created	October 17, 2026
Latest release	v2.5.0 on October 21, 2026
TV requirement	Android 8.0 or newer for the launcher; Fire TV gets debloat only
Binaries	None; the launcher is built from source

How the debloat works

The TV needs USB debugging or network debugging enabled in its developer options. The user accepts one authorization prompt from the computer. From there, the tool uses Android's package manager to run pm disable-user against a list of packages for that TV model.

The lists live in the repository's examples/ folder, one folder per model. The README counts contributions from TCL, Sony, Xiaomi, Philips, Hisense, Amazon Fire TV, Panasonic, Skyworth, BenQ, Sharp, Toshiba, NVIDIA Shield, and many smaller brands. The README's current table lists 158 models from 52 brands, while the v2.5.0 release notes count 147 TVs from 49 brands. Both numbers reflect a project that is adding models faster than the prose can be updated.

The process is not fully automatic. The README's quick start asks the user to paste PROMPT.md into a coding assistant that has shell access, such as Claude Code, Codex, or Gemini CLI, and fill in the TV model. The assistant discovers the bloat on that specific set and asks before touching anything the user might use. The README estimates about fifteen minutes with the repository as context.

That design is unusual. CleanYourTV is not a single binary that applies a universal list. It is a repository of scripts, model-specific lists, and a prompt for an AI assistant. The assistant is the installer. The scripts are the safety rails. The restore record is the undo button.

The launcher

The home screen is the part most people will notice. It has a pitch-black background, a dock of four favorite apps in a rounded band, and a top shelf that names the focused app in large type. The ambient glow takes its color from the app's artwork. The rest of the apps sit in a grid that slides up when the user presses Down.

The focused tile lifts and shows its name under a glass highlight. Three frosted buttons open Inputs, the network panel, and Settings. The Inputs panel lists HDMI ports with what is connected, plus AV and Live TV. A thin clock, date, and greeting sit on the home screen, and a Clock screensaver drifts so it does not burn in.

Themes can change colors, wallpaper, dock size, and column count without rebuilding the APK. Module cards let other apps declare a tile, but the launcher asks the user once per app and keeps it hidden by default. The launcher also supports 4 to 7 apps per row, reduced motion for weak GPUs, an optional app library, and Hebrew right-to-left layout.

The launcher follows the TV's language. The README lists English, German, Spanish, Catalan, French, Russian, Serbian Latin, Polish, Czech, Italian, Portuguese, Turkish, Arabic, Hebrew, Simplified Chinese, Traditional Chinese, Thai, Finnish, Vietnamese, Bangla, Indonesian, Norwegian Bokmål, Dutch, and Uzbek Latin. The v2.5.0 release notes count 21 languages, so the README is again ahead of the tag.

What you need to run it

The computer and the TV must be on the same Wi-Fi network. The TV needs developer options enabled and USB debugging or network debugging turned on. On some sets the option is called Network debugging or ADB debugging. The user accepts the computer's key on the TV.

The launcher needs JDK 17 and Android SDK platform 34 to build. The README gives one-line setup for macOS and Linux, and PowerShell scripts for Windows. There is no prebuilt APK. The stated reason is control: the build takes about a minute, and the README says you control what goes on your TV.

The TV must run Android 8.0 or newer for the launcher. The install script checks the TV's API level before building and reports both numbers if the set is too old. Fire TV is different: Fire OS will not hand the home screen to a third-party launcher, so the project supports the debloat half only. On the 2015 Fire TV Box with Fire OS 5, even the debloat half fails because every package manager command is refused.

Restore, updates, and recovery

Before disabling anything, the tool writes a record to ~/.cleanyourtv/ with one file per TV and list. The record stores whether each package was enabled, already disabled, or not installed, which component owned Home, and the three animation scales. A second run of the same list keeps the first record instead of overwriting it with the cleaned-up state.

The restore script replays that record. A package the user had already disabled stays disabled, and Home returns to the component that owned it. If the record is missing, the script falls back to enabling the whole list and resetting the animation scales, which is less precise but still reversible.

Firmware updates usually bring the disabled packages back. The project includes tools/check.py, which compares the TV against the list that was applied, prints what drifted, and exits with a failure code. With --fix, it reapplies the list after the same safety checks. A weekly cron job on a home server or Raspberry Pi can run the check and report drift.

The most important install rule is to disable the stock launcher only after the new launcher is installed and running. If the stock launcher is disabled first, Home can land on a factory test app or nothing at all. The install script orders this correctly.

The real risks

CleanYourTV does not flash firmware, root the TV, or delete packages. The README's own safety answer is that the risk is disabling something the TV depends on, which is why lists are applied in small batches with a restore script ready. That is accurate, but it understates two failure modes.

The first is the setup wizard on Google TV. The package com.google.android.tungsten.setupwraith can take Home and crash-loop the system after the stock launcher is disabled. The README says to disable it too, but first check whether the saved Wi-Fi network belongs to it. If it does, disabling it deletes the network. The tool skips setupwraith until the user re-joins Wi-Fi, which is a good guard, but the failure mode is still a TV that cannot get online.

The second is ADB exposure. Network ADB is plaintext on port 5555. On firmware with ro.adb.secure=0, any device on the LAN can get a shell. The README tells users to turn debugging off and revoke authorizations when finished, then confirm the port is closed. That step should not be skipped. A TV left with open ADB is a bigger security problem than a sponsored row.

The legal and warranty picture is simpler. Disabling packages in the user profile does not void a warranty by itself, but a manufacturer may treat it as unsupported modification. The README states the project is MIT licensed and that the Outfit typeface is under the SIL Open Font License, though GitHub's own licence field for the repository reports no recognised licence. Either way, the licence covers the code, not the TV.

FAQ

Does CleanYourTV root my TV?
No. It uses Android's debug bridge and pm disable-user to disable packages in the user profile. Nothing is flashed and nothing is uninstalled.

Can I undo the changes?
Yes. The tool records the TV's state before it changes anything and includes a restore script that replays that record. Keep the debloat list you applied.

Does it work on Fire TV?
Only the debloat half. Fire OS will not hand the home screen to a third-party launcher. The 2015 Fire TV Box with Fire OS 5 refuses even the package disable commands.

Do I need to know how to code?
Not much, but you do need a computer on the same network and a coding assistant with shell access. The README's quick start is built around pasting a prompt into Claude Code, Codex, or Gemini CLI.

Why is there no APK download?
The project does not ship prebuilt binaries. You build the launcher from source with JDK 17 and Android SDK 34, which takes about a minute after setup.

What happens after a firmware update?
The disabled packages usually come back and Home returns to the stock launcher. Run tools/check.py to see what drifted, then --fix to reapply the list.

Our take: CleanYourTV is the rare debloat project that treats reversibility as a feature instead of a footnote. The restore record, the stock-launcher ordering, and the setupwraith guard are the work of someone who has actually broken a TV and had to fix it. The caution is the install path: an AI assistant with ADB access can disable the wrong package, and the README's guardrails only work if the user reads them. Build it, keep the list, and turn debugging off when you are done.

Bottom Line

CleanYourTV is an Android TV and Google TV debloat tool with a small Java replacement launcher. It disables packages over ADB without root or uninstalling anything, records the original state, and provides a restore script. Version 2.5.0 arrived on October 21, 2026, with release notes counting 147 TVs, 49 brands, and 21 languages; the README's contributed list is already larger. It had 15,898 stars and 1,742 forks at the time of check. The launcher is about 405 KB, needs Android 8.0 or newer, and has no prebuilt APK. Fire TV gets debloat only, and ADB should be closed after setup.

Sources

Gathin23/cleanyourtv repository README, including the debloat method, launcher features, supported TVs, restore process, firmware update behavior, safety notes, and licence, verified October 21, 2026.

Gathin23/cleanyourtv releases, including v2.5.0 published October 21, 2026 with 147 TVs, 49 brands, 21 languages, and a 405 KB launcher.

GitHub repository metadata for Gathin23/cleanyourtv, verified October 21, 2026: 15,898 stars, 1,742 forks, 135 open issues, Java, created October 17, 2026, last push October 21, 2026, and a licence field GitHub reports as unrecognised.

Lantern Engine: Bloodborne's Code Runs Native on Windows

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:50 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Lantern Engine is a Windows launcher and engine that runs a copy of Bloodborne on a PC. The project says the game's own code executes natively on the CPU while its graphics are translated to Vulkan. It is built on bbport and shadPS4, then reworked for performance.

The requirement is the part that matters: Lantern Engine ships no game files. The user must supply a decrypted Bloodborne dump, CUSA03173 with update 1.09, and the v0.9.34 release notes add that CUSA00900, the US release, with update 1.09 is fully supported. That makes this a compatibility project for people who already own the game, not a free download. It also puts the project in the same legal gray zone as every other Bloodborne PC effort.

The repository appeared on October 11, 2026, and had 731 stars and 35 forks ten days later. Stable v0.9.36 and beta v0.9.37-beta.1 both arrived on October 21. The developer says updates are slower now because of the hate the project has attracted.

Image: GitHub social preview for the Lantern Engine repository. Image credit: Awakenchan/LanternEngine, via GitHub.

What Lantern Engine is

The README describes Lantern Engine as a launcher and an engine that runs the user's own copy of Bloodborne on Windows. The game's code runs natively on the CPU, and its graphics are translated to Vulkan. That is a different approach from a general-purpose PlayStation 4 emulator, which has to emulate the console's CPU as well as its GPU.

The launcher asks the user to pick the game folder containing eboot.bin, then starts the game. The package also includes Bloodborne.exe, which can be added to Steam as a non-Steam game for a direct launch. The launcher updates itself.

The project credits PC menus and controls, an integrated upscaler, online play through The Hunter's Dream, crash fixes, and performance work on the shader compiler, draw calls, draw pipeline, and memory. Those are the developer's claims. The README does not publish a frame-rate table, a compatibility matrix, or a list of known-broken areas.

Item	Detail from the repository
Project	Lantern Engine
Game	Bloodborne, CUSA03173 with update 1.09, and CUSA00900 per the v0.9.34 notes
Platform	Windows
Language	C#
Licence	GPL-2.0 for the repository; engine GPL-3.0-or-later per the README
Stars and forks	731 stars, 35 forks, 12 open issues at check
Repository created	October 11, 2026
Latest stable	v0.9.36 on October 21, 2026
Latest beta	v0.9.37-beta.1 on October 21, 2026
Game files	None included; user supplies a decrypted dump

What it is built on

Lantern Engine did not start from a blank page. The README credits DarkIzuku's bloodborne_pc as the engine starting point, which is itself built on bbport and shadPS4. It also credits separate work for VRAM cleanup, frame pacing, crash and controller fixes, cheats, online sign-in, occlusion culling, trophy research, and game patches.

The engine's graphics path is Vulkan. The launcher is C# and uses WPF-UI. The credits list Dear ImGui for the in-game menu, SDL's controller database for mappings, LibAtrac9 for sound, and FSR and DLSS components for upscaling. The README does not say which of those are enabled by default or which GPUs are supported.

The project is therefore a fork of a fork with a long dependency chain. That is normal for Bloodborne PC work, but it means a bug can come from the game code translation, the Vulkan layer, a shader compiler, a third-party upscaler, or the launcher. The release notes reflect that: one hotfix fixes a frame-rate drop over long sessions, another fixes a controller tap after closing a dialog, and the beta fixes an OptiScaler compatibility mode that cost frames.

The engine is open source, but the game is not. The launcher is GPL-2.0-or-later, and the README says the engine is GPL-3.0-or-later with licences in the package's licenses/ folder. Those licences cover the project's code, not the Bloodborne assets the user supplies.

What you need

The minimum is a Windows PC and a decrypted Bloodborne dump for CUSA03173 with update 1.09, or CUSA00900 with update 1.09 for the US release. The README does not publish a CPU, GPU, RAM, or storage minimum. It says the graphics are translated to Vulkan, so a Vulkan-capable GPU is implied, but the project does not state a support list.

The download is substantial. The v0.9.36 Windows zip is about 196 MB, and the standalone launcher is about 76 MB. The game dump is much larger, and the user has to provide it. The README does not say whether the launcher is code-signed.

The installation flow is short: download the zip, extract it, run Lantern Engine.exe, pick the folder with eboot.bin, and press Play. The README says the launcher updates itself. There is no installer and no dependency checklist, which is convenient but also means the project assumes a fairly clean Windows environment.

Release pace and fixes

The release cadence is fast. v0.9.34, v0.9.35, and v0.9.36 all landed on October 21, along with a beta. The stable notes focus on hotfixes: a frame-rate decay over long sessions, a crash after long play, missing PC menus with a modded Hunter's Dream, and a controller input bug after closing a dialog.

The beta adds better diagnostics more than new features. It restores frame rate with OptiScaler by limiting the safe mode to the case where OptiScaler's own frame generation is on. It writes a full crash report when the GPU is lost while presenting a frame. It also logs when mods replace gameplay files, because co-op partners with different mods will desync enemies, and it counts network packets once a minute during a session.

That is a project still building its diagnostic foundation. The beta notes are written for people who will report bugs, not for a mass audience. The README carries a note from the developer saying the project is being kept going on his own and that updates are slower now because of the hate the project has been getting. That context matters when reading the issue tracker or asking for support.

Release	Date	Main change
v0.9.36	October 21, 2026	Hotfix for controller input, OptiScaler safe mode, and motion blur with upscalers
v0.9.37-beta.1	October 21, 2026	Restores OptiScaler frame rate, adds GPU crash reports, logs mod desync and packet counts

The legal problem

Bloodborne is owned by Sony Interactive Entertainment and developed by FromSoftware. Lantern Engine is not affiliated with either company. The README does not claim permission, and the project ships no game code or assets. The user must provide a decrypted copy of the game.

That requirement is where the legal risk sits. Making a personal backup of a game you own may be allowed in some jurisdictions and not others. Circumventing encryption or downloading a decrypted dump from someone else can violate copyright law and the platform's terms. The project's open-source licence does not grant rights to the game, and it does not protect a user from a rights-holder claim.

There is also a practical security risk. A project that asks users to run an unsigned launcher and point it at a game dump is a natural target for malicious repacks. The official repository is the only source that should be trusted, and even that trust is not a legal shield. Users should verify the release source and avoid third-party builds.

The project's existence also says something about the demand for Bloodborne on PC. Sony has not announced a PC version. Until it does, projects like this will keep appearing, and rights holders will keep deciding whether to act. That is the unresolved backdrop to every performance improvement in the release notes.

What to watch

The next useful signal is not another hotfix. It is a compatibility list. The README currently says the engine runs Bloodborne, but it does not say which areas, bosses, effects, or hardware configurations are known to work. A public table of tested GPUs, known crashes, and missing effects would do more for users than another beta tag.

The second signal is the online component. The README credits The Hunter's Dream for online servers, but it does not explain authentication, account requirements, or what data leaves the PC. Online play in a fan project raises privacy and security questions that a single credit line does not answer.

The third signal is legal. If Sony sends a takedown, the repository disappears and the builds stop. That risk is not hypothetical for console compatibility projects. Users should treat every release as temporary.

FAQ

Does Lantern Engine include Bloodborne?
No. The repository and releases contain no game files. You need your own decrypted Bloodborne dump for CUSA03173 with update 1.09, and the v0.9.34 notes say CUSA00900, the US release, with update 1.09 is fully supported.

Is Lantern Engine an emulator?
Not in the traditional sense. The project says the game's code runs natively on the CPU while its graphics are translated to Vulkan. It is built on shadPS4 and bbport, but it is not a general-purpose PS4 emulator.

What are the PC requirements?
The README requires Windows and a decrypted Bloodborne dump. It does not publish a CPU, GPU, RAM, or storage minimum, so there is no official support list to check before downloading.

Does it support mods and online play?
The README credits mod-related fixes and online play through The Hunter's Dream. The beta release notes warn that co-op partners with different gameplay mods will desync enemies.

Is Lantern Engine legal?
It is an unofficial fan project with no affiliation or permission from Sony or FromSoftware. Making or using a decrypted dump can violate copyright law or platform terms depending on where you live. The open-source licence covers the code, not the game.

How often does it update?
Frequently. Three stable builds and a beta landed on October 21, 2026 alone. The developer says updates are slower now because of harassment, so the pace may not hold.

Our take: Lantern Engine is technically ambitious and unusually transparent about its dependencies. The problem is not the code. It is the missing compatibility and legal clarity. A 731-star project that ships hotfixes every few hours is moving fast, but users still have no published hardware list, no area-by-area status, and no answer to what happens when a rights holder objects. Bloodborne on PC is a real achievement. It is not yet a supported product.

Bottom Line

Lantern Engine is a GPL-licensed Windows launcher and engine that runs a user-supplied Bloodborne dump by executing the game's code natively and translating its graphics to Vulkan. It is built on bbport and shadPS4 and requires CUSA03173 with update 1.09, with CUSA00900 supported per the v0.9.34 notes. The project had 731 stars and 35 forks ten days after creation, with stable v0.9.36 and beta v0.9.37-beta.1 released on October 21, 2026. It includes no game files, publishes no hardware minimum, and has no affiliation with Sony or FromSoftware. The legal and security risks are real, and the compatibility picture is still thin.

Sources

Awakenchan/LanternEngine repository README, including the engine description, install steps, Bloodborne dump requirement, credits, developer note, and licences, verified October 21, 2026.

Awakenchan/LanternEngine releases, including v0.9.36 and v0.9.37-beta.1 published October 21, 2026, with the OptiScaler, crash reporting, co-op desync, and packet logging notes.

GitHub repository metadata for Awakenchan/LanternEngine, verified October 21, 2026: 731 stars, 35 forks, 12 open issues, C#, GPL-2.0, created October 11, 2026, last push October 21, 2026.

BlackBox: A PS5 Download Manager With 1,796 Game Links

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:48 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: BlackBox is a homebrew storefront and download manager for jailbroken PS5 consoles. Its catalogue file lists 1,796 entries pointing at third-party file hosts, and its own repository contains no application source: only a README, the catalogue JSON, and a small update feed. The latest release, 1.12.3, published on October 21, 2026, fixes a stack buffer overflow that could take the whole payload down during a repair scan.

That combination is the story. The tool is polished in the ways homebrew apps often are, with ten languages, a TV app, an automatic update stream and torrents. The repository behind it publishes a catalogue of links and a checksum feed, not the code that does the work. The README states the project is licensed under GPL-3.0 or later and points to a third-party notices file for the components it builds on.

Image: GitHub social preview for the BlackBox repository. Image credit: D3ATHLY/blackbox, via GitHub.

What BlackBox actually is

BlackBox describes itself as a homebrew storefront and download manager developed for PS5 consoles. The README is unusually explicit about what it is not. It says the project does not host any game files, copyrighted content, or third-party software, that the catalog/ directory contains only game metadata, and that the updates/ directory contains only the application's own update stream.

The repository tree matches that description exactly. It holds the README, a catalog/ folder with six JSON files, and an updates/ folder with three more. There is no source directory, no build script, and no compiler output. GitHub's own metadata for the repository reports no detected language and no detected licence, even though the README states the project is licensed under GPL-3.0 or later.

The application itself ships as release assets instead: blackbox.elf, a payload of about 67.8 MB, and PPSA01453.ffpkg, a TV app package of about 182 MB. The README also says users should not download and use files from the repository directly, because the application updates itself through its own mechanism.

Those points make the project worth writing about, but they also frame the legal question that follows it everywhere. A catalogue of 1,796 commercial game entries, with sizes, formats, cover art and direct download URLs, is not a neutral piece of metadata no matter how carefully the README is worded.

Item	Detail from the repository
Project	BlackBox, a homebrew storefront and download manager for PS5
Repository contents	README, six catalogue JSON files, three update JSON files, no source
Detected language and licence	None reported by GitHub; the README states GPL-3.0 or later
Stars and forks	604 stars, 59 forks, 6 open issues at check
Repository created	October 6, 2026
Latest release	1.12.3 on October 21, 2026
Release assets	`blackbox.elf` at about 67.8 MB and `PPSA01453.ffpkg` at about 182 MB
Languages	Ten, with Turkish listed first
Catalogue size	1,796 entries covering 1,475 distinct titles

The catalogue and where the files live

The catalogue file holds 1,796 entries. Each one carries a title ID, a byte size, a format, a file name, a URL, cover and hero artwork links, a publisher, a release date, and a timestamp for when the entry was added. A separate games.json file carries 678 entries of longer metadata, including descriptions and taglines for a subset of titles, and a revision.json file records the current catalogue revision, which was 48 at the time of check. Two more catalogue files, one for PlayStation 4 titles and one with a second naming scheme, sit alongside them.

Two providers account for the links. The catalogue lists 877 entries pointing at Archive.org and 919 pointing at Vikingfile. By format, 953 entries are PKG packages, 687 are exFAT images and 156 are FFPFSC containers. The sizes add up to roughly 51 TB across the whole catalogue, which is what a full console library weighs.

So the answer to where the files live is simple and worth stating without euphemism: they live on other people's file hosts. BlackBox does not host them, and the README says so. The project is the index and the downloader, and the hosts are where the copies are.

The release notes describe how the tool handles those hosts. Version 1.11.0 added automatic Viking link capture, a mirror domain fallback and multi-connection support, along with a note that the Viking download system may not start on the first attempt and that the user should exit the browser and try again. Version 1.12.0 added torrent downloads in both magnet and .torrent form. The in-app announcement for 1.12.0, which is stored in the repository as a language file, says the release brought torrent downloads, more than 1,900 games, browse sort fixes and translations into ten languages. The catalogue file itself lists 1,796 entries covering 1,475 distinct titles, so the app's own number and the file's number are not the same measure.

The update system and three bugs in two days

The update mechanism is simple and self-contained. A file named updates/payloads.json lists the payload URL, its SHA-256 checksum, its file name and its version, plus the same for the ffpkg package, and a minimumVersion of 1.0.8. A second file, updates/tv-app.json, does the same for the TV app with a minimumVersion of 1.9.0. The payload for the current release, 1.12.3, is listed with a checksum, and the app verifies the download against it before installing.

Four releases landed between October 18 and October 21. The first added features, and the next three fixed something real.

Version 1.12.1, published October 20, fixed consoles that were stuck. The payload had grown to 67.7 MB, which exceeded the old 64 MiB ceiling, so downloads stopped at the limit, the TV app reported that no BlackBox copy could be found, and the update retried on every launch. The fix raised the ELF update ceiling to 96 MiB and staged the companion package even when the ELF lane failed.

Version 1.12.2, published October 21, made PKG games convert automatically to exFAT on firmware 12.00 through 13.60 once the download finished, deleted the old package and left the game playable. It also fixed a background cleaner that could delete the freshly written image mid-job and fail the whole conversion, so that the package is now only spent after a verified dump. The same release finished the translation pass, with every screen message fully translated in all ten languages.

Version 1.12.3, also published October 21, fixed a reopen problem. The scan that repairs the TV app's install record used a 1 MiB stack buffer on a 1 MiB thread stack, and could overflow it, taking the whole payload down so the app looked unreachable and would not reopen. That scan now runs on the heap. The release also made catalogue updates accept feeds with a trailing newline, and let consoles carrying a revision number from the older numbering repair themselves and follow the fresh feed again.

Release	Date	Main change
1.12.0	October 18, 2026	Torrent downloads, firmware detection, corrected internal and USB storage support, 10 language files
1.12.1	October 20, 2026	ELF update ceiling raised to 96 MiB after consoles got stuck with a 67.7 MB payload
1.12.2	October 21, 2026	Automatic PKG to exFAT conversion on firmware 12.00 to 13.60, translation pass completed
1.12.3	October 21, 2026	Install-record repair scan moved off a 1 MiB stack buffer, feed parsing and self-repair fixes

How it installs, and why ShadowMount is involved

The install path is the same one jailbroken PS5 homebrew uses. The README says to delete the old PPSA01453.ffpkg from /data/homebrew, uninstall the previous version, copy the new package into that folder, and then use shadowmount to trigger and install it, rebooting if there are warnings. A shortcut to run ShadowMount was added to the Downloads menu in version 1.12.0, and the release notes recommend it whenever a game fails to install after a download.

There is one warning in caps in every recent release, and it is about process. The blackbox.elf file is embedded inside the ffpkg package and updates itself. Loading the ELF manually through a payload manager is not supposed to happen, because the two paths conflict.

Firmware support is versioned in a way that matters for the PKG format. The release notes for 1.12.0 state that downloading PS5 .pkg files works on firmware 11.60 and earlier, that on firmware between 12.00 and 13.60 PKG downloads are not available, and that exFAT and FFPFSC downloads still work without issues on those versions. Version 1.12.2 then added the automatic conversion that makes the second path viable.

The legal line, stated plainly

The README carries a long disclaimer, and it is worth reading closely rather than skimming. The project says it does not host any game files and that all file downloading and installation is at the user's own risk. It says the tool is intended exclusively for content to which the user has legal access rights, that it does not guarantee the security, legality, or copyright compliance of files accessed from third-party sources such as Archive.org and Vikingfile, and that this responsibility belongs entirely to the user. It also says the project is not designed for unauthorized access, piracy, or illegal activities.

That disclaimer describes the position the project wants to hold. It does not change what the catalogue contains. The entries are commercial console games, the URLs point at file hosts that host them, and in most jurisdictions making and using an unauthorized copy of a commercial game is copyright infringement regardless of which application fetched the file.

There is also a platform risk that has nothing to do with copyright law. BlackBox needs a jailbroken PS5, and the ShadowMount tool it tells you to use is the same kind of homebrew payload, one whose own README warns that mounting images can cause shutdown problems and data corruption on internal drives. A console running homebrew is also a console where uninstalling a game through system settings can delete its saves, which is exactly the warning ShadowMountPlus prints in its troubleshooting section. The README's own content and liability table is a useful summary of that posture: no warranties, no liability for data loss, and the user agreeing to follow applicable law and terms of service by using the tool.

The interesting engineering story here, the one worth following, is the update system. A self-updating homebrew app with checksum-verified payloads, a ten-language TV interface, live conversion of one image format to another across a firmware generation boundary, and three consecutive releases that fixed a stuck update lane, a mid-conversion deletion race and a stack overflow, is a well-built piece of software. That is worth saying, and it is separate from whether the catalogue should be there.

FAQ

What is BlackBox?
It is a homebrew storefront and download manager for jailbroken PS5 consoles, distributed as an ELF payload plus a TV app package. The catalogue and the update feed live in the GitHub repository.

Does BlackBox host the game files?
No, per its own README. The repository holds metadata only, and the catalogue links point at third-party file hosts, 877 entries at Archive.org and 919 at Vikingfile at the time of check.

Is the source code available?
Not in the repository. The tree contains the README, six catalogue JSON files and three update JSON files, and GitHub detects no language for the project. The README states the project is licensed under GPL-3.0 or later, and points to a third-party notices file for the components it builds on.

Which PS5 firmware can install PKG games?
Per the 1.12.0 release notes, PKG downloads work on firmware 11.60 and earlier. On firmware 12.00 through 13.60 they are not available, but exFAT and FFPFSC downloads still work, and version 1.12.2 added automatic conversion from PKG to exFAT on those firmwares.

How does it update itself?
Through the app. The update feed lists the payload and TV app URLs with SHA-256 checksums and a minimum version, the app verifies the download, and it is installed as a whole. Manually loading the ELF through a payload manager conflicts with that path.

Is using BlackBox legal?
That depends entirely on where the files come from. The tool itself declares that it hosts no game files and puts responsibility for legality on the user, but copying commercial games without the right to do so is copyright infringement in most countries, whatever application downloaded the file.

Our take: BlackBox is a better piece of software than its repository layout suggests, and the update system is the part that deserves attention. A checksum-verified self-update path, a firmware-aware conversion pipeline and a stack overflow fixed in a day are all signs of an active maintainer. The problem is not the engineering. It is that a catalogue of 1,796 commercial games with direct links to file hosts, in a repository that publishes no source, is a distribution operation wearing a metadata file's clothes. Whether you install it is your call and your legal exposure, not ours.

Bottom Line

BlackBox is a GPL-3.0-or-later homebrew storefront and download manager for jailbroken PS5 consoles. Its repository publishes a catalogue of 1,796 entries covering 1,475 titles and roughly 51 TB, split between 877 Archive.org links and 919 Vikingfile links, plus a checksummed update feed, but no application source. The latest release, 1.12.3, published on October 21, 2026, moved a 1 MiB stack buffer to the heap after it could take the payload down, following two releases on October 20 and 21 that raised the update size ceiling and added automatic PKG to exFAT conversion on firmware 12.00 through 13.60. It needs a jailbroken PS5 and the ShadowMount tool to install, ships in ten languages, and puts copyright responsibility on the user according to its own disclaimer.

Sources

D3ATHLY/blackbox repository README, including the project description, install steps, disclaimer, content and liability table, and licence statement, verified October 21, 2026.

D3ATHLY/blackbox releases, including 1.12.0, 1.12.1, 1.12.2 and 1.12.3 published between October 18 and October 21, 2026, with the update ceiling, conversion, translation and buffer fixes.

The repository's own catalog/catalog.json, catalog/games.json, catalog/revision.json, catalog/catalogue-ps4.json, catalog/catalogue-v2.json, catalog/releases.json, updates/payloads.json, updates/tv-app.json and updates/messages.json files, for entry counts, providers, formats, sizes, checksums and the in-app release announcement, verified October 21, 2026.

GitHub repository metadata for D3ATHLY/blackbox, verified October 21, 2026: 604 stars, 59 forks, 6 open issues, no detected language or licence, created October 6, 2026, last push October 21, 2026.

wotlk-browser: A WotLK Client in a Browser With Your Own Game Data

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:46 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: wotlk-browser is a same-day project that serves a World of Warcraft 3.3.5 client to a browser and tunnels the game's TCP connections to a real server. The repository was created on October 21, 2026, and by the time of check it had 3 stars, 1 fork, no releases and no open issues. It ships a WebAssembly client, a Node bridge, a MySQL account backend, and a README whose account manager section is headed "Account Manager (Untested)".

That last detail is the one to read twice. The project does not pretend to be finished. It also does not pretend to be a hosted realm: no realm server is included, and no play button leads to one. What it is, per its own README, is a delivery and connectivity layer. You slice your own game files, serve them over HTTPS, and point the tunnel at a server you already run.

Image: GitHub social preview for the wotlk-browser repository. Image credit: Pucur/wotlk-browser, via GitHub.

What the project actually is

The repository description is "World of Warcraft Client in Browser", and the README opens by calling it a browser-based delivery and connectivity layer for World of Warcraft: Wrath of the Lich King. The feature list names the pieces: web client delivery over HTTPS, game data streaming with block-based delivery and HTTP byte-range requests, a WebSocket-to-TCP tunnel, configurable routing, account management, session management, MySQL integration, diagnostics and logging, and environment-based configuration.

The architecture table splits the system into four parts. The web client is the browser-side code and its resources. The bridge server handles HTTPS file delivery, game data streaming and the WebSocket-to-TCP tunneling. The account backend handles authentication, registration and sessions. A MySQL-compatible database stores account and session data.

Nobody publishes a player count for this project, because it is not a realm. It is the plumbing a realm needs in order to live in a browser tab. The README is explicit about the prerequisites: a compatible Node.js runtime, the project dependencies, the necessary client files and game data, a configured MySQL database, valid TLS certificates for HTTPS deployments, and access to a compatible WotLK game server.

Item	Detail from the repository
Project	wotlk-browser, a browser delivery and connectivity layer for WotLK
Language	JavaScript, Node.js with ES modules
Licence	MIT
Stars and forks	3 stars, 1 fork, 0 open issues at check
Repository created	October 21, 2026, and last pushed the same day
Releases	None published
Dependencies	Express 5, ws, mysql2, pngjs
Client build	A WebAssembly client in `web/Wow.wasm` at about 8.7 MB
Game data	The user's own files, sliced into 64 KB blocks
Account backend	MySQL with SRP-style login, headed "Account Manager (Untested)"

The client and the 17.8 GB of data you supply

The browser client is a WebAssembly build. The repository carries web/Wow.js at about 323 KB and web/Wow.wasm at about 8.7 MB, plus a loading font of about 62 KB, four loading images between 1.09 MB and 1.25 MB each, and a small art manifest. The JavaScript glue carries the marks of an Emscripten build, with references to SharedArrayBuffer and threads.

That binary is the smallest part of the job. The game data comes from the user. A manifest in the repository lists 195 data files with their sizes, and the total of those sizes is about 17.8 GB. The largest entry is Data/patch.MPQ at about 4.01 GB, followed by Data/lichking.MPQ at about 2.55 GB, Data/common-2.MPQ at about 2.44 GB and Data/common.MPQ at about 2.24 GB. Further down sit the English expansion speech archive at about 241 MB and the English backup archive at about 166 MB.

The README describes the pipeline in one line: the slicer cuts the enUS build into 64 KB chunks, written from the game folder into the blocks folder. The script behind it reads the manifest, matches each entry against the game directory, and cuts every file into chunks of 64 KB. The bridge then serves those chunks, and the page fetches the ones the client asks for.

Two prefetch lists sit alongside the manifest, one for the data the client reads on its way to the login screen and one for the data the client reads entering the world. The page fetches them in the background and warms the browser cache while the WebAssembly module downloads. A comment in the page notes that a stale list costs bandwidth rather than correctness, which is an honest way to describe a cache warm-up that can drift.

The bridge and its redirect default

The bridge is a Node program called bridge.mjs, and its configuration is entirely environment variables. The defaults in the source are a port of 443 and a bind address of 127.0.0.1, with the web directory, the game data directory, the statistics directory, the allowed realm hosts, the allowed realm ports, and the redirect rules all read from the environment.

The allowed ports are 3724 and 8085, the two a WotLK realm needs, one for authentication and one for the world connection. The allowed hosts default to 127.0.0.1, and the README's own examples show a comma-separated list of hosts the tunnel will accept, plus per-port redirects that send every connection on a port to a fixed address instead of the host the client asked for.

There is a default worth knowing about. The source sets the 3724 redirect to a fixed address, 51.254.7.68, unless the environment overrides it. The comment block above the code shows a different example value, and the README's second example shows a second redirect on 8085. Anyone deploying this should set the redirect variables deliberately rather than accept the shipped default, because an unnoticed redirect points realm traffic at a host the operator did not choose.

The bridge also sets the headers that make the WebAssembly client work. It sends Cross-Origin-Opener-Policy: same-origin and Cross-Origin-Embedder-Policy: require-corp, because the client's threads need SharedArrayBuffer, which browsers only expose to a cross-origin isolated page. It also sends Cross-Origin-Resource-Policy: same-origin and X-Content-Type-Options: nosniff, and it resolves game file paths case-insensitively while refusing to traverse above the game directory.

The account manager the README calls untested

The account section of the README carries a heading that is unusual in a project README: "Account Manager (Untested)". Anyone writing about this project should quote that heading rather than paraphrase it. The section below it describes a MySQL-backed account service with the usual endpoints for registration, a challenge, a login, a password change, an email change and account information, and it says responses are plain text, one value per line, with "OK" on success.

The login flow is a challenge and response, in the shape of the protocol the game itself uses. The client sends a username and receives a salt and the server's public value. The client computes its own value and a proof from the password, the salt and the server's value, and sends them back. On success the server returns a session token, which the other endpoints take. The code confirms the shape: a large fixed prime, a small generator, a multiplier constant, and SHA-1 helpers, all in one module.

Errors are as plain as the successes. The README shows a status code with a single-line reason, such as a wrong-password response or a username-already-taken response. The server module adds the details the README does not: the database pool, a challenge lifetime of 60 seconds, a session lifetime of 24 hours, and a ban table with its own start and end dates. A banned account is refused at login, and the account information endpoint reports the ban state as part of its output.

The same module records the things a private realm's operator needs to see. It tracks the last login time and the last IP for an account, the number of failed logins, and the expansion the account is set to, and it flags whether the account is banned. The admin page, which is a separate program, reads the event and session files the bridge writes and renders page loads, browser families, load failures by reason, and world sessions with their device, average session length and average frame rate.

The admin program requires an admin password, takes HTTP basic authentication, and listens on port 8081 by default. Its host default is 0.0.0.0, which is worth changing if the machine is not on a trusted network.

What the browser client demands

The page that loads the client is unusually explicit about its own limits, which is the most useful thing in the repository. It reports the reason a load failed rather than showing a blank screen.

WebGL 2 is required, and the page says so in plain words with a suggested fix: use a recent Chrome, Edge or Firefox on a computer. The page also refuses to run inside the built-in browsers of social apps, and it explains why in the code comment: those apps do not give the page a cross-origin isolated context, which the client's threads require. The page offers a link to open the realm in a real browser instead.

Phones are treated separately, with a lower memory ceiling of 2048 MB, a smaller game-data block cache, and a service worker so the realm can open without a connection. Opening the page with a stats parameter adds an overlay showing the frame rate, the WebAssembly heap size in megabytes and the viewport, which is a small sign the author debugs on real devices.

There is no shop and no browser-side account form. The page's own description says you create an account on the login screen, which is the native client's own screen, and the page's code comments point at the client's account code by name. Session state beyond the browser cache lives in the MySQL backend, not in the page. And the launcher download link that the Conquest of AzerothCore front page points at is still marked as coming soon, which is not this project's problem to solve: this repository is the layer underneath, not the realm in front of it.

FAQ

What is wotlk-browser?
A delivery and connectivity layer for World of Warcraft 3.3.5. It serves a WebAssembly client and the game data over HTTPS and bridges the browser's WebSocket connections to a real realm server over TCP.

Do I need my own copy of the game?
Yes. The manifest covers 195 data files worth about 17.8 GB, and the README says the slicer works on the enUS build from the user's own game folder. The repository ships no game data.

Is the account manager ready to use?
The README heads that section "Account Manager (Untested)". It describes registration, a challenge and response login, password and email changes, and account information, and the code adds session expiry, ban checks and failed login tracking. Read the heading as written.

What does the client need from a browser?
WebGL 2, a cross-origin isolated page for the client's threads, and a recent Chrome, Edge or Firefox on a computer. The page refuses to run inside social apps' built-in browsers and tells you to open it in a real browser.

Is there a server included?
No. The prerequisites list access to a compatible WotLK game server. This project is the layer that puts a client in a tab; it is not a realm and it does not run one.

What should I check before deploying it?
The realm host and port allowlists, and the redirect variables. The shipped source sets the 3724 redirect to a fixed address unless the environment overrides it, so a deployment that does not set it deliberately sends realm traffic somewhere the operator did not choose.

Our take: The interesting part is the honesty about the limits. A project with 3 stars and a same-day commit history that tells you its account manager is untested, names the exact browser requirements in the page, and ships a size manifest covering 17.8 GB of data you already own is a better-engineered starting point than most things with a launch trailer. The one thing to fix before you point it anywhere is that default redirect.

Bottom Line

wotlk-browser is an MIT-licensed Node.js project created on October 21, 2026 that puts a World of Warcraft 3.3.5 client in a browser tab. It serves a WebAssembly client and the user's own sliced game data over HTTPS, bridges WebSocket traffic to a realm's TCP ports, and ships a MySQL account backend whose own README section is headed "Account Manager (Untested)". It had 3 stars, 1 fork, no releases and no open issues at the time of check. It needs a WotLK 3.3.5 client the user already owns, a MySQL database, TLS certificates, and a realm server to reach, and its page demands WebGL 2 and a cross-origin isolated browser.

Sources

Pucur/wotlk-browser repository README, including the features, architecture, prerequisites, install steps, security notes and the account manager section, verified October 21, 2026.

The repository's own source files, bridge.mjs, server.mjs, admin.mjs, srp.mjs, web/index.html, web/game-manifest.txt and web/slicer.py, for the default configuration, headers, session and ban handling, the browser requirements, the 195-entry manifest and the 64 KB slicer, verified October 21, 2026.

GitHub repository metadata for Pucur/wotlk-browser, verified October 21, 2026: 3 stars, 1 fork, 0 open issues, JavaScript, MIT, created October 21, 2026, last push October 21, 2026.

Conquest of AzerothCore: A Public Test Realm That Publishes Its Own Bug Count

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:43 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Conquest of AzerothCore is a public bug-test realm for the AzerothCore emulator core, and its website treats bug fixing as the product. The site's own tracker API reports 3,859 fixes against 6,983 reports since September 7, 2026, with 1,300 commits, 1,254 pull requests and 56 contributors. The GitHub repository behind it, jealous-sound/azerothcore-wotlk-coa, had 272 stars and 193 forks at the time of check, and GitHub counted 1,564 open items against it once its open pull requests are included.

What makes it unusual is the disclosure. The front page says, in its own words, that this is testing only, that accounts and characters and progress can be wiped at any time without notice, and that it is not a permanent or live server. The site publishes a live changelog written from the development commits, a roadmap with per-day fix counts, a mirror of the GitHub issue tracker, an Armory that ranks characters by class, and a talent calculator for every class. It is a private server project that shows its own bug backlog as a feature.

Image: the Conquest of AzerothCore social card. Image credit: coa-development.org.

What the site actually is

Conquest of AzerothCore describes itself as a public bug-test realm for the AzerothCore emulator core. AzerothCore is the open-source server emulator that the large World of Warcraft private server community runs on. This project takes that core, adds a large layer of custom content on top, and then runs the result as a place where players are expected to break it.

The site's own warning is the clearest statement of purpose anywhere in the project, and it is worth reading in full rather than skimming:

Conquest of AzerothCore is a public bug-test realm for the AzerothCore emulator core. It exists purely to help find and fix bugs before changes ship. It is not a permanent or "live" server. Because of that, accounts, characters and progress on this realm can be wiped or reset at any time, without notice, whenever it's useful for testing. Please don't treat anything you build here as permanent.

That is a deliberate trade. The team gives up the retention that makes a private server worth joining and gets something more valuable in exchange: a large, motivated group of testers who are already logged in and already have characters. A confirmed bug report counts towards a Bug Hunter badge in the site's Armory, and the top testers panel ranks players by confirmed and fixed reports, with a currency called Groot Coins attached to the leaderboard.

Behind the site sits a fork of AzerothCore maintained under the GitHub account jealous-sound. The repository description is simply "Conquest of AzerothCore", the language is C++, and the licence is GPL-2.0. The account also hosts smaller pieces: a volumetric fog and light shaft project for the 3.3.5a Direct3D 9 client with 24 stars, a bug-report helper, and forks of the playerbots module and a project called Ascension Extensions that reconstructs a DLL function by function.

Item	Detail
Project	Conquest of AzerothCore, a public bug-test realm for the AzerothCore core
Repository	jealous-sound/azerothcore-wotlk-coa, C++, GPL-2.0
Stars and forks	272 stars, 193 forks, and 1,564 open issues and pull requests at check
Repository created	September 7, 2026
Last push	October 21, 2026
Client	World of Warcraft 3.3.5a, build 12340
Realms listed	Conquest of Azeroth, Wildcard, Warcraft Reborn, Area 52, Freepick, Hero
Test policy	Accounts, characters and progress can be reset at any time

The numbers, straight from the API

The site publishes its own metrics as JSON, which makes verification unusually easy. The roadmap API reports the following totals at the time of check.

Metric	Value
Fixed	3,859
Reported	6,983
Still open	1,516
Closed as other	1,608
Commits	1,300
Pull requests	1,254
Contributors	56
News posts	47
Days tracked	45, from September 7, 2026
Best single day	461 fixes on September 21, 2026

The bug API adds the split between the two trackers. It reports 1,517 confirmed bugs, 3,860 fixed and 355 fixed in the last week. On the GitHub side it reports 1,516 open issues, 3,794 fixed and 600 listed, and it links most website reports back to the GitHub issue they came from, with two placeholder entries that carry no link.

The changelog is generated from the development commits, one entry per day, and the pace is visible in the feed. October 20 recorded 57 changes, October 19 recorded 63, October 18 recorded 76, and October 17 recorded 51. The largest day in the feed is October 7 with 141 changes. The entries are not release notes in the marketing sense. They read like a work log, and the October 21 entry was still filling itself in as we checked, adding tooltip text fixes on the Starcaller and a fix that lets Lunar Cascade mark full-health allies.

Not every entry is a bug fix. A standalone October 17 post announced Manastorm Endless, an SQL-driven endless dungeon mode with floor-by-floor progression, gadgets, affixes and depth-scaled rewards, narrated by Millhouse Manastorm. It requires level 10 or above, works solo or with a party of up to 5, and is entered with a command or through a native queue in the client.

The website as a bug dashboard

Most private server sites are marketing pages. This one is closer to a status page with a game attached.

The realm status page polls every 30 seconds and keeps 30 days of history, with a stated rule that a break of up to 15 minutes counts as a restart and anything longer counts as an outage. It tracks the login server separately from the realms, because a realm can be up while nobody can log in. When the site was checked, three realms answered, with player counts moving minute to minute: CoA with 31 players, Warcraft Reborn with 10 and Wildcard with 3. The site states plainly that bots and game masters are not counted.

The changelog page carries filters for realm, for class, and for fix versus new, with a class picker that filters the whole history down to one class's changes. The roadmap page archives every week and every month, with a running total of fixes, and it keeps two hours of finished periods as they were so later edits to the changelog do not rewrite history.

The known bugs page mirrors the GitHub tracker and adds a website-reported layer on top, with a search box and filters for all bugs, not fixed and fixed. The Armory looks up any character on the realm and shows gear, stats, professions and achievements, and its leaderboard API returns the top character per class at a maximum level of 60, along with achievement counts, top professions, guild sizes with average levels, and the top bug hunters with their report counts. The talent calculator is per class, with one specialization and the class tree, and it produces a shareable link.

Account features are the sort a hosted service would have and a home-kept server usually would not: two-factor authentication with an authenticator app, recovery codes, a bug reporting panel inside the account page, and a message when the realm still thinks an account is online after a crash.

The class roster and where it came from

The site's own metadata advertises 21 custom classes, and the class roster echoes the 21-class Conquest of Azeroth roster this site covered on October 17, with the same named classes such as the Necromancer, the Chronomancer and the Runemaster. The two projects are run by different maintainers, and this is a different deployment: the October 17 coverage described a Phase 1 launch with a level 20 cap, while this is a bug-test realm for the AzerothCore core, running three live realms and a 60 level cap.

The site ships a class map in one of its JavaScript files, and it says where it came from: extracted from the custom client DBCs for characters, classes, races and areas, and regenerated when those change. That map holds 32 class IDs in total. Ten of them carry the stock 3.3.5a names, from Warrior through Druid, and the remaining 22 are custom, running from Hero and Barbarian through Witch Doctor, Felsworn, Witch Hunter, Stormbringer, Knight of Xoroth, Guardian, Templar, Bloodmage, Ranger, Chronomancer, Necromancer, Pyromancer, Cultist, Starcaller, Sun Cleric, Tinker, Venomancer, Reaper, Primalist and Runemaster, each with its own colour. That is one more custom class than the site's own 21 headline figure, which is the kind of gap a changing roster produces.

The client also carries extra races in the same map, including Zandalari Troll, Worgen, Kul Tiran, Vulpera and Pandaren, plus a long tail of custom identifiers such as Thin Human, Fel Orc, Naga, Broken, Skeleton, Vrykul, Tuskarr, Forest Troll, Taunka, Northrend Skeleton, Ice Troll, Drakkari Troll, Leper Troll and Troglodyte.

The changelog shows what a class roster like this costs in maintenance. The October 21 entry alone carries Starcaller fixes, including tooltip text for the Scattered Stars effect in Aspect tooltips, Lunar Cascade marking full-health allies, an explanation of Eclipsing Arrows inside Stellar Convergence, restoring the visible healing area of Elune's Cradle, and keeping the Nightborne Armaments buff off above 35 percent health. The October 20 entry, all 57 changes being fixes, ranges across vendor gossip, creature spawns, a Crossroads dungeon, item vendor scaling, and a set of class-specific corrections.

Why a test realm tells you to expect a wipe

The FAQ is unusually blunt about the deal. It says to expect bugs, restarts and changes, and that restarts are announced in the site banner and on Discord. It says the launcher is not signed with a code-signing certificate, so SmartScreen will warn, and that antivirus products that flag file-patching launchers are a common false positive, with instructions for restoring the file and for excluding the game folder rather than switching protection off. It says the game needs a Wrath of the Lich King 3.3.5a client, build 12340, and that other versions cannot log in.

It also says that if a game is still downloading from the launcher, the client files may not match the realm, so the launcher should be run again, the cache folder deleted, and addons disabled for a moment to rule them out. The site also warns that school, work and hotel networks block game ports, and that a firewall rule for the game's executable may be needed.

One detail stands out: the front page still says the launcher download link is coming soon, even though the FAQ and the step-by-step guide both point at a launcher. Anyone planning to test this needs to check whether the download is live before they start.

FAQ

What is Conquest of AzerothCore?
A public bug-test realm for the AzerothCore World of Warcraft server emulator, run from the GitHub fork jealous-sound/azerothcore-wotlk-coa. It exists to find and fix bugs before changes ship to the wider private server community.

Can I keep my character there?
Do not plan around it. The site says accounts, characters and progress can be wiped or reset at any time without notice, whenever it is useful for testing, and tells players not to treat anything they build as permanent.

What client does it need?
World of Warcraft 3.3.5a, build 12340, per the FAQ. The launcher adds the project's files on top of that client, and other client versions cannot log in.

How do I report a bug?
Through the Bug reports panel in My Account, or as an issue on the GitHub tracker. Reports the team confirms or fixes count towards the Bug Hunter badge in the Armory.

Is my bug already known?
Each class page lists the open issues for that class, the changelog records what was fixed, and the known bugs page mirrors the full GitHub tracker with filters for all bugs, not fixed and fixed.

Does it cost anything?
The site does not publish a price. The related CoA Server Manager project this site covered in October was community tooling for a private server, and this repository is licensed GPL-2.0.

Our take: Publishing your bug count is unusual and it cuts both ways. It proves the project is doing real maintenance work at a real pace, and 3,859 fixes in 45 days is not a rounding error. It also tells any prospective player that this is a laboratory, not a destination, and the wipe policy makes that unambiguous. For a community that has spent years watching private servers promise permanence and deliver chaos, that honesty is worth more than a player count.

Bottom Line

Conquest of AzerothCore is a GPL-2.0 C++ fork of the AzerothCore emulator core, maintained at github.com/jealous-sound/azerothcore-wotlk-coa, that runs a public bug-test realm for its own project on a site that publishes its own bug backlog. Its tracker API reported 3,859 fixes against 6,983 reports since September 7, 2026, with 1,300 commits, 1,254 pull requests, 56 contributors and a best day of 461 fixes. The repository had 272 stars and 193 forks at the time of check, and the site's mirror of the GitHub tracker listed 1,516 issues still open. The site advertises 21 custom classes, mirrors the GitHub tracker, and states plainly that accounts and characters can be reset at any time.

Sources

The Conquest of AzerothCore website at coa-development.org, including the front page, the FAQ, the known bugs page, the roadmap, the changelog and its RSS feed, verified October 21, 2026.

The site's own JSON endpoints for server status, bug tracking and the roadmap, and its class map script, for the totals, the per-realm player counts, the leaderboards and the class and race names, verified October 21, 2026.

GitHub repository metadata for jealous-sound/azerothcore-wotlk-coa, verified October 21, 2026: 272 stars, 193 forks, C++, GPL-2.0, created September 7, 2026, last push October 21, 2026.

decomp.anything: A Fork Keeps a Discontinued AI Decompiler Alive

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:41 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: decomp.anything is a fork of decomp.wtf, the AI-driven matching decompiler whose author shut it down and archived it after the decompilation community pushed back. The upstream repository, i12bp8/decomp.wtf, had 531 stars and 294 forks and was archived with its last version tagged v0.10.6. The fork, giveen/decomp.anything, was created on October 19, 2026, and extends the same engine to Windows executables and Linux binaries rather than only game images.

That makes it a different kind of project from the one that was discontinued. The upstream author's stated reason for stopping was that much of the community was against AI decompiling of that kind, and he asked anyone who used it or built on it to respect the rules of the projects they worked with. The continuation keeps the MIT licence, keeps the byte-exact rule, and says in its own README that it never sends anything upstream.

Image: decomp.anything banner. Image credit: giveen/decomp.anything, via the repository.

The fork and the shutdown it came from

The upstream project, decomp.wtf, is an application that hands a game to an AI agent one function at a time. The agent writes C or C++, the original compiler builds it, and the result is compared byte for byte. The scoring is unforgiving: a score of zero means the bytes match, and there is no partial credit.

Its author stopped work on it and archived the repository. The README says so in its first line, and the reason is stated plainly: over a few days he noticed that much of the decompilation community, the people who made the tools and projects everything is built on, was against AI decompiling of that kind. He writes that he did not make decomp.wtf to do harm, that he respects their opinion, and that the code stays up under the MIT licence for anyone who wants to fork it and keep going. He adds that he will not fix issues, review pull requests or make releases, and that anyone using it or building on it should only use it on games they own, should not share the game code it produces, and should respect the rules of the projects they work with.

GitHub's metadata for the archived repository records 531 stars, 294 forks, a Python codebase, an MIT licence, a creation date of October 13, 2026 and a last push of October 15, 2026, with tags running from v0.10.1 to v0.10.6 and no release objects attached.

The fork, giveen/decomp.anything, was created on October 19, 2026 and last pushed on October 20, 2026. It had 22 stars, 3 forks, no open issues and no releases at the time of check. The repository tree holds 268 files, with 127 under the source directory, 66 under projects and 53 under tests.

Item	Detail
Upstream	i12bp8/decomp.wtf by Sam Wolfs, archived, 531 stars, 294 forks, last version v0.10.6
Continuation	giveen/decomp.anything, created October 19, 2026, an independent fork rather than a GitHub network fork
Fork size	22 stars, 3 forks, 0 open issues, no releases at check
Language and licence	Python, MIT
Platform	Linux x86-64, or Windows through WSL 2. macOS is not supported
Requirements	Python 3.12 or newer, uv, plus Node.js for npm-based agents and REA, and Java 21 or newer for Ghidra
Network	Local-first: no account, no telemetry, nothing uploaded

What byte-exact matching means here

The workflow is the part worth understanding, because it is the claim that makes the tool different from an ordinary reverse engineering tool that hands you a plausible decompilation.

The target is split into functions. splat and spimdisasm do that for MIPS games, and the PE reader does it for Windows programs using the exception directory, the COFF symbol table, exports, the entry point and direct call targets. The Linux reader uses the symbol table, or the unwind tables in a stripped program. Other tools cover GameCube, Wii, DS, GBA and PlayStation targets.

Agents then claim functions, and each one receives the target assembly, the data it touches, its callers, a machine-made draft and notes from earlier attempts. Every attempt is compiled with the target's own compiler and linked at the function's real address, so calls, globals, strings and jump tables are verified too, and a miss returns an instruction diff that goes back to the agent. A hit is saved, named and shared with the other agents.

The README's own worked example shows the standard. A short comparison function from the target compiles to a small branch sequence, and the agent's C, a clamp with two comparisons, compiles to exactly those bytes with KMC GCC 2.7.2 at optimization level two. The matchers cover MIPS, PowerPC, ARM, x86 and x86-64.

Two features sit under that. The autopilot takes free matches before any agent spends a token, first from an identical function already matched elsewhere, then from a machine draft run through the judge. The permuter rewrites near misses in the background, and the README puts its threshold at 95 percent and up, changing statement and declaration order, casts and temporaries. The permuter changes what to try, never what counts as a match.

What the fork adds

The continuation keeps the upstream engine, the project workflow, the app and the installers, and its README lists six things it adds on top.

The first is Windows and Linux targets. Windows executables and DLLs, 32-bit and 64-bit, import and match with a relocation-aware x86 matcher, and Linux x86 and x86-64 executables and shared libraries import the same way. The matching path compiles the agent's C to COFF and compares bytes with address fields blanked on both sides, so a call or a RIP-relative load has to point at the same function or global. For Linux, the compiler is taken from what the binary's own comment section says built it, and position-independent programs compile with the position-independent flag set. The README is clear that only C and C++ are matched byte for byte: Rust and Go binaries are identified by their compiler markers, but not matched.

The second is a console for REA, the reverse engineering tool that this site covered in October. The fork describes it as a graphical front end for REA's tool catalog, opened from a procedure work item or from a PC function. It browses native binaries, Android apps, .NET assemblies, JavaScript and Electron apps, web pages, firmware and plain files, and it shows functions, pseudocode, strings, flow graphs and call graphs, plus a generated form for every REA tool. It runs its own REA process so it never switches the target the agents are reading, tools that run programs or write files ask for confirmation, and everything REA shows about a target is treated as untrusted text. Agents can ask the same service for pseudocode, signatures and callers.

One of the six is a console for opencode 2, where the app writes its own V2 configuration, turns Code Mode off and runs the binary in standalone mode, and the other is a set of reference folders, so helpers can read and search a matched decomp, notes or symbol lists by path, with reads confined to that folder.

Separate from that list, the README also documents archive support. Drop in a ZIP or TAR and every regular file inside becomes a work item, with a contents panel that previews the tree. A binary or nested archive can be opened as a linked child project, and findings from the child roll up into the parent's summary. The extraction limits are generous for your own files and stated plainly: 2 GiB of archive input, 20,000 entries, 1 GiB uncompressed, a 1000 to 1 per-entry compression ratio, 256 MiB per entry and 512 MiB extracted per project.

The README also documents a set of quality-of-life pieces that are useful outside the game world: an intake report written beside each project that records what the import already established, including the format, the compiler and flags with the evidence for them, the debug information, the function and string inventory, subsystem hints and the questions still open, and which agents are told to read it first. It adds version comparison, which pairs functions across a patch, demo or port by normalized instructions and reports same, moved, changed, added and removed, with a similarity score for the changed functions. It adds runtime coverage import from a debugger or emulator trace, which maps hit counts onto functions and gives agents a way to find the hot code. It adds reference folders, so helpers can read and search a matched decomp, notes or symbol lists by path, with reads confined to that folder.

Which agents drive it

The application talks to agents over the Model Context Protocol, and the fork lists Claude Code, Codex, the Antigravity CLI and opencode among the ones it drives headlessly. For opencode 2 specifically, the app writes its own V2 configuration, turns Code Mode off and runs the binary in standalone mode.

There is also a built-in loop for any OpenAI-compatible API, which covers DeepSeek, OpenRouter and OpenAI, as well as a model running on your own machine through llama.cpp, Ollama or LM Studio. For local models the README asks for at least 16K tokens of context, and says small models should be given small functions and big models the hard ones.

The tool boundary is deliberate. Agents get decomp.anything's own tools and nothing else: no shell, and no access to files beyond a fork's own source tree and any reference folders the project names. They write only C and C++ sources and headers, never targets, assembly or build scripts, and inline assembly is refused. A change that breaks an already-matched function is undone, and the project's file layout and build wiring are changed by the application itself, never typed by an agent.

For a team, one computer hosts the room and hands out functions so nobody duplicates work, checks each match against its own copy, and shares names, structs and notes. No original bytes or assembly cross the wire, because every member reads the code from their own copy. The address can be a free Cloudflare quick tunnel, your own LAN, Tailscale or a server.

Where the fork draws its line

The README has a section on responsible use, and it is worth reading in full because it widens the upstream rules rather than replacing them. The upstream notice said decomp.wtf contains no game code and provides no ROMs, and asked users to stick to games they own, to avoid getting around copy protection, and not to publish decompiled code of games whose rights they do not hold. The continuation keeps each of those and generalizes them: it contains no game or application code, provides no ROMs, executables or firmware, asks that you only analyze software you own or are authorized to analyze, and says not to publish decompiled code whose rights you do not hold.

It also says that many community decomps do not accept AI-written contributions, and that the ones that do expect a person to stand behind every line. The continuation never sends anything upstream on its own: the agents' work stays on your local branch, and before opening a pull request you are told to read the project's contribution rules, review the code yourself, and say how it was made.

That is a coherent position for a project that exists because its author asked for it. The upstream notice asked anyone continuing the work to keep the MIT licence and to respect the rules of the projects involved. The continuation kept the licence, kept the byte-exact standard, and wrote the rules down for software rather than games alone, since the byte-exact path now covers PC binaries as well.

The remaining question is the one the whole scene is arguing about, and the continuation does not settle it. Whether AI agents should be allowed to produce matching decompilation at all is a question for the people who own the projects, and several of them have said no, as this site reported in October. What the continuation does is make the case that the output quality can be judged mechanically, byte for byte, and leave the decision to the project owner.

FAQ

What is decomp.anything?
An MIT-licensed fork of decomp.wtf, a matching decompilation workbench where an AI agent writes C or C++ for one function at a time and the target's original compiler proves the result byte for byte.

Why was decomp.wtf discontinued?
Per its README, the author stopped because much of the decompilation community was against AI decompiling of that kind, and he respects their opinion. The code stays up under MIT and he asks that anyone continuing it respect the rules of the projects they work with.

What does the fork add?
Six things, per its README: a graphical console for the REA tool catalog that also serves agents, Linux targets, Windows targets, REA access for agents, reference folders, and opencode 2 support. The README also documents archive support, an intake report, version comparison and runtime coverage import.

Does it need the original game files?
The project does not provide them: the README says it contains no game or application code and provides no ROMs, executables or firmware. Your own copy of the target is what you match against, which is why it asks that you only analyze software you own or are authorized to analyze.

Will it send my work upstream?
No. The README says the agents' work stays on your local branch and nothing is sent anywhere by itself. It also says many community decomps do not accept AI-written contributions, so check a project's rules before opening a pull request.

What does it need to run?
Linux x86-64 or Windows through WSL 2, Python 3.12 or newer, uv, Node.js for npm-based agents and for REA, and Java 21 or newer for Ghidra. macOS is not supported. There is no account, no telemetry, and nothing is uploaded.

Our take: The interesting decision here is not technical. Extending byte-exact matching to Windows and Linux binaries is a real widening of what the engine can do, and the RE console opens the same workbench to anything that can be analysed rather than matched. The decision that matters is that the continuation kept the MIT licence and the byte-exact standard, and restated the responsible-use rules for software rather than games alone. Whether the community wants AI-produced matching decompilation is still an open fight, but a project that argues for itself with a compile-and-compare judge is a more useful answer than a manifesto.

Bottom Line

decomp.anything is an MIT-licensed Python fork of decomp.wtf, the AI-driven matching decompiler its author archived in October 2026 after community pushback against AI decompiling. The fork was created on October 19, 2026 and had 22 stars, 3 forks, no releases and no open issues at the time of check. It keeps the upstream byte-exact rule, where a score of zero means the compiled function matches the target's bytes and nothing less counts, and extends it to Windows PE executables and Linux ELF binaries, adds a graphical console for the REA tool catalog that agents can also query, and treats archives as projects whose files become work items. It runs on Linux or Windows through WSL 2 with Python 3.12 or newer, and it asks that you analyze only software you own and never publish decompiled code whose rights you do not hold.

Sources

giveen/decomp.anything repository README, including the fork statement, what stays from upstream, what the fork adds, the target and platform tables, the agent table, the tool boundary, the pinned tool versions, the repository layout and the responsible use section, verified October 21, 2026.

i12bp8/decomp.wtf README and repository metadata, including the discontinuation notice, the last release and the archive state, verified October 21, 2026.

GitHub repository metadata for giveen/decomp.anything, verified October 21, 2026: 22 stars, 3 forks, 0 open issues, Python, MIT, created October 19, 2026, last push October 20, 2026.

Dragon Age Origins Recomp: v0.10.0-dev.4 Recompiles Four Missed Jump Tables

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:39 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Dragon Age: Origins Recompiled shipped v0.10.0-dev.4 on October 21, 2026, and the build does two things worth a second look: it recompiles four jump tables the recompiler had previously missed, and it stops the game from dying on a failed downloadable-content check instead of showing its own error message. It also arrives on a launcher that now has a Release and a Dev update channel, keeps a spare copy of every save, and rebuilds the game for the exact processor you are standing at.

This site covered the project on October 20, when v0.10.0-dev.3 was the newest build. That piece covered what the recompilation is and what it adds. This one covers the part that changed since: the launcher's release engineering, and the fixes in the build published today.

Image: GitHub social preview for the 96dot/Dragon-Age-Origins-Recomp repository. Image credit: 96dot, via GitHub.

Four jump tables the recompiler had missed

The v0.10.0-dev.4 release notes, published on October 21, 2026, list two crash fixes and a temporal upscaling improvement. The crash fixes are the interesting part.

The first is narrow and nasty. Loading a save while an area is still streaming in, a case the notes describe as seen when loading Ostagar after Awakening, could crash the render thread. The fix makes effects whose parameters have not loaded yet draw without them, rather than waiting on data that is not there.

The second is the one with consequences for anyone who owns the DLC. Four jump tables that the recompiler had missed are now recompiled. One of them sits in the Title Update 6 downloadable-content check, and when that check found a problem the game stopped instead of showing its own "downloadable content is corrupt" message. The other three live in both the disc code and the Title Update 6 code, and had not been seen to crash.

That last detail is the sort of thing a static recompiler is exposed to and an emulator is not. A jump table the translator did not recompile is not a wrong translation of an instruction, it is a missing piece of the program's own data, and the symptom appears far from the cause. The release notes put it plainly: the game stopped, and it should have said something useful.

The third change is cosmetic by comparison. The screen border no longer jitters, and full-screen 2D, meaning menus and loading screens, is no longer jittered either. The notes credit the community contributor who reported it. A separate fix in the same series, from v0.10.0-dev.1, stopped motion vectors crashing under Wine and Proton through vkd3d-proton, where stream output was unbound with an empty view instead of unbound with none.

Release and Dev, and what the update check sends

The launcher grew an update channel switch in the v0.10.0-dev.1 build, and it is the kind of small feature that changes how a project is used.

Per the README, there are two channels, chosen in the launcher's Configure, Advanced, Update channel screen, or with update_channel in the [launcher] section of the configuration file. Release, the default, offers releases only, which at version 0.x are marked pre-release on GitHub and offered all the same. Dev offers the dev releases too, for trying what is next early. The release notes are explicit that dev builds can have bugs, and that an install made from a dev release starts on the Dev channel.

Switching back to Release does not roll the install back. Per the v0.10.0-dev.1 notes, the launcher then waits for the next release, v0.10.0, rather than going back to an older one. That is the right behaviour for a build that has already been recompiled on your machine, and it is also why the release notes for every dev build carry the same advice: if you would rather wait, stay on v0.9.0, because its launcher does not offer dev releases.

What the check sends is worth stating. Per the README, the launcher asks api.github.com/repos/96dot/DAORecomp/releases for the list of releases, and the request carries only the program's version. That can be turned off in the same Configure, Advanced screen or with check_updates = false in the configuration file. Builds from source do not check at all, and when the launcher is hidden there is no check either; the notes say to open it now and then by holding Shift.

On the update path itself, the launcher downloads the new release's executable into an update/ folder inside the install folder, checks it against the SHA-256 lists the project publishes on GitHub, and runs it against the same folder. The checksum catches a damaged or altered download. An update rebuilds only what changed, so the disc image is not needed a second time, and saves, settings and mods are kept.

Release	Published	What it changed
v0.9.0	October 14, 2026	Cheats menu, AMD Anti-Lag 2 and NVIDIA Reflex, NVIDIA DLSS 4.5, Streamline SDK download
v0.10.0-dev.1	October 15, 2026	Release and Dev update channels, START menu rows for Photo Mode and Cheats, neural rendering frame rate guard, motion vector fix under Wine
v0.10.0-dev.2	October 15, 2026	Game language selection, English, Italian or Spanish
v0.10.0-dev.4	October 21, 2026	Four recompiled jump tables, streaming-load render thread crash fix, temporal upscaling border and 2D jitter fix

Save backups and the restore path

Every save overwrite now keeps the version it replaces, in a save_backups folder inside the install. Per the README, it keeps one backup per save, the one before its latest overwrite, and a save the game crashed while writing does not replace it.

When there is not enough room to keep the old save, or the new one's folder cannot be made, the save is not overwritten at all. The game shows a message over the top saying it was not saved, and the old one stays. Closing the window while the game is writing a save waits for the write to finish, up to 10 seconds.

The restore path is in the launcher's Save Backups screen. It lists the backups by character, with when each backup and the save in its place were made. Restoring asks first and will not run while the game is running, and the save it replaces becomes the backup, so restoring again switches back. Two folder names have meanings worth knowing: a .new folder is a backup that could not be put in place, with the reason in the log, and a ~current folder is the save a restore set aside, which is put back when the list is next read with the game closed.

This matters more than it sounds. The cheats menu that arrived in v0.9.0 changes the save like anything else in the game, and the release notes for that build say to back the save up first if you may want it as it was. Before this week there was a way to do that, through a separate save backup screen; now the protection is built into the overwrite path itself, and it fails safe by refusing the write rather than losing the old file.

A build made for your processor, and your processor only

The README is unusually clear about a limitation that other projects bury. The game is built for your processor. With AVX2, which it uses on Intel from 2013 and AMD from 2015, the build takes those newer instructions, and a copy built that way and moved to a PC without them says so when it starts. The instruction is to run the installer on that PC to build the game for its processor.

The install itself needs an x86-64 PC running Windows 10 or 11 or Linux, with a Direct3D 12 capable GPU and a processor with AVX, which the README dates to most processors from 2011 or later. About 25 GB of disk space is needed, 8 GB of it for the game and the rest for the tools and the build, and the first build takes 30 to 60 minutes depending on the CPU.

What the installer downloads is pinned and checksummed, per the README's dependency list: Xenia's third-party libraries and rapidjson, XenonRecomp, SDL 3, the AMD FidelityFX SDK with the signed FSR and frame generation libraries, the Intel XeSS SDK, and NVIDIA's Streamline SDK 2.14.1 for Reflex and DLSS. On the toolchain side it takes LLVM or Clang 22, or the system's LLVM 18 or newer when it has the right tools, plus CMake and Ninja, and MinGit on Windows without git. The Microsoft C and C++ runtime headers and the Windows SDK come from xwin under Microsoft's licence, which the installer asks you to accept, and no Visual Studio installation is needed.

One warning belongs in this story rather than a footnote. OptiScaler, and forks of it, can take over the temporal upscaling, but they are not included. Per the README you copy the files next to the game executable without replacing the AMD FidelityFX libraries already there, rename the OptiScaler library to dxgi.dll, and leave temporal upscaling on, because that is what it replaces. A dxgi.dll sitting there runs as part of the game and the launcher every time it starts, so the README says to use only a build you trust.

The DLC path and the title update check

The DLC flow is unchanged and worth restating for anyone new to the project. Awakening, Leliana's Song, The Darkspawn Chronicles, The Golems of Amgarrak and Witch Hunt need Title Update 6 or 7, and the other DLC works without it. You supply the title update package and the DLC packages from a console or your own backup of them.

Only packages in the read-only format Xbox Live delivered them in, LIVE or PIRS, are read, and the packages themselves are left alone. The title update is checked against the supported disc's media ID, extracted into the install folder, and the game's code is translated a second time from the updated executable, which takes about as long again. The program then holds both versions and runs the updated one.

At startup the build checks title_update/default.xexp to confirm it is Title Update 6's or 7's, and anything else is passed over in favour of the disc's own version. The launcher's Downloadable Content screen shows which DLC and title update are installed and whether the game can use them.

With the DLC check now recompiled, the "downloadable content is corrupt" message the game was supposed to show can actually appear. That is the whole point of the fix, and it is a better outcome than the alternative, which was the game stopping with nothing to say.

FAQ

What is new in Dragon Age: Origins Recompiled v0.10.0-dev.4?
Four jump tables the recompiler had missed are now recompiled, one of which was in the Title Update 6 downloadable-content check where the game stopped instead of showing its error message. It also fixes a render thread crash when loading a save while an area is still streaming, and stops temporal upscaling from jittering the screen border and full-screen 2D.

How do I get dev builds?
Choose the Dev channel in the launcher's Configure, Advanced, Update channel screen. An install made from a dev release starts on that channel. To go back, choose Release, and the launcher waits for the next release rather than going back to an older one.

Can I keep my character if I switch channels?
Your saves, settings and mods are kept, and an update only rebuilds what changed. What does not come back is a previous build, since going back to Release waits for the next release instead of downgrading.

What happens to my saves?
The version a save replaces is kept in the save_backups folder, one per save, and a save the game crashed while writing does not replace it. If there is no room to keep the old save, the game refuses the write and says so rather than losing the file. Restores are in the launcher's Save Backups screen.

Can I move my install to another PC?
Sometimes, with a catch. The build is compiled for your processor, so with AVX2 it uses instructions a 2013 Intel or 2015 AMD or newer has, and an AVX2 build moved to an older PC says so at startup. Run the installer on the destination PC to rebuild it for that processor.

Do I still need my own disc image?
Yes. The installer extracts the Xbox 360 disc image, translates the game's code to C++ on your computer and builds the program there, and the repository ships no game code or data. Supported discs are the English, Italian and Spanish edition, which is also disc 1 of that edition's Ultimate Edition.

Our take: The jump table fix is the kind of detail that says more about the project's maturity than any feature list. A recompiler that silently drops a piece of the original program's data will produce crashes whose error messages point nowhere near the cause, and the fix here restores a message the developers wrote years ago and never saw. The update channel work matters for a different reason: it lets a project ship test builds to people who want them without asking the people who do not to take them.

Bottom Line

Dragon Age: Origins Recompiled is a community static recompilation of the Xbox 360 version of Dragon Age: Origins into a native x86-64 Windows program, and its v0.10.0-dev.4 build, published on October 21, 2026, recompiles four jump tables the recompiler had missed, including the Title Update 6 downloadable-content check where the game used to stop instead of reporting the problem. The same build fixes a render thread crash on loading a save mid-stream and stops temporal upscaling from jittering the border and full-screen 2D. The launcher now carries Release and Dev update channels, keeps one backup of every save it replaces with a restore path in the launcher, and rebuilds the game for the processor you are standing at, which means an AVX2 build will refuse to run on an older CPU. It still needs your own Xbox 360 disc image, and it ships no game code or data.

Sources

96dot/Dragon-Age-Origins-Recomp repository README, including the install requirements, the update channel and update check behaviour, the save backup rules, the DLC and title update handling, the dependency list and the OptiScaler notes, verified October 21, 2026.

96dot/Dragon-Age-Origins-Recomp releases, including v0.10.0-dev.4 published October 21, 2026 with the jump table recompiles, the streaming-load crash fix and the temporal upscaling fix, plus v0.10.0-dev.1, dev.2 and v0.9.0, verified October 21, 2026.

GitHub repository metadata for 96dot/Dragon-Age-Origins-Recomp, verified October 21, 2026: 20 stars, 1 fork, 1 open issue, C, MIT, created October 10, 2026, last push October 21, 2026.

benilla-re: 2,067 Function Rows Now Name a Test That Proves Them

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:36 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: benilla-re, the reverse engineering of the World of Warcraft 1.12.1 client, has moved from naming the client's functions to reimplementing them and proving the reimplementation. 2,067 of its 37,967 function rows now name a test, and twenty-one commits landed between October 17 and October 21, eighteen of them touching the function table. The new work in that window is concentrated in a handful of client systems: the auto-loot sweep, the game clock, screen scaling and the worldport console command each gained a reimplementation and a test.

This site covered the project on October 17, when the headline was a symbol table of 37,967 named functions with every claim graded by the evidence behind it. That part has not changed. What has changed is what the project does with the names. The crates under crates/ are no longer a naming exercise. Each one reimplements a slice of the client's behaviour, and the harness in crates/difftest runs the client's own code inside an emulator beside the reimplementation and compares the two bit for bit.

Image: benilla-re repository banner. Image credit: samwhosung/benilla-re, via the repository.

The harness: the real client running beside a copy

The method document describes the arrangement in one line: a bit-exact test runs the exe's own code in the Unicorn emulator beside the reimplementation. The client is not shipped. It is the reader's own copy of WoW.exe from their own 1.12.1.5875 install, and the harness refuses any other build by SHA-256.

The crate is called benilla-re-difftest, and its own comment explains the trick. It runs the executable's own code in the Unicorn x86-32 emulator on chosen inputs so a reimplementation can be compared with the client bit for bit. The executable is located through the BENILLA_RE_WOW_EXE environment variable, or at WoW/WoW.exe inside the repository, and when it is absent the crate returns nothing and writes one skip line. Set BENILLA_RE_REQUIRE_EXE=1 and the same situation panics instead, so that a skip can never pass as a test.

That is the design detail that matters most. A differential test that quietly skips when the target is missing is a test that reports success while proving nothing, and the project made the loud version the opt-in. The install's Data folder is treated the same way, with BENILLA_RE_REQUIRE_DATA=1, and a skip log file lets a runner count the skips rather than infer them.

Emulation has to be bounded, and the harness bounds it by instruction count rather than by time, so the limit is the same under load. The budget is 200 million instructions per emulation, and the crate records that the largest healthy case runs about 10 million.

There is also an inventory tool. A coverage variable names a file to which each test appends every executed basic-block address once, and a separate variable collects the x87 instructions a test runs that the emulator computes in the host's double precision, as a census with the address and the mnemonic. More on that below, because it is the interesting failure mode.

The x87 problem and the census that tracks it

The client runs the x87 floating point stack at 53-bit precision, rounding to nearest, with every exception masked, and the method document pins the control word at 0x023F. It traces how that gets set: WinMain calls __clearfp and then __control87(0x9001F, 0xFFFFFFFF), which turns the exception bits into the mask, adds the denormal bit and the 53-bit precision bit, and loads the result with fldcw. The Direct3D 9 device is then created with flags that include D3DCREATE_FPU_PRESERVE, so the graphics device never lowers the precision either.

Every finding and test in the project assumes that control word unless it says otherwise. The emulator's x87 is not the client's at every instruction, and the project does not pretend otherwise. Ten transcendental and remainder instructions, f2xm1, fyl2x, fyl2xp1, fsin, fcos, fsincos, fptan, fpatan, fprem and fprem1, convert their operands to the host's double, call its C library and convert back.

The consequences are specific and the project writes them down. A result carries 53 significant bits where the real x87 rounds at 64, and the two remainders are computed in double rather than exactly. The edges differ too: fsin, fcos, fsincos and fptan reduce an operand that rounds to plus or minus 2 to the 63rd power as a double, which the emulator reduces as a double, answering with infinity and the C2 flag set, where the x87 reports the operand out of range and returns the indefinite. The two remainder instructions give the host's positive NaN for an infinite or NaN operand, a zero divisor, or exponents far enough apart that their scale overflows a double, and with exponents 53 or more apart their partial remainder is not the x87's at all.

The rule that follows is the one a test follows. A test whose compared values pass through one of those instructions shows the code around the instruction, not the client's result there, and keeps its inputs away from the other differences. The census file records every address of the ten that a test runs, and a test that runs one on an operand where the answers differ beyond rounding writes a second line marked edge. A test with an edge line compares no value that passes through it there.

That is a level of honesty about the limits of a verification method that most projects do not reach. The alternative would be to quietly compare a double against a 53-bit x87 and call a mismatch a bug in the reimplementation.

What landed in four days

Between October 17 and October 21 the repository took 21 commits. Eighteen touched the function table, thirteen touched the crates, and six touched the tables. The work is not random: it reads like a client being walked system by system.

The clearest example is auto-loot. The commit that added it is titled "Auto-loot's sweep and LOOT_OPENED, and the DB cache's callback order", and the new document for it describes what the client's shift-click sweep does at 0x4c1fa0 in LootFrame.cpp. The sweep returns at once while item records are pending, sends CMSG_LOOT_MONEY when the money cell is not zero, plays the coin sound and prints the money line once per source outside a party and a raid, sends each row of loot type 0 as a click does, except that a bind-on-pickup row of quality 2 or more raises LOOT_BIND_CONFIRM on the first such row and leaves the rest alone. It then clears the player's auto-loot byte, and closes the loot with release when nothing is left, or signals LOOT_OPENED again if the flag it was given is zero.

That last part is the kind of detail that makes the approach worth reading. The sweep's sends change no row and not the money cell: a row disappears when the server's SMSG_LOOT_REMOVED arrives, and the coin row when SMSG_LOOT_CLEAR_MONEY does, both after the sweep has already returned. The reimplementation in crates/ui/src/auto_loot.rs models the calls the sweep makes and the two cells it stores, and the test reads the 16 rows and the money cell back after each sweep to find them unchanged.

The game clock got the same treatment. The new document covers the server's game-time messages and what they do to the clock, and the crate in crates/gametime reimplements the time-of-day computation, noting that the client runs the x87 arithmetic at 53-bit precision. The decode of the packed date record is the sort of thing that only matters because something else depends on it: seven int32 fields are read out of a dword, minute at bits 0 to 5 and hour at 6 to 10 and weekday at 11 to 13 and day at 14 to 19 and month at 20 to 23 and year at 24 to 28, with a two-bit field at 29 and 30, and a field whose bits are all ones is stored as minus one. No field is range-checked, so a decoded hour can reach 30 and a minute can reach 62.

Screen scaling was documented and reimplemented as a pair, four dwords set from the display's aspect ratio with 446 other call sites scaling, unscale or reading through them, and the worldport console command modelled as a module, reading what CCommand_WorldTeleport takes from its argument text, including the delimiter set at 0x82ecc8, which is tab, carriage return, line feed, quote and space, and no comma, and the usage line the command prints when the arguments do not parse.

Chat channels, the item record's owner signal, the ammo callback, the chat token expander and PickupContainerItem all landed in the same window, alongside a commit that covers the channel notify message that confirms, renames and drops the client's ten joined channel slots.

The shape of the repository now

The tree is bigger than the symbol table. The project holds 1,648 files, of which 507 are markdown documents under systems/ covering 40 documented systems, 23 are tab-separated tables under tables/, 971 belong to 37 Rust crate directories, and 66 sit in tools/. The symbol table names 45 distinct systems, and seven of them, including the base, the C runtime and the Direct3D libraries, have no document of their own.

The tables have their own rules. Each one records facts a program reads, with a header row, one row per fact, and a generator in tools/tables/. Nothing in them is edited by hand, because per the tables README a wrong row is a wrong generator. The generators are Python 3 with the standard library only, they read the executable and refuse any other build by size and SHA-256, and running one twice gives byte-identical output. The check command compares each table's header against the columns the README gives it, and with the executable present it regenerates every derived table into a scratch directory and fails on any byte that differs from the committed one.

That is the point of the whole arrangement, and the Oct 17 piece made it well. A symbol table that nobody checks drifts into folklore. A symbol table where every row is graded, every table is regenerated, and every claim can be turned into a test against the real binary does not drift quietly, because the build fails.

Measure	Value in the repository on October 21, 2026
Function rows in the symbol table	37,967
Rows that name a test	2,067
Distinct test names	2,115
Rows claiming two independent readings reconciled	165
Rows with no owning system yet	5,889
Documented systems	40
Markdown documents under `systems/`	507
Tab-separated tables	23
Rust crate directories	37
Commits between October 17 and October 21	21
Commits since October 13, 2026	121

On the licence, the README is specific and slightly unusual. The code, the symbol table, the docs and the tables are all under MIT or Apache-2.0 at the reader's option. The exception is the test harness, which links the Unicorn emulator, a GPL-2.0 project, as a test-only dependency, and the crate's own comment notes that it is only ever a dev-dependency of the system crates, so nothing that ships links it.

What is still unfinished

The gap between a documented client and a reimplemented one is visible in the numbers. 5,889 of the 37,967 function rows still have no owning system, which means the project has a name for the function and has not yet worked out which part of the client it belongs to. The README's method document describes how a system claims a file's functions, by a listed address span, a source path, an initializer run or a stretch of code, and says a row stays unassigned until its file is proved.

The repository is also young, at 121 commits, and it is not a place where a claim is made lightly. The same method document that pins the x87 control word also lays out checklists for reading the bytes, for proving a finding by absence, for writing a finding and for writing an observation, which is a lot of ceremony for a project that started on October 13.

What it buys is a reference that can be checked by a machine instead of trusted. That is the part worth carrying forward.

FAQ

What is benilla-re?
A public reverse engineering of the World of Warcraft 1.12.1 client, read at its bytes and cited by address. Everything ships as plain files, and only the tests, the table generators and the listing read WoW.exe from your own copy, which they check by SHA-256 and refuse to substitute.

How does it test a reimplementation?
The harness in crates/difftest runs the client's own code in the Unicorn x86-32 emulator on chosen inputs, beside the reimplementation, and compares the results bit for bit. The executable is found through an environment variable or at WoW/WoW.exe, and if it is missing the test skips with one line unless you set the require variable, which makes it panic instead so a skip cannot pass as a test.

What does the project not cover?
The 200 million instruction budget per emulation bounds what a test can show, and the largest healthy case runs about 10 million. Separately, the emulator's x87 differs from the client's at ten transcendental and remainder instructions, so the project keeps a census of every address a test runs through one of them and refuses to compare a value that passes through it there.

Is the client binary included?
No. The README says the symbol table, the docs and the tables are plain files that need no executable, and that nothing here comes from a WoW install. Bring your own legally obtained client.

What licence applies?
MIT or Apache-2.0 at your option, for the code, the symbol table, the docs and the tables. The test harness links Unicorn, which is GPL-2.0, as a test-only dependency, so nothing that ships links it.

How much of the client is understood?
37,967 function rows carry a name and a confidence grade, 2,067 of them name a test, and 165 claim two independent readings reconciled at the bytes. 5,889 rows are still marked unassigned to a system, which is the honest measure of what is left to do.

Our take: The interesting shift is not the volume of the work but its shape. A reference that names functions is useful; a reference whose names come with a running test against the original binary is a different kind of artifact, because the next person can disagree with it and find out immediately. The x87 census is the part to admire, since a project that quietly compared doubles against a 53-bit stack would have shipped confident wrong answers and never known.

Bottom Line

benilla-re is the MIT or Apache-2.0 licensed reverse engineering of the World of Warcraft 1.12.1 client, and as of October 21, 2026 it is no longer only a map of the binary. 2,067 of its 37,967 function rows name a test, 1,541 distinct tests exist, and the project reimplements client behaviour in Rust and runs the client's own code beside the reimplementation in the Unicorn x86-32 emulator, comparing results bit for bit. Twenty-one commits landed between October 17 and October 21, adding reimplementations of auto-loot, the game clock, screen scaling and the worldport command. 5,889 function rows are still unassigned to a system, which is the honest measure of the work left. The client's own binary is never included, and the test harness links the GPL-2.0 Unicorn emulator as a test-only dependency so that nothing shipped carries it.

Sources

samwhosung/benilla-re repository README and its method document, including the harness rules, the x87 control word and the emulator's ten differing instructions, the confidence and proof grades, the licence terms and the citation file, verified October 21, 2026.

The repository's own symbols/functions.tsv, crates/difftest/, the crates for auto-loot, gametime, screencoord and worldport, and the system documents for auto-loot, game time, chat channels and screen coordinates, for the counts, the addresses and the test names, verified October 21, 2026.

GitHub commit history for samwhosung/benilla-re, for the 21 commits between October 17 and October 21 and the 121 commits since October 13, verified October 21, 2026.

Cyberpunk 2077 Update 2.5: Cross-Platform Mod Support Comes to Console

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 12:43:34 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: The official patch notes for Cyberpunk 2077's Edgerunners 2 Update, numbered 2.5, are now public, and they confirm what the October 20 announcement left open. The update adds cross-platform mod support through mod.io, dual-wielding as a Cool tree perk, hand cyberware skinning, and DLSS 4.5 with up to 6X Multi Frame Generation. It also removes FSR 3 upscaling from the options list. The update reaches PC, Mac, PlayStation 5 and Xbox Series X|S on October 22, 2026, with Nintendo Switch 2 and the Mac App Store version to follow in the days after.

What the notes do not contain is as informative as what they do. There is no new district, no new ending, and no level cap change. The anime content is a side quest, a radio station, a gear set and a row of outfits, and the rest of the update is quality of life, graphics options and a long fix list.

Image: promotional image for the Edgerunners 2 Update. Image credit: CD PROJEKT RED.

What the Edgerunners content actually is

The content list in the patch notes is short, and it is worth reading item by item because the announcement promised a lot less detail than the patch notes now deliver.

There is a new side quest called Clampdown, set before the events of the anime. Per the notes it unlocks after completing The Heist and two gigs for Wakako, then waiting for a message from her. A new 5-piece gear set based on the character King can be obtained during that quest. And there is a new radio station, HKGN 102.0 FM, carrying a selection of tracks from the Cyberpunk: Edgerunners series.

That is the whole Edgerunners story content. One side quest, one armour set, one radio station. The prior article on the October 20 announcement flagged the open question of whether the anime tie-in would reach beyond cosmetics, and the notes answer it: a quest, an outfit, and a radio station are what shipped, and the quest is a side quest rather than a new storyline.

The rest of the update is not anime content at all. It is a Night City tour, tarot cards, photo mode additions, console settings and a long list of fixes.

Item	What the patch notes say
Clampdown	New side quest set before the anime's events, unlocked by The Heist plus two Wakako gigs
King's set	New 5-piece gear set obtained during Clampdown
HKGN 102.0 FM	New radio station with tracks from the Edgerunners series
New outfits	Five new outfit sets from clothing vendors, plus NUS Infiltrator in Kabuki and Japantown
Cyberware skins	Hand cyberware can be skinned at 5 Glamdoc shops, with colours
Double Trouble	New Cool tree perk, plus four auxiliary perks, for dual-wielding Power category pistols and revolvers
Cross-platform mods	mod.io powered mod browser and installer, in-game
Welcome to Paradise	New side quest narrated by Richard Night, unlocked by The Heist and level 15 Street Cred
Tarot cards	12 new Minor Arcana cards, in a side quest extension of Fool on the Hill
Upscalers	DLSS 4.5 with up to 6X Multi Frame Generation and the 2nd Generation Transformer Model; FSR 3 upscaling removed

The cosmetic layer and the new perk

The bulk of the Show Off section is exactly that, show. Five new outfit sets are purchasable from clothing vendors, with named locations: Maxiwear in Heywood, Arroyo and Little China, Jinguji in City Center, NetWatch in Kabuki and Japantown, and IconAmerica in Pacifica, Badlands and Northside. A fifth set is promised for later, and the NUS Infiltrator set is now purchasable in Kabuki and Japantown. Four new TWINTONE skins for the Quadra Turbo-R V-Tech arrive as Twitch Drops.

Two cosmetic changes are more than recolouring. Hairstyles that were previously exclusive to NPCs are now available for V, with some of them limited to feminine or masculine body types. And hand cyberware can be customised for the first time: the notes name the Mantis Blades, Gorilla Arms, Monowire and Projectile Launch System, and say you visit one of 5 Glamdocs, marked with a spray paint icon on the map, each in its own shop with an intro scene with voiceover and animations, offering a catalogue of cyberware skins with customisable colours.

The one gameplay addition in that section is the Double Trouble perk in the Cool tree, with four auxiliary perks, which the notes say allows dual-wielding of pistols and revolvers from the Power category. That is the "new dual-wielding playstyle" the October 20 announcement teased, now with a specific location in the skill tree. Photo Mode also gets frames from the Cyberpunk Trading Card Game, so screenshots can look like custom cards, and new poses for showing off dual-wielding.

Cross-platform mods through mod.io

The most consequential line in the notes is short. Cross-Platform Mod Support, powered by mod.io, lets players browse and install a selection of player-created mods directly from the in-game menu, and the notes call it a first step into cross-platform modding with more mods to come over time. The feature is listed for PC, Mac, PlayStation 5, PlayStation 5 Pro and Xbox Series X|S.

It is worth being precise about what that means in practice. Mods on a console require the platform to allow them, so the scope of what is installable is set by what mod.io can offer on each platform, and the notes describe a selection rather than the whole catalogue. A first step is the wording CD Projekt chose, and the phrase "cross-platform" describes the ambition rather than a promise that the same mod runs identically everywhere.

Two related quality-of-life entries make mods a first-class part of the save system. Modded saves in the Load Game menu are now marked with a dedicated icon, and the notes warn that achievements and trophies cannot be earned on console while playing from a modded save file. That is a specific and unusual limitation, and it is the kind of detail that will be tested rather than assumed.

The graphics changes, including FSR 3's removal

The Technology section is where the update has the most reach for players who care about frame rates.

DLSS 4.5 arrives, including up to 6X Multi Frame Generation or Dynamic Multi Frame Generation, and the 2nd Generation Transformer Model. The notes also add support for glasses-free 3D effects on compatible Samsung Odyssey 3D monitors, with a Motion Comfort setting to balance 3D depth and viewing comfort. MetalFX Frame Interpolation and Denoising come to Mac. There is a toggle in settings for FSR Anti-Lag 2, various performance optimizations including improvements specifically for ray tracing and for PlayStation 5 Pro, and improved HDR rendering for more accurate colours, brightness and contrast on compatible displays.

And then the removal. The FSR option list has been updated, and FSR 3 upscaling is no longer included. FSR 2.1, FSR 4, and FSR 3.1 Frame Generation remain available.

That is a strange cut, because FSR 3 upscaling is the only entry in the list that combined upscaling with frame generation, and the game added FSR 3 support in September 2024, which this site covered at the time. FSR 3.1 Frame Generation surviving without FSR 3 upscaling leaves AMD users with FSR 2.1 for upscaling and FSR 4 for the newer path, and the notes give no reason. The update is scheduled to land tomorrow, so the practical answer will come from testing rather than from the patch notes.

One more graphics-adjacent change is worth flagging for console players: an FOV slider now exists on consoles. The console version of the game shipped without one, and the patch notes do not describe it as a new feature so much as an option that was missing.

Quality of life and the fixes

The Quality of Life list is short and specific. Saves in the Load Game menu are now grouped by separate playthroughs, which is the kind of change that matters for anyone with several characters in one file. Item filters can now be applied to apartment and vehicle stashes. The CD PROJEKT RED Account connection, My Rewards and telemetry consent menus have been updated.

The fixes cover a mix of platforms and ages of the game. On Nintendo Switch 2, saving could produce a "Save failed - insufficient space" error. On PlayStation 5, adjusting HDR settings had no effect. The Delamain cab could get stuck after crashing into another vehicle, blocking quest progression. Enabling Film Grain had no effect. Tagging enemies while scanning could also make V crouch or stand up under some control schemes. Third-person camera controller settings were missing. The notes also fix various typos in various languages, and various other issues with features and content introduced since Patch 2.1.

The Delamain cab bug is the one to watch on day one, because a stuck cab that blocks quest progression is a save-scraper, and the Switch 2 save error is the one that could cost players progress on the least mature platform in the list.

FAQ

When does the Cyberpunk 2077 Edgerunners 2 Update release?
October 22, 2026, for PC, Mac through Steam, GOG and Epic Games Store, PlayStation 5 and Xbox Series X|S. Players on Nintendo Switch 2 and on the Mac App Store version of the game get it in the days that follow.

Is the Edgerunners 2 Update free?
Yes. CD PROJEKT RED describes it as a free update, and it is numbered 2.5, following the 2.1 and later patches that added features such as FSR 3 and XeSS 1.3 support.

Does the update add new story content?
Two side quests, one of them tied to the anime, plus a new gear set, a radio station, outfits and cyberware skins. The patch notes list no new district, no new ending and no level cap change.

Does Cyberpunk 2077 still support FSR 3?
FSR 3 upscaling is no longer included in the updated options list. FSR 2.1, FSR 4 and FSR 3.1 Frame Generation remain available, and the notes give no reason for the removal.

Can you install mods on console?
The patch notes list Cross-Platform Mod Support through mod.io for PC, Mac, PlayStation 5, PlayStation 5 Pro and Xbox Series X|S, described as a first step with a selection of mods available. Achievements and trophies cannot be earned on console from a modded save.

What is new for console players?
A field of view slider, which the console version of the game did not have, plus the cross-platform mod browser. The update also fixes an HDR settings bug on PlayStation 5 and a save error on Nintendo Switch 2.

Our take: The patch notes are a content-light, feature-wide update, which fits a free drop timed to an anime premiere. The two things worth watching are technical rather than thematic. Cross-platform mod support on console, if it holds, is a bigger deal for the game's remaining life than the quest it ships with. And dropping FSR 3 upscaling while keeping its frame generation is the kind of change that only makes sense commercially, which is a poor reason to hand an AMD player a worse option on the eve of release.

Bottom Line

Cyberpunk 2077's Edgerunners 2 Update, version 2.5, is a free patch with official notes published on October 21, 2026 and a release on October 22 for PC, Mac, PlayStation 5 and Xbox Series X|S, with Nintendo Switch 2 and the Mac App Store version following. It adds a Clampdown side quest, a 5-piece gear set, the HKGN 102.0 FM radio station, five outfit sets, cyberware skinning at 5 Glamdoc shops, the Double Trouble dual-wielding perk, the Welcome to Paradise quest, 12 Minor Arcana tarot cards, mod.io cross-platform mod support, a console FOV slider and DLSS 4.5 with up to 6X Multi Frame Generation, while removing FSR 3 upscaling and leaving FSR 2.1, FSR 4 and FSR 3.1 Frame Generation in place. There is no new district, no new ending and no level cap change.

Sources

CD PROJEKT RED official patch notes for the Edgerunners 2 Update, including the Edgerunners content, Show Off, Mod the World, Upgrading the World, Quality of Life, Technology and Fixes sections, published October 21, 2026, verified the same day.

Cyberpunk 2077: Edgerunners 2 Update 2.5 - Overview

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 10:50:30 -0400

New Edgerunners, new chrome, new playstyle? Sounds like we need another REDstreams! Join us this Wednesday, October 14th at 5 PM CEST for a closer look at Edgerunners 2 Update (2.5) and some of the features coming to Cyberpunk 2077. We're talking cyberware customization, a new dual-wielding playstyle, and more! See you in chat, chooms 🤘

KickDropsMiner: A Queue Builder for Kick Drop Campaigns

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:51:41 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: KickDropsMiner is a Kick-only desktop app that automates a queue of live channels, tracks watch time toward drop campaigns, and gives you a browser for active Kick campaigns. Its latest release is v1.0.30 from June 13, 2026, and the repository has not been pushed since that same day, so it is the oldest build in this group despite having 82 stars.

The app is written in Python with a CustomTkinter interface and Selenium plus undetected-chromedriver underneath. It does not try to be a cross-platform service or a headless miner. It opens Chrome, logs in with your cookies, plays the channels you queue, and keeps the timer moving while the stream is live. That makes it closer to a browser automation controller than to a Twitch API client.

Image: The KickDropsMiner main window. Image credit: HyperBeats/KickDropsMiner, via the repository README.

A queue instead of a campaign list

The core workflow is a table of Kick live URLs, each with a target number of minutes. You press Add link, paste a channel URL, and enter the minutes you want; zero means infinite watch time. The app processes the queue from top to bottom. Finished entries stay in config.json and can be reset by removing the finished flag or adding the URL again with a new target.

The timer only increments while the channel is actually live, per the README. If Kick reports the stream as offline mid-watch, the worker pauses and resumes when it comes back. Offline channels are tagged Retry and revisited later instead of being dropped. Status tags show LIVE, PAUSED, FINISHED, or STOP, and you can stop any selected entry.

Queue feature	What the README says it does
Target time	Per-entry minutes, with 0 for infinite watch time
Offline handling	Tags the entry Retry and revisits it later
Live check	Timer increments only while Kick reports the channel live
Status labels	LIVE, PAUSED, FINISHED, or STOP
Persistence	Queue entries, elapsed time, preferences, and finished flags in config.json
Stop control	Stop selected ends that entry's worker without clearing the queue

The design assumes you know which channels you want. That is different from a miner that discovers campaigns and picks a stream for you. The campaign browser fills part of that gap, but the queue remains the source of truth.

The campaign browser

The Drops campaigns button fetches active campaigns from https://web.kick.com/api/v1/drops/campaigns. The app spins up a tiny hidden Chrome session so it can pass Cloudflare and reuse your stored cookies, per the README. You can browse each campaign, review the rewards, and see the participating channels with avatars when Kick provides them.

From there, Add this channel pushes one entry into the queue, and Add all channels targets every eligible channel of that campaign. The campaign helper reuses the minutes selector, so you decide how long to farm. It also inherits the authenticated state stored for kick.com, which the README says keeps the fetch rate-limit friendly.

This is the feature that makes the app more than a URL queue. It gives you a live view of what Kick is offering and a one-click path from campaign to watch target. It is also the feature most exposed to Kick's site changes, because it parses a private web API and relies on a browser session rather than a documented public interface.

Cookies, Chrome, and the login path

Sign in (cookies) launches a Chrome window pointed at the selected domain, usually kick.com. You log in, press OK in the app, and the cookies are saved to cookies/kick.com.json. If browser_cookie3 is installed, the app can pull cookies directly from Chrome, Edge, or Firefox without opening a window, which the README says avoids extra 2FA prompts or 429 errors and is the fastest way to refresh authentication.

The cookies are stored per domain and applied before a worker navigates to a stream or before the campaign scraper hits kick.com. That is convenient and also the main security consideration: the app is holding a reusable Kick session in a local JSON file. The README tells you to delete the folder when you are done, which removes config.json, cookies, and chrome_data. There is no vendor account and no cloud sync described.

The stack is a real Chrome instance driven by Selenium and undetected-chromedriver. The README requires Google Chrome, Python 3.10 or newer, and an internet connection so webdriver-manager can download a matching ChromeDriver. It also lists undetected-chromedriver as required and recommends keeping Chrome updated so the driver version stays compatible. That is a maintenance treadmill: a Chrome update can break the driver before the app ships a fix.

Mute, hide, and the mini player

The playback controls are aimed at keeping the browser out of the way. Mute enforces muted audio and reapplies it periodically so Kick cannot unmute the tab. Hide player removes the video element while keeping playback going, and with hide enabled Chrome runs in a headless-like invisible window for the worker. Mini player spawns a small always-on-top overlay plus a reduced Chrome window.

Hide takes precedence over the mini player, so you leave Hide off if you want the overlay. If you load a Chrome extension through the app, Chrome has to stay visible because Google disallows extensions in headless mode, and the app adjusts visibility rules automatically. These are small details, but they show the app was built by someone who actually ran it against Kick rather than wrapping a command line tool.

The README does not describe an automatic claim step. Its focus is watch time, queue state, and campaign discovery. According to the GitHub repository description, however, the app allows you to AFK mine timed Kick drops "with automatic drop claiming and channel switching." That is a direct conflict with the documentation, and it is the kind of gap that should be tested before you rely on the app to claim anything.

Requirements and platform support

The supported target is Windows 10 or 11, with Python 3.10 or newer and Google Chrome installed. Linux and macOS can work but are not the main target, per the README. The quick start uses PowerShell commands to create a virtual environment, install CustomTkinter, Pillow, Selenium, webdriver-manager, undetected-chromedriver, setuptools, and optionally browser_cookie3, then runs main.py. On Windows you can also double-click run.bat.

The persistence layout is simple. config.json holds the queue and preferences. cookies/ holds per-domain JSON exports. chrome_data/ holds a reusable Chrome profile. chromedriver-win64/ keeps downloaded drivers for offline reuse. The README's troubleshooting covers a Chrome launch failure, a timer that does not move, 429 and Too many requests errors, a mini player that is too small, and rewatching a finished link.

Requirement	Detail
Operating system	Windows 10 or 11 primary; Linux and macOS possible but not the main target
Python	3.10 or newer, tested on 3.10
Browser	Google Chrome installed and updated
Network	Required for webdriver-manager to fetch ChromeDriver
Optional package	browser_cookie3 for automatic cookie import
Storage	config.json, cookies/, chrome_data/, chromedriver-win64/

No license and an old release

Two repository facts matter more than the feature list. First, GitHub reports no license for the project, and the repository tree contains no LICENSE file. The source is public, but the absence of a license leaves the legal terms for copying, modifying, or redistributing it unclear. That does not stop you from running it locally, and it does matter if you want to fork it or ship it inside another project.

Second, the latest release is v1.0.30 from June 13, 2026, and the last push to the repository is the same day. As of October 21, 2026, that is more than four months without a push. The release notes for v1.0.30 describe a move toward direct Kick API requests with a headless Chrome fallback, a campaign UI, bulk channel controls, and updated translations. It is a substantial release, but it is not recent.

A browser automation app that depends on undetected-chromedriver ages quickly. Chrome updates, Kick changes its player or campaign API, and Cloudflare changes its checks. The README's own troubleshooting section is a list of ways the browser path can fail. An old build can still work, but it is working against a moving target without recent maintenance.

FAQ

Does KickDropsMiner claim rewards automatically?
The documentation is inconsistent. The README does not describe a claim step, but the repository description says the app supports automatic drop claiming and channel switching. Test the claim path before relying on it.

Does KickDropsMiner support Twitch?
No. It is Kick-only. The repository description, features, campaign helper, and login flow are all specific to Kick.com.

Do I need Python to run KickDropsMiner?
Yes. The README requires Python 3.10 or newer, Google Chrome, and several pip packages, then tells you to run main.py or double-click run.bat on Windows.

Does KickDropsMiner work on macOS or Linux?
It can work, but Windows 10 and 11 are the main target. The README says Linux and macOS are not the primary platform, and the quick start uses Windows PowerShell commands.

Is KickDropsMiner open source?
The source is publicly available, but GitHub reports no license and the repository has no LICENSE file. That leaves the terms for reuse and redistribution unclear, even though you can read and run the code.

Why does KickDropsMiner need Google Chrome?
It uses Selenium with undetected-chromedriver to drive a real Chrome session. The README requires Chrome and a matching ChromeDriver, which webdriver-manager downloads automatically.

Our take: KickDropsMiner is the best fit in this group for someone who wants manual control over a Kick queue and a campaign browser, not a hands-off account service. The queue model is clear, the campaign helper is genuinely useful, and the playback controls show attention to the realities of browser automation. The problems are maintenance and licensing. A four-month-old build against a moving target is a risk, and no license means you should not build a project on top of it without asking the author. Use it as a local tool if the workflow fits, and keep a fallback ready for the next Chrome update.

Bottom Line

KickDropsMiner is a Kick-only Python desktop app with a CustomTkinter interface, a Selenium and undetected-chromedriver backend, a per-entry minute queue, a campaign browser, cookie import, and mute, hide, and mini player controls. Its latest release is v1.0.30 from June 13, 2026, and the repository has not been pushed since, with 82 stars, 17 forks, and 8 open issues on October 21, 2026. It automates watch time well on paper, but its README is silent on automatic claiming while its repository description promises it, it carries no license, and it depends on a browser stack that changes often.

Sources

HyperBeats/KickDropsMiner repository README, including the feature highlights, requirements, quick start, cookie workflow, queue behavior, campaign helper, playback controls, persistence layout, and troubleshooting, verified October 21, 2026.

HyperBeats/KickDropsMiner releases, including v1.0.30 published June 13, 2026 and its release notes.

GitHub repository metadata for HyperBeats/KickDropsMiner, verified October 21, 2026: 82 stars, 17 forks, 8 open issues, Python, no license reported, created November 3, 2025, last push June 13, 2026.

Lurkloot: One Chrome Popup Farms Twitch and Kick Drops

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:51:25 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Lurkloot is a free, open source Chrome extension that farms Twitch and Kick drops from one popup, and its latest stable release, v1.18.0, arrived on October 16, 2026, per the project's GitHub releases. It runs inside the browser you already use instead of asking you to install a separate desktop miner, and a beta command line and Docker image can run the same engine without a browser window.

The project sits at 33 stars, 3 forks, and 28 open issues on GitHub as of October 21, 2026, with TypeScript source under the Apache License 2.0. It was created on May 31, 2026, so it is younger than the long-running TwitchDropsMiner, but it is also more active: the repository was pushed on the same day this article was checked, and the maintainer shipped three releases between October 14 and October 16.

Image: The Lurkloot popup with a drop in progress and a reward queue. Image credit: jamezrin/lurkloot, via the repository README.

What Lurkloot farms

Lurkloot covers two platforms separately, with its own switch, game list, order, and channel list for each. On Twitch, the README lists drops, subscriber badges, channel points, the Idle Watchlist, and two Twitch channel extensions: Fortnite and NoPixel. On Kick, it lists drops, daily rewards, and the Idle Watchlist. You can enable one platform or both.

The optional extras are not part of the main drop loop. Fortnite collects sprites and phase rewards, while NoPixel earns daily packs and can join giveaways. Both are off until you enable them in the popup, and enabling either one asks the browser for access to that extension's site. Drop campaigns are still watched first unless you change the order.

Feature	Twitch	Kick
Timed drops	Yes	Yes
Badges or daily rewards	Subscriber badges	Daily rewards
Channel points	Yes	Not listed
Idle Watchlist	Yes	Yes, also counts toward profile level and badges
Channel extensions	Fortnite and NoPixel, optional	Not listed
Background watching	Most watching runs without stream video or video ads	Same engine, with a muted tab fallback

The Idle Watchlist is the feature that fills the gaps. When no drop, extension, or other reward is eligible, Lurkloot picks a live channel from the channels you added, so the browser is not sitting idle between campaigns. On Kick, that watch time also counts toward your profile level and badges, per the README.

How the extension watches

You sign in to Twitch or Kick as usual, enable the platform, and leave Lurkloot running. The extension detects what you can still earn, picks a live channel that counts toward it, and switches on its own when the stream goes offline, the reward finishes, or a better target becomes available.

Most of that watching happens in the background, with no stream video and no video ads, per the README. If a watch cannot continue that way, Lurkloot opens a muted tab for that channel. That tab plays the stream, so a platform ad can run there. The distinction matters: the low-bandwidth path is the default, but it is not a promise that no tab will ever open.

The popup is a single workspace for both platforms. The queue shows what is being farmed. You can pin campaigns and drag them into order, rank favorite games, limit farming to chosen categories, block categories, exclude channels, and choose which reward Lurkloot prefers when several are available. That is more control than the typical one-channel-at-a-time miner exposes in its main window.

The beta CLI and Docker path

The headless version is the part that needs the most careful reading. The CLI and Docker image are in beta, and the README says Twitch drop claiming from the CLI alone is not fully verified yet. They run the same farming engine on a server, a NAS, or any machine that can run a container, with no browser window and no stream video.

The published image is multi-architecture, for amd64 and arm64, at ghcr.io/jamezrin/lurkloot-cli. A Twitch device login uses Twitch's Smart TV client, which means it cannot see every drop campaign. For full Twitch discovery, you export credentials from the extension with Settings, then Export credentials, and import that file into the CLI. Kick login on the same image uses its own device login command.

Fortnite and NoPixel run only in the extension. The CLI farms Twitch and Kick. So the browserless path is not a complete replacement for the extension if those two channel extensions are the reason you wanted the tool.

Sessions, telemetry, and the AI disclosure

Lurkloot uses the Twitch and Kick sessions already in your browser. It does not ask for your password, does not create a Lurkloot account, and does not send activity to a Lurkloot server, per the README. There is no telemetry. The one path for session data to leave the extension is the manual credential export used to set up the CLI, and that writes a file on your machine.

The repository also carries an unusually direct AI disclosure. The README says Lurkloot is maintained by one person using frontier models and workflows written for the repository, and that the models write most of the code while the maintainer owns the architecture, design, and product decisions. Issues and pull requests must come from a person who has read and checked them, and fully automated submissions are not welcome. That is a more specific statement than most projects make, and it belongs in the assessment rather than being buried.

None of that is a security audit. The extension runs with access to your logged-in Twitch and Kick sessions, which is exactly the kind of permission a drop miner needs and exactly the kind that deserves scrutiny before you install it. The README's privacy claim is a project statement, not an independent review.

What changed in v1.18.0

The October 16 release added a Rewards to farm menu in the Queue and in Settings, available per platform. On Twitch, you can choose in-game items, badges, emotes, or any mix, and set subscription rewards to include, skip, or only. Campaigns outside your choice stay listed under Skipped with the reason, and Farm all resets the filter.

The same release added a Farm last option for campaigns without a linked account. Those campaigns are watched only when no linked campaign can be farmed, and the extension switches back as soon as one can. Favorite games can now be reordered by dragging or by typing a new position. The Kick tab shows viewer level and daily watch streak. The interface also gained Japanese, Indonesian, Polish, Korean, Ukrainian, and Czech translations.

Repository fact	Value on October 21, 2026
Latest stable release	v1.18.0, published October 16, 2026
Stars and forks	33 stars, 3 forks
Open issues	28
Primary language	TypeScript
License	Apache License 2.0
Repository created	May 31, 2026
Last push	October 21, 2026

The limits worth knowing

The extension is the supported path; the CLI is beta. Firefox builds on GitHub Releases are best-effort and untested, while the Chrome Web Store listing works in Chrome, Edge, Brave, Opera, and Vivaldi. The browser has to stay running for the extension to keep farming. Twitch drop claiming from the CLI is not fully verified, and the Smart TV login cannot see every campaign without exported credentials.

Adoption is still small. Thirty-three stars and 28 open issues describe a project with a real user base and a real bug list, not a finished product with a long track record. The README itself warns that platform behavior and terms can change, and that automating viewing may be restricted by those terms. That is the risk that matters here: the tool can be well built and still put an account in front of a platform policy it does not control.

FAQ

Does Lurkloot work in Firefox?
Only as an untested best-effort build. The README says Firefox builds on GitHub Releases are best-effort and untested, while the Chrome Web Store listing is the supported install for Chrome, Edge, Brave, Opera, and Vivaldi.

Does Lurkloot need my Twitch or Kick password?
No. It uses the sessions already signed in inside your browser, and the README says it does not ask for a password. The CLI uses a device login or credentials exported from the extension.

Can Lurkloot run without a browser?
Yes, through the beta CLI and Docker image, but with two caveats. Twitch drop claiming from the CLI alone is not fully verified, and full Twitch campaign discovery requires credentials exported from the extension.

Does Lurkloot collect telemetry?
No, per the repository README. There is no Lurkloot account and no telemetry, and the only session data that leaves the extension is the credential file you export yourself for the CLI.

Does Lurkloot farm channel points and badges as well as drops?
On Twitch it lists drops, subscriber badges, channel points, and the Idle Watchlist. On Kick it lists drops, daily rewards, and the Idle Watchlist, with Idle Watchlist time also counting toward profile level and badges.

Is automated viewing allowed by Twitch and Kick?
The README does not claim that it is. It says platform behavior and terms can change and that automating viewing may be restricted by those terms, so the decision and the account risk are yours.

Our take: The browser-extension approach is the strongest idea here. Most drop miners ask you to install a runtime, keep a desktop app open, or hand a server a cookie file. Lurkloot asks you to install an extension in the browser where you are already logged in, which removes the part that usually breaks. The beta CLI is the part we would not rely on yet, because the README itself says Twitch claiming there is not fully verified and full discovery depends on exported credentials. The project is small and the open-issue count is not tiny, but the core pitch is clear and the privacy story is more specific than most.

Bottom Line

Lurkloot is a free, Apache-2.0 Chrome extension that farms Twitch and Kick drops from one popup, with optional Fortnite and NoPixel extension support, an Idle Watchlist, and a beta CLI plus multi-architecture Docker image. Its latest stable release is v1.18.0 from October 16, 2026, and the repository had 33 stars and 28 open issues on October 21, 2026. The extension is the supported path; Firefox builds are untested, CLI claiming is not fully verified, and automated viewing can conflict with platform terms.

Sources

jamezrin/lurkloot repository README, including the feature list, browser support, CLI caveats, privacy statement, AI disclosure, and disclaimer, verified October 21, 2026.

jamezrin/lurkloot releases, including v1.18.0 published October 16, 2026 and the v1.18.0 release notes quoted in the body.

GitHub repository metadata for jamezrin/lurkloot, verified October 21, 2026: 33 stars, 3 forks, 28 open issues, TypeScript, Apache-2.0, created May 31, 2026, last push October 21, 2026.

GrubDrops: A Self-Hosted Miner That Warns You It Is Broken

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:51:23 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: GrubDrops is a self-hosted, multi-account Twitch and Kick drop miner written in Go, and its own README still carries a blunt warning: "GrubDrops is currently not working as expected. Critical Auth Issues." The project's latest stable release is v1.4.2 from September 28, 2026, and the repository was still being pushed on October 19, so the warning sits on top of active maintenance rather than an abandoned codebase.

That combination is the story. GrubDrops has the architecture a homelab user would ask for: one Docker container, a SQLite file, a web dashboard, multiple accounts per platform, auto-claim, priority lists, and Discord alerts. It also tells you, before the first feature bullet, that authentication is not reliable right now. Most projects hide that sentence in an issue thread. This one puts it above the logo.

Image: The GrubDrops console with watch-time statistics and per-account mining. Image credit: aalejandrofer/GrubDrops, via the repository README.

The warning comes first

The README opens with two callouts. The first says drop discovery is limited because Twitch no longer lists campaigns for new logins, so GrubDrops only finds drops for games on your whitelist. The second is the authentication caution quoted above. Both are load-bearing: they change what the tool can do on a fresh install and how much trust you should place in an unattended setup.

The authentication warning is not a leftover from an old release. The README was verified on October 21, 2026, and still contains it. The repository received a push on October 19, 2026, after the last tagged release on September 28. A user cannot tell from the release page alone whether the warning refers to the published v1.4.2 build, to unreleased master, or to both. That ambiguity is itself a reason to treat the project as experimental.

The whitelist warning is more concrete. GrubDrops does not scan every campaign and offer you the catalog. A fresh install mines nothing until you add at least one game by name, either globally or per account. Until then the drops page shows a prompt and accounts sit in a "no games yet" state. This is a deliberate design choice to work around Twitch's discovery limits, not a bug, but it means the tool cannot be judged by installing it and waiting.

What the miner does

GrubDrops watches the right Twitch and Kick streams, banks the watch time, and claims drops across several accounts at once, per the README. One small self-hosted web app provides the dashboard, the account list, the campaign catalog, the priority list, and the claim history. The stack is Go, HTMX with Go templates for the interface, and SQLite for storage.

The feature list includes Twitch and Kick side by side, many accounts in one dashboard, automatic discovery and claiming, proxy support, Discord alerts, update checks, multiple languages, and optional OIDC single sign-on. The README also documents a per-account game whitelist and priority list. The console shows lifetime statistics, per-account mining state, and a live event feed. The history page keeps a claim log across every account.

Area	GrubDrops behavior per the README
Deployment	Docker image or Go 1.26+ build from source
Storage	One SQLite database, with sessions encrypted using an age X25519 key
Accounts	Multiple Twitch and Kick accounts in one dashboard
Discovery	Whitelist only, global or per account
Claiming	Automatic, with a claim log in History
Extras	Proxy, Discord webhook, OIDC, multiple languages, update checks

Two platforms, two login paths

Twitch uses the official device-code flow at twitch.tv/activate. Your password and cookies never touch GrubDrops, and the miner talks to Twitch through GraphQL and PubSub to track progress and claim rewards. That is the cleaner of the two login paths.

Kick has no public login API, so GrubDrops asks for your existing kick.com session as a cookies.txt file exported from your browser. The README walks through installing a cookie-export extension, signing in at kick.com, exporting the current site, and uploading or pasting the export into the account's Authorize page. When the session goes stale, discovery logs Cloudflare or 401 errors and you re-export.

The Kick watch path is the resource-heavy part. Kick detection and claims use an HTTP client with a Chrome TLS fingerprint, but watch time needs a real player. GrubDrops runs an on-demand Chrome sidecar per Kick account for IVS playback, created and stopped over the Docker socket. The README recommends Docker strongly for Kick, says the sidecar uses about 4 GB of RAM per account on ARM, and notes that the arm64 image is heavy. A multi-account Kick setup on a small server can consume memory quickly.

Platform	Login	Watch path	Docker need
Twitch	Device code at twitch.tv/activate	Direct HTTP, no browser	Optional
Kick	cookies.txt exported from a browser	WebSocket first, Chrome IVS sidecar fallback	Strongly recommended; sidecar needs the Docker socket

Docker, SQLite, and the Kick sidecar

The setup is one miner service with a bind-mounted data directory and a mounted Docker socket. The image runs as distroless nonroot with UID 65532, so the data directory has to be writable by that user or the database cannot be created. The README warns that a permission mistake shows up as a login failure with "failed to persist session," which is a confusing symptom for a filesystem problem.

GRUB_MASTER_KEY is required and must be a valid age X25519 identity, the kind that starts with AGE-SECRET-KEY-1. A random string will not parse, and the miner crashes at startup. The README provides a keygen command that prints a valid key, and says to reuse the same key on every restart because it encrypts the stored session tokens.

There are two security settings that deserve more attention than a default install usually gets. GRUB_AUTHBYPASS disables all authentication when set truthy, and the README marks it as such. OIDC is optional, but with no allowlist set, anyone the identity provider authenticates becomes an admin. Those are normal knobs for a self-hosted app and dangerous defaults for an exposed instance. The README also says the app is single-tenant and recommends a reverse proxy if it is exposed.

How it chooses a campaign

Each account mines one campaign at a time. When several whitelisted campaigns are eligible, GrubDrops chooses by the account's priority mode, then by the campaign closest to a claim, then by restricted team campaigns ahead of open ones, then by the live channel with the highest viewer count. Twitch also probes for a channel actually serving the target drop. A campaign with no live stream is skipped, not slept on.

The three priority modes are ordered, which follows your whitelist rank; ending_soonest, which puts the nearest deadline first; and low_avbl_first, which favors the campaign with the fewest available channels. Whitelist and priority can be set globally and overridden per account. That is a sensible model for a multi-account setup, where one account may be chasing a nearly finished campaign while another starts a new one.

The recent auth fixes

Authentication work dominates the last three releases. v1.4.0 on September 28 fixed new Twitch logins after Twitch stopped accepting device-code sign-in from the Android app client around September 18. It moved new logins to the Twitch for TV client and made each account remember which client signed it in, because Twitch accepts mismatched requests but credits no watch time for them. The same release fixed campaign discovery for TV-client accounts, stopped mining subscription-only drops, and changed an empty claim response to count as a failure instead of a success.

v1.4.1, also on September 28, stopped Kick accounts from needing re-authentication after seven days. Kick had added a seven-day expiry that nothing renewed, so every Kick account went idle a week after login. The release ignores that expiry and lets the regular auth check decide when a session has actually died. It also added retries for a startup session failure.

v1.4.2, later the same day, handled Twitch accounts on the older Android login whose token refresh is now refused. GrubDrops checks the token with Twitch first and keeps using it while it is valid, instead of marking the account idle at every restart. These are real fixes, and they explain why the repository is still active. They do not erase the README warning, which remains the first thing a new user sees.

Release	Date	Main auth change
v1.4.0	September 28, 2026	New Twitch logins move to the TV client; discovery and claim handling fixed
v1.4.1	September 28, 2026	Kick accounts no longer expire after seven days; startup retries added
v1.4.2	September 28, 2026	Older Android-login Twitch tokens keep mining while still valid

Operational limits

The first limit is the warning itself. A self-hosted miner that may fail authentication is a poor fit for unattended operation, which is the exact use case the project targets. The README's promise of set-and-forget mining is undercut by a caution banner that says it is not working as expected.

The second limit is discovery. Whitelist-only discovery means you have to know which games have campaigns and add them by name. The v1.4.0 release notes say a campaign that runs only on channels outside a game's top ten live drops channels can be missed until it appears in your inventory. That is a functional gap, not a preference.

The third is operational cost. Twitch is light, but Kick runs a Chrome sidecar per account, and the README puts the ARM sidecar at about 4 GB of RAM each. A handful of Kick accounts can outgrow a small VPS. The Docker socket mount also gives the miner control over containers, which is a meaningful privilege to hand to an application that stores session cookies.

The fourth is policy. The README says to stay within each platform's terms of service, on your own accounts, at your own risk. It does not claim platform approval, and self-hosting does not change the platform's rules.

FAQ

Is GrubDrops working right now?
According to its README, "GrubDrops is currently not working as expected. Critical Auth Issues" as of October 21, 2026. The project has shipped authentication fixes in v1.4.0 through v1.4.2, but the warning has not been removed, so treat it as experimental.

Does GrubDrops need Docker?
Twitch can run from a plain Go binary without Docker. Kick is strongly recommended to run in Docker because the Chrome IVS fallback needs the Docker socket, and the README says the sidecar uses about 4 GB of RAM per account on ARM.

How does GrubDrops log in to Kick?
It does not have a Kick login API. You export your existing kick.com session as a cookies.txt file from a browser and upload or paste it into the account's Authorize page, then re-export when the session goes stale.

Why does a fresh install mine nothing?
GrubDrops is whitelist-driven. It only discovers and mines games you add by name, either globally or per account, because Twitch no longer lists campaigns for new logins in the way the miner needs.

Does GrubDrops support multiple accounts?
Yes. The README describes many Twitch and Kick accounts in one dashboard, each with its own mining state, whitelist, and priority overrides, backed by one SQLite database.

Is GrubDrops safe to expose to the internet?
The README calls it single-tenant and recommends a reverse proxy if exposed. It also includes a setting that disables all authentication and an OIDC option where no allowlist makes every authenticated identity an admin, so the defaults need review before exposure.

Our take: We would not recommend GrubDrops to someone who wants a drop miner that just works. The README says it does not, and the warning is specific enough to take seriously. We would recommend reading it if you run a homelab and want to see how a multi-account miner is put together, because the architecture is more ambitious than most projects in this space: per-account state machines, a real discovery loop, encrypted sessions, a Kick sidecar, and a dashboard that shows what every account is doing. The honest warning is a point in the maintainer's favor, but honesty is not the same as readiness. Wait for the caution banner to come down before trusting it unattended.

Bottom Line

GrubDrops is a Go, MIT-licensed, self-hosted Twitch and Kick miner with multi-account support, a SQLite database, a web dashboard, whitelist-only discovery, and automatic claiming. Its latest stable release is v1.4.2 from September 28, 2026, and the repository had 31 stars, 8 forks, and 11 open issues on October 21, 2026. The README still warns of critical authentication issues and limited discovery, Kick accounts need a Chrome sidecar and cookies.txt, and the project is best treated as an experimental homelab build rather than a set-and-forget service.

Sources

aalejandrofer/GrubDrops repository README, including both warning callouts, the feature list, login paths, Docker guidance, priority logic, configuration notes, and responsible-use section, verified October 21, 2026.

aalejandrofer/GrubDrops releases, including v1.4.0, v1.4.1, and v1.4.2 published September 28, 2026, and their release notes quoted in the body.

GitHub repository metadata for aalejandrofer/GrubDrops, verified October 21, 2026: 31 stars, 8 forks, 11 open issues, Go, MIT, created June 4, 2026, last push October 19, 2026.

StreamDropCollector: A Windows Miner With a Nightly Build Channel

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:51:21 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: StreamDropCollector is a Windows desktop drop miner for Twitch and Kick, built with C# and WPF on .NET 10, and its latest stable release is v1.4.5 from October 12, 2026. The repository also published a fresh nightly build from commit 3599699 on October 21, which makes it one of the few miners in this group that ships a continuous build channel alongside tagged releases.

The project is a Windows-only alternative to the cross-platform Python and browser-based miners. It runs both platforms at once, logs in through embedded WebView2 browsers, tracks campaigns in a dashboard, and claims rewards automatically. On GitHub it had 32 stars, 4 forks, and 3 open issues on October 21, 2026, with MIT-licensed C# source and a repository created on November 14, 2025.

Image: The StreamDropCollector dashboard with live account progress. Image credit: tsgsOFFICIAL/StreamDropCollector, via the repository README.

What StreamDropCollector does

The README's short version is that the app mines streams in the background, earns campaign rewards, and claims them automatically. The feature list is concrete: Twitch and Kick at the same time, automatic selection of the best active campaign and streamer, support for general category drops, auto-claiming, a lowest quality mode, mature content bypass for age gates, live percentage and channel tracking, a dark mode WPF interface, background operation, and an optional nightly update channel.

The lowest quality mode is the bandwidth feature. It sets streams to the minimum quality so the app barely touches your bandwidth or CPU, per the README. The mature content bypass handles age gates automatically. The dashboard shows each account mining in parallel with its own live progress and network usage, according to the screenshot caption in the repository.

Capability	StreamDropCollector behavior
Platforms	Twitch and Kick, simultaneously
Operating system	Windows 10 or 11, 64-bit only
Framework	.NET 10 with a WPF interface
Campaign selection	Best active campaign and streamer, including general category drops
Claims	Automatic when a reward is ready
Playback	Lowest quality mode to reduce bandwidth and CPU use
Updates	Stable releases, with an optional nightly channel
License	MIT

A Windows-first design

The requirements are short and strict: Windows 10 or 11 64-bit, plus a Twitch or Kick account. There is no Linux or macOS build. The README tells contributors to test on Windows before submitting a pull request and states plainly that WPF and WebView2 have no Linux or macOS builds. If you run a homelab server, this is not the miner for that machine.

That constraint buys a native desktop experience. The app is a single WPF window with a dark mode interface, responsive cards, and a dashboard rather than a browser tab or a terminal. It can run in the background while you use the machine. It does not require Docker, Python, or a separate browser, though it does require the .NET runtime unless you use a self-contained build.

The release page offers a Setup.exe for a guided install and both framework-dependent and self-contained ZIPs for a portable copy. The README's quick start describes extracting the ZIP to the appdata folder and running the executable, then signing in when the embedded browsers appear. The self-contained option matters on a clean Windows machine where installing the .NET 10 desktop runtime is an extra step.

WebView2 logins and the Helix split

Drops logins happen inside embedded WebView2 browsers, the same engine Edge and Chrome use, per the README. That means you sign in on Twitch and Kick in a browser window controlled by the app, not by pasting cookies into a text file. The README says credentials never leave your machine outside the WebView2 engine and local token storage, and it points to the encrypted WebView2 profile and helix-auth.dat file under the appdata folder.

Twitch gets a second, separate login. A one-time Helix device-code authorization is used only for live stream metadata. The README is explicit that this is not connected to the drops login and that the data still goes only to Twitch. The split exists because the app needs Twitch metadata to choose a stream, and it keeps that permission separate from the drop session.

This is still a local application holding session tokens. The README says the token stores are encrypted and never leave the device, and that there are no third-party APIs or services: all network traffic goes only to Twitch and Kick, including Helix, GraphQL, EventSub, and Kick's own REST APIs. Those are project claims, not an independent audit, but they are specific enough to check against the code.

Stable releases and nightly builds

The stable channel is the one to use. v1.4.5 was published on October 12, 2026, with a Setup.exe and two ZIP variants. The release notes are short and point to the installation guide, Discord, GitHub issues, and a Full Changelog compare link rather than listing changes. The repository was pushed on October 21, 2026, so development has continued since that tag.

The nightly channel is built from every push to master and republished under a nightly tag. On October 21, 2026, the nightly was version 1.4.7-nightly.3599699, built from commit 3599699 and published at 11:07 UTC. The README says nightly builds are untested beyond CI passing and that you should expect the occasional rough edge, and it recommends a regular release if you want something stable. The version number includes the short commit hash, so you can tell exactly what you are running.

Channel	Version on October 21, 2026	Published	What it means
Stable	v1.4.5	October 12, 2026	Tagged release, recommended by the README
Nightly	1.4.7-nightly.3599699	October 21, 2026	Automated build from commit 3599699, untested beyond CI

A nightly channel is a good sign for maintenance and a bad default for users. It means fixes can reach testers quickly, and it also means the newest code is not the version the README recommends. If you install this app, turn the beta channel off unless you are deliberately testing a fix.

Local tokens and no telemetry

The privacy section is more detailed than most. The app uses no third-party APIs, sends traffic only to Twitch and Kick, keeps credentials in WebView2 and helix-auth.dat, and collects no analytics or telemetry, per the README. Logs live under the appdata folder. There is no account system and no vendor server in the described flow.

The caveat is the same one that applies to every tool in this category: the app needs authenticated sessions to watch and claim on your behalf. Local storage reduces the number of parties involved, but it does not remove the platform's ability to detect automated viewing. The README says the tool is for personal use, tells users to respect Twitch and Kick's terms of service, and warns that automated viewing may violate platform rules in some contexts. That warning should be read before the feature list, not after.

The limits of a young Windows app

The first limit is the platform. Windows 10 or 11 64-bit only is a hard boundary, and there is no server or Docker path. The second is maturity. The repository was created in November 2025, the stable release is nine days old as of October 21, 2026, and the issue count is small enough that most bugs may still be undiscovered rather than absent.

The third is the nightly temptation. A continuous build channel is useful, but the README itself says nightly builds are untested beyond CI. The version string is honest about the commit, which helps with bug reports, but it does not make the build safer for daily use.

The fourth is the nature of the automation. The app claims rewards automatically and bypasses age gates, which are exactly the behaviors platforms watch. A small star count and a short release history are not evidence of safety. They are evidence that the project has not been stress-tested by a large audience yet.

FAQ

Does StreamDropCollector run on Linux or macOS?
No. The README requires Windows 10 or 11 64-bit, and it states that WPF and WebView2 builds for Linux and macOS do not exist.

Can StreamDropCollector mine Twitch and Kick at the same time?
Yes. Dual-platform support is the first feature in the README, and the dashboard describes every Twitch and Kick account mining in parallel with its own progress.

Why does StreamDropCollector ask for a second Twitch login?
It uses a separate one-time Helix device-code authorization for live stream metadata. The README says this is not connected to the drops login and that the data still goes only to Twitch.

Does StreamDropCollector collect telemetry?
No, per the README. It says there are no analytics, telemetry, or data collection by the app, and that logging stays local in the appdata logs folder.

Should I use the stable release or the nightly build?
Use the stable v1.4.5 release unless you are deliberately testing. The README says nightly builds are untested beyond CI passing and recommends a regular release for stability.

Does StreamDropCollector claim rewards automatically?
Yes. The README lists auto-claiming as a feature that detects and claims ready rewards instantly, alongside live progress tracking for each campaign.

Our take: StreamDropCollector is the most conventional desktop app in this set, and that is its appeal. It does not ask for Docker, a Python environment, or a cookie export. It gives Windows users a native window, embedded logins, a dashboard, and automatic claiming, then backs it with a stable release and a nightly channel. The tradeoff is that it is Windows-only and still young. We would use the stable build, keep the beta channel off, and treat the first few months of releases as a public test period rather than a proven track record.

Bottom Line

StreamDropCollector is an MIT-licensed C# and WPF drop miner for Windows 10 and 11, built on .NET 10, with simultaneous Twitch and Kick support, WebView2 logins, automatic claiming, lowest quality playback, and a dashboard for per-account progress. Its stable release is v1.4.5 from October 12, 2026, and a nightly build from commit 3599699 arrived on October 21, 2026. It is a polished Windows-only option with a short track record, no telemetry per its README, and the usual platform-terms risk that comes with automated viewing.

Sources

tsgsOFFICIAL/StreamDropCollector repository README, including the feature list, privacy and safety section, requirements, quick start, nightly build warning, and important notes, verified October 21, 2026.

tsgsOFFICIAL/StreamDropCollector releases, including v1.4.5 published October 12, 2026 and the nightly build 1.4.7-nightly.3599699 published October 21, 2026.

GitHub repository metadata for tsgsOFFICIAL/StreamDropCollector, verified October 21, 2026: 32 stars, 4 forks, 3 open issues, C#, MIT, created November 14, 2025, last push October 21, 2026.

Rust-Droper: A Tray App That Claims Rust Drops by Default

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:51:00 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Rust-Droper is a Windows tray app that collects Rust Twitch and Kick drops, and its auto-claim setting is on by default, including for irreversible in-game items. The first stable release, v1.0.0, was published on October 21, 2026, the same day the repository was created, and the README is unusually candid that game-delivered item claims are only just being tested and Kick claiming has not been tested live.

This is a narrow tool with a narrow promise. It does not mine every game, it does not manage a farm of accounts, and it does not try to be a hosted service. It follows Rust campaigns, keeps one Twitch and one Kick account watching, shows progress in a tray window, and claims finished rewards. The README says the design choice is deliberate: Rust only, one account per platform, for lower risk and a simpler app.

Image: The Rust-Droper dashboard in English. Image credit: Pinhan/Rust-Droper, via the repository README.

Auto-claim is the headline

Most drop miners treat claiming as a convenience. Rust-Droper treats it as a default that can consume a reward before you have decided what to do with it. The README states the warning directly: auto-claim also claims irreversible in-game items, it is on by default, and it can be turned off in Settings.

That sentence should decide whether you install the app. A Twitch badge or emote can be claimed and ignored. An in-game Rust item is the kind of reward the README calls irreversible, and auto-claim accepts it before you inspect it. If you want to see a drop before accepting it, turn auto-claim off before the first campaign completes. The setting exists, but the default does not wait for you.

The claim path is also the least proven part of the app. The README's honest status section says automatic claiming is verified against Twitch's live API for Twitch-native rewards such as badges and emotes. Claiming game-delivered Rust items, the Steam and Facepunch kind, is only just being tested on a real Rust campaign. A fallback exists in the form of a claim it manually notification plus a link, but the README says to expect rough edges in the first days. Kick drop progress is verified live, while Kick claiming has not been tested live yet.

Claim type	Status per the README
Twitch-native rewards such as badges and emotes	Automatic claiming verified against Twitch's live API
Game-delivered Rust items through Steam and Facepunch	Only just being tested on a real Rust campaign; manual claim fallback with a notification and link
Kick drop progress	Verified live
Kick claiming	Not tested live yet

One game, two platforms

Rust-Droper follows Rust's Twitch and Kick drop campaigns automatically, including general drops and streamer drops. The v1.0.0 release notes list a filter that follows only campaigns giving Rust rewards, with an option to turn it off in Settings. The interface has a separate screen per platform with a Twitch and Kick switch.

The account model is intentionally small. The README's FAQ says the app will not add other games or multiple accounts: Rust only, one Twitch plus one Kick account, by design. That is a narrower scope than the multi-account miners in this group, including GrubDrops. The tradeoff is a smaller blast radius and a simpler state machine.

The interface includes drop cards with images, progress, and states such as queued, watching, ready to claim, and claimed. There is a claimed rewards section, drag-and-drop priority order for streamer drops, and a health check page for internet, sign-in, the Facepunch and Steam link, campaign discovery, the engine, and sleep prevention. The v1.0.0 release notes also list an event card with the Facepunch event name, dates, and countdown. The tray icon and notifications keep it out of the way.

How it watches without a browser

Twitch is watched through the real stream feed at the lowest quality, with no browser window, at about 2 MB per minute, per the README. The app requests the stream's HLS playlist and segments like a player. Progress is reconciled with Twitch's Inventory and DropCurrentSession endpoints, and a watchdog switches channels when progress stalls.

Kick needs a real player, so the app uses a hidden, muted WebView2 window. Discovery reads campaigns from Twitch's GraphQL API and Kick's public API, and only follows campaigns that give Rust rewards. Claiming uses Twitch PubSub drop-claim events with a fallback that builds the claim ID itself, and Kick through its claim endpoint.

The app also tries to keep the machine awake. A sleeping PC cannot make drop progress, so Prevent sleep while running is on by default and can be turned off. The screen may still turn off. Crash recovery resumes after an unexpected close, and the app recovers after internet outages and PC sleep.

What is verified and what is not

The README separates the two in a way that is rare for a launch post. Twitch-native claiming is verified. Game-delivered Rust items are in testing. Kick progress is verified. Kick claiming is untested. The app has a fallback for the game-delivered case, but a fallback is not the same as a completed claim path, and the README says so.

The task the app is solving is also harder than it looks. Rust drops can be Twitch-native cosmetics or items delivered through the game's own account link. The README tells you to link Steam and Twitch once at twitch.facepunch.com/connect so items reach your Steam inventory. If that link is missing, a claim may succeed on Twitch and still not produce the item you expected. The health check page includes the Facepunch and Steam link for that reason.

The app was designed after studying TwitchDropsMiner, TwitchDropsBot, and StreamDropCollector, per the README. It says no code was copied and credits them as inspiration and analysis in a third-party notices file. That is a normal lineage for a new miner, and it is worth stating because the space is small and the same platform quirks recur.

Download, SmartScreen, and resources

The download is a portable ZIP from the releases page, not an installer. You unzip it anywhere and run RustDroper.exe. The README warns not to use the green Code download button or the Source code ZIP files, because those contain source and no executable. The build targets Windows 10 and 11, 64-bit. Kick needs the Microsoft Edge WebView2 Runtime, which is already present on current Windows 10 and 11.

The executable is not code-signed, so Windows SmartScreen may show Windows protected your PC, and some antivirus products may flag a tray app that automates a website. The README calls that a common false positive for unsigned, self-contained automation tools, and offers two ways to check: every release has a SHA256SUMS.txt file, and the source can be built locally with the .NET 10 SDK. The release includes RustDroper-1.0.0-win-x64.zip and SHA256SUMS.txt.

Resource	Figure per the README
RAM, Twitch only, in the tray	About 60 MB
RAM, Kick active	About 0.6 to 0.8 GB, plus a fraction of one CPU core
Twitch bandwidth	About 2 MB per minute at the lowest quality
Sleep behavior	Prevents sleep while running by default; the screen may still turn off
Storage	%LOCALAPPDATA%\RustDroper for settings, logs, and encrypted sign-in
Requirements	Windows 10 or 11, 64-bit; WebView2 Runtime for Kick

A day-one project with a short history

Rust-Droper had 2 stars, 0 forks, and 0 open issues on October 21, 2026, the day of its first release. The repository and the release were both created that day. That is not a criticism of the code; it is a statement about evidence. There has been no time for a user base to find the edge cases, and the README already names several.

The day-one status also means the download counts are tiny. The release assets showed 2 downloads for the Windows ZIP and 2 for SHA256SUMS at the time of the GitHub API check. Those numbers will change, but they describe a project that has not been exercised by a crowd.

The MIT license and public source lower the trust problem: you can read the code or build it yourself. They do not lower the platform-terms risk. The README says Twitch's and Kick's terms of use may restrict automation, a warning or restriction on your account is possible, and there is no guarantee of any kind. It also says the app is not affiliated with Twitch, Kick, Facepunch, or Rust.

FAQ

Does Rust-Droper support games other than Rust?
No. The README says Rust only, no other game, and the FAQ says other games will not be added. It follows Rust's Twitch and Kick drop campaigns specifically.

Does Rust-Droper claim rewards automatically?
Yes, auto-claim is on by default, and the README warns that it also claims irreversible in-game items. You can turn it off in Settings before a campaign completes.

Does Rust-Droper support multiple accounts?
No. It supports one Twitch account and one Kick account, which the README describes as a deliberate design choice for lower risk and a simpler app.

Has Kick claiming been tested?
Kick drop progress is verified live, but Kick claiming has not been tested live yet, per the README's honest status section. Twitch-native claiming is verified, while game-delivered Rust item claiming is still being tested.

What does Rust-Droper require on Windows?
Windows 10 or 11, 64-bit, and the Microsoft Edge WebView2 Runtime for the Kick player. The app is a portable ZIP with no installer, and the executable is not code-signed.

How much memory does Rust-Droper use?
About 60 MB in the tray with Twitch only. With Kick active, the README says to expect roughly 0.6 to 0.8 GB of RAM and a fraction of one CPU core.

Our take: Rust-Droper is the clearest example in this group of a tool that should be judged by its caveats, not its feature list. The scope is honest, the resource numbers are specific, and the README tells you which claim paths are verified and which are not. The auto-claim default is the part we would change. A day-one app that can irreversibly claim in-game items should make the user opt in, not opt out, especially while the game-delivered claim path is still being tested. If you play Rust and want a narrow tray tool, this is worth watching. If you want something to run unattended on valuable items today, wait for the claim path to settle.

Bottom Line

Rust-Droper is an MIT-licensed Windows tray app for Rust Twitch and Kick drops, with one account per platform, Twitch watching through the HLS feed at about 2 MB per minute, Kick watching through a hidden WebView2 player, and auto-claim on by default. Its first stable release, v1.0.0, arrived on October 21, 2026, with 2 stars at check time. Twitch-native claiming is verified, game-delivered Rust item claiming is still being tested, Kick claiming has not been tested live, and the unsigned executable can irreversibly claim in-game items unless you turn auto-claim off.

Sources

Pinhan/Rust-Droper repository README, including the warning section, honest status, feature list, download and SmartScreen notice, quick start, architecture summary, FAQ, resource figures, build instructions, and license, verified October 21, 2026.

Pinhan/Rust-Droper releases, including v1.0.0 published October 21, 2026, its release notes, RustDroper-1.0.0-win-x64.zip, and SHA256SUMS.txt.

GitHub repository metadata for Pinhan/Rust-Droper, verified October 21, 2026: 2 stars, 0 forks, 0 open issues, C#, MIT, created October 21, 2026, last push October 21, 2026.

TwitchDropsMiner: The 3,380-Star Veteran Has No Stable Download

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:50:58 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: TwitchDropsMiner is the veteran of open-source drop miners, with 3,380 stars and 390 forks, but its latest stable tag is still v15 from December 25, 2022. The project's own v15 release notes say it has been "NOT STABLE since February 2023" and point users to the latest master or pre-release build, and the current development build was published on October 9, 2026.

That gap is the whole practical story. The repository is not abandoned: it was pushed on October 9, 2026, and the development build carries a reference commit from that day. But the v15 release page no longer offers a download at all. Its notes say downloads were removed to prevent reports because the build does not work anymore, and the release has no assets. The only current downloadable build is the development pre-release, with all the instability that label implies.

Image: The TwitchDropsMiner main window. Image credit: DevilXD/TwitchDropsMiner, via the repository README.

The most-adopted miner in the group

TwitchDropsMiner was created on December 3, 2021, and it has the adoption numbers to show for it: 3,380 stars, 390 forks, and 73 open issues on October 21, 2026, per GitHub. The next most-starred project in this set has 82 stars. That does not automatically make it the best choice, but it does mean more people have run it, broken it, and reported what broke.

The application AFK mines timed Twitch drops without downloading stream video or sound. Every several seconds it fetches stream metadata, which the README says is enough to advance drops, and it keeps channel status current through a websocket connection that receives events about streams going up or down and viewer counts changing.

Repository fact	Value on October 21, 2026
Stars	3,380
Forks	390
Open issues	73
Language	Python
License	MIT
Created	December 3, 2021
Last push	October 9, 2026
Latest stable	v15, December 25, 2022
Latest pre-release	Development build, October 9, 2026

Why the stable release is not the one to use

The v15 release page is unusually direct. Its notes begin with a warning that the build has been not stable since February 2023, and tell users to use the latest master or pre-release version while waiting for v16. The same notes describe v15 as the long-awaited release that introduced a new SmartTV account linking login method, fixed campaign ACL handling, fixed missed short campaigns, and added a Portuguese translation, among other changes. The notes end with a line saying downloads were removed to prevent reports from this version, as it does not work anymore. The v15 release has no assets.

The development build published on October 9, 2026 is an automatically generated pre-release that includes the latest master branch changes. Its own notes warn that the build is not stable and may terminate with a fatal error. It records the last build date, 2026-10-09 07:55:14 UTC, and the reference commit f693b0ca41301ac2282f3978713479efebe0490e. There is no newer tagged stable release.

Build	Published	Status	What the project says
v15, stable tag	December 25, 2022	Old stable	Not stable since February 2023; downloads removed because it does not work anymore
Development build	October 9, 2026	Pre-release	Automatically generated from master; not stable and may terminate with a fatal error

This is not a case of a project quietly forgetting to mark a release. The maintainer has kept the warning in the release notes and published a development build on October 9, 2026. The practical advice is simple: if you install TwitchDropsMiner today, there is no current stable download, only a pre-release the author says may crash. The old stable tag remains as a label, not as an installable build.

Stream-less mining and 199 channels

The core technique is why the project earned its following. TwitchDropsMiner pretends to watch a stream by fetching its metadata, without downloading the video or audio. The README says this is enough to advance drops and saves bandwidth. A sharded websocket connection tracks up to 199 channels at the same time, so the miner can watch for a higher-priority game coming online without opening a player for every channel.

The feature list includes game priority and exclusion lists, automatic campaign discovery based on linked accounts, stream tag and campaign validation to avoid mining a stream that cannot earn the drop, automatic channel switching when the current stream goes offline or a higher-priority game comes online, a saved login session in a cookies file, and automatic start and stop as campaigns appear and finish.

Account linking is still your job. The README tells you to link your Twitch account to game accounts on the Twitch drops campaigns page. The miner discovers campaigns based on the links that already exist; it does not create them. If a campaign requires a link you have not made, it is not a campaign the miner can complete for you.

What the project refuses to do

The project goals section is as important as the feature list. TwitchDropsMiner is designed as a Twitch drops tool, easy for an average person, packaged as a ready-to-run executable, intended to start with the PC and run in the background, and to require minimal attention. It is not intended for unattended operation, 100 percent uptime, remote server hosting, multiple managed accounts, mining campaigns the account is not linked to, or mining anything besides Twitch drops.

The README goes further and lists features that are most likely never going to exist: running without a GUI or with only a console, automatic restart after an error, Docker or any other remote deployment, using more than one managed account, mining unlinked campaigns, anything that increases site processing load, and additional notification systems beyond what is implemented. It also says channel points are out of scope.

That refusal is a design statement, not a bug report. It means the many users who want a 24/7 Docker miner on a VPS should not choose this project and then complain that it lacks Docker. It also means the app expects a desktop session, a GUI, and a human who checks it once or twice a day.

Windows, Linux, and macOS builds

The Windows build is a PyInstaller executable. The README warns that some antivirus engines, including Windows Defender, may report the packaged executable as a trojan because PyInstaller has been used to package malicious Python code in the past. It says those reports can be ignored, and that anyone who does not trust the executable can install Python and run from source. The executable uses the temp directory for runtime files and the executable's directory for persistent files. Autostart is a registry entry under the current user, and moving the app breaks it until you toggle the option off and on.

Linux is distributed as an AppImage and a PyInstaller build. The AppImage is the recommended format. It should work on any modern distribution with glibc 2.35 or newer and a working display server. The Linux build is significantly larger because it includes gtk3 and its dependencies for tray and notification support. Running the Windows build through Wine is also described as working well.

macOS is a PyInstaller app bundle in a ZIP. It is not signed with a paid Apple Developer certificate, so Gatekeeper will block it on first run with a message that the application is damaged and cannot be opened. The README provides the xattr command to clear the quarantine flag. Persistent files such as cookies.jar, settings.json, lock.file, and the cache folder live inside the app bundle under Contents/MacOS.

Cookies, account conflicts, and risk

TwitchDropsMiner stores persistent cookies in cookies.jar, and it restores authorization from that file on each run. The README warns that the file can give another person access to your Twitch account even without your password. Treat it like a password, not like a cache.

The README also warns that watching a stream in a browser, or by any other means, on the same account the miner is using will usually cause the miner to misbehave, report false progress, and get stuck on the current drop. Using the same account to watch other streams during mining is discouraged. That is a practical conflict: the miner expects to be the only viewer on that account while it works.

A successful login may trigger a New Login notification email from Twitch. The README says this is normal, the detected browser will be Chrome, and you can verify the login came from your own IP address. The app presents itself as Chrome to the Twitch server, which is why the notification says that.

The project does not claim platform approval. It is a third-party automation tool, and Twitch's terms and anti-bot systems can change. The README's own goals include keeping interactions with Twitch to the minimum required for reliable operation, at a level achievable by a diligent site user, but that is a design target rather than a guarantee.

FAQ

Is the latest stable TwitchDropsMiner release current?
No. The latest stable tag is v15 from December 25, 2022, its own release notes say it has been not stable since February 2023, and its downloads have been removed because it does not work anymore. The current development build was published on October 9, 2026.

Does TwitchDropsMiner support Kick?
No. It is Twitch-only. The README explicitly says it will not add support for another site that relies on watching Twitch streams, and its feature list is built around Twitch campaigns.

Can I run TwitchDropsMiner on a server or in Docker?
No. The project goals list remote server hosting, Docker, remote deployment, running without a GUI, and automatic restart as things it does not intend to support.

Does TwitchDropsMiner support multiple accounts?
No. It is designed for one managed account, and the README lists using it with more than one managed account as outside the project's goals.

Does TwitchDropsMiner mine channel points?
No. The README says the project is about drops only and lists channel points as something it is not intended to mine.

How does TwitchDropsMiner avoid downloading stream video?
It fetches stream metadata every several seconds, which the README says is enough to advance drops, and uses a websocket for channel status. It does not download the actual video or audio.

Our take: TwitchDropsMiner still deserves its reputation as the reference implementation. Stream-less mining, 199-channel tracking, priority lists, and campaign validation are the features every newer miner compares itself against, and the project's refusal to become a multi-account Docker service is a coherent design choice rather than neglect. The problem is the install experience. A first-time user sees a stable tag from 2022 and has to read the release notes to learn that the tag is the one not to use. The maintainer should either mark v15 as deprecated on the release page or publish a current stable. Until then, the honest description is a mature codebase distributed through an unstable channel.

Bottom Line

TwitchDropsMiner is an MIT-licensed Python Twitch drop miner with stream-less metadata mining, tracking for up to 199 channels, game priority and exclusion lists, campaign validation, automatic channel switching, and saved cookies. It had 3,380 stars and 390 forks on October 21, 2026, but its latest stable tag is v15 from December 25, 2022, the project itself says that build has been not stable since February 2023, and its downloads have been removed because it does not work anymore. The working path is the October 9, 2026 development pre-release, which the project warns may crash. It is Twitch-only, one account only, and explicitly not built for Docker, remote hosting, channel points, or unattended 24/7 operation.

Sources

DevilXD/TwitchDropsMiner repository README, including the stream-less mining explanation, feature list, usage notes, warnings, platform-specific build notes, project goals, and credits, verified October 21, 2026.

DevilXD/TwitchDropsMiner releases, including v15 published December 25, 2022 and the development build published October 9, 2026 with reference commit f693b0ca41301ac2282f3978713479efebe0490e.

GitHub repository metadata for DevilXD/TwitchDropsMiner, verified October 21, 2026: 3,380 stars, 390 forks, 73 open issues, Python, MIT, created December 3, 2021, last push October 9, 2026.

Pear Drops: A New Kick Miner That Watches With the Real Player

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:50:56 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Pear Drops is a Kick desktop app that collects emote and badge drops by running Kick's real player instead of sending watch events itself. The repository was created on October 17, 2026, and its latest release, v0.1.17, arrived on October 21. At check time the project was less than four full days old and had zero stars.

That age is the reason to be careful, not the reason to ignore it. Most miners in this space either scrape an API or automate a browser. Pear Drops takes a third path: it loads the actual kick.com channel page in an Electron window, keeps the player running, and lets Kick's own player report the watch time. The README says the app never sends watch events itself, so the traffic matches a normal viewer's. Whether that approach holds up is an open question, but it is a more interesting design than another Selenium wrapper.

Image: The Pear Drops social preview artwork. Image credit: StreamOrchard/pear-drops, via the repository README and .github assets.

The real-player approach

Kick's web player sends a tracking.user.watch.livestream event over its WebSocket about every two minutes while the video is playing. Each credited event adds two minutes to a campaign's progress, per the README. Pear Drops relies on that behavior by running the real kick.com channel page, so the page produces the events and the app only manages which channel is playing.

The app never sends those events itself. That is the key distinction from a miner that talks directly to an API. Pear Drops is not pretending to be a player at the network layer; it is running a player and staying out of the way. The README says the traffic matches a normal viewer's, which is the project's claim about why the approach should be less fragile than a hand-built client.

The visual side is deliberately quiet. The window is muted, but the player stays unmuted inside the page because browsers throttle muted background video, per the README. Muting at the operating system window level keeps the audio out of your speakers without pausing the playback that Kick credits. That is a small detail with a real technical reason behind it.

What Kick relies on	How Pear Drops handles it
Watch events about every 2 minutes	Runs the real kick.com player and lets the page send them
One stream credited per account at a time	Runs one stream and warns when another player takes the credit
Muted playback can be throttled by the browser	Mutes the window, keeps the player unmuted inside the page
Streams end, raid, freeze, or change category	Detects the event and switches or reloads the player

Emote and badge drops first

Pear Drops is built for Kick emote and badge drops. It finds a live stream that counts for each drop, with channels you follow first, watches it, claims rewards when they unlock, and moves on to the next category when one is done. Game drops that need a linked game account are skipped unless you turn them on.

The queue groups every active emote and badge campaign by category, with live stream counts and a bar of minutes watched plus each reward's icon where it unlocks. You can drag to reorder, switch categories off, or let Smart order sort the list. Smart order puts what runs out of time first and tries to catch drops that are rarely live while they are on. It learns how often each category has a stream over the past week, and Settings controls how much that history counts and how readily the app switches.

The Done tab keeps finished drops and missed drops that ended with rewards left. Each drop's details, last known progress, and artwork are saved locally, so they remain after Kick removes the campaign. If Kick lists a drop again with a new end date, it goes back into the queue. That archive is a nice touch for anyone who has watched a campaign disappear before they could check what they missed.

Queue, smart order, and recovery

The recovery list is the most practical part of the README. Pear Drops notices streams that end or raid within seconds using the player's video ending, Kick's own stream stopped and raid events, and the player moving on. It checks a frozen picture with Kick within seconds and reloads after a minute. It reloads and then switches streams after a set time without credit, 20 minutes by default. It handles offline, banned, or crashed streams, streamers changing category, and sleep and wake.

The dashboard shows what is wrong at once. The LIVE chip turns Frozen, Buffering, or Checking, and a notice says what happened and where the app went next. There is a live monitor of watch events versus credits, a credited-minutes chart with a forecast to each reward, session stats, and a stream and chat dock beside the dashboard. The tray icon is a progress ring, and its popover shows the current drop, ETAs, and quick controls.

That is a lot of interface for a project this young. It is also a warning about maturity: when a project ships this much dashboard before it has a user base, the risk is not a lack of features but a lack of time in the field. The README documents edge cases well, which suggests the author tested them, but documentation and a track record are different things.

The one-stream conflict alarm

Kick credits only one stream per account at a time. If you watch a Kick stream in another tab, a multiview, or the phone app while Pear Drops is farming, the other player can take the credit and Pear Drops will stall. The app watches for the pattern: the player keeps sending watch events, but none are credited. It then chimes and shows a banner explaining that another Kick stream on the account is taking the credit.

That is a better failure mode than silently watching a stream for an hour with no progress. The README also says the app can link with Stream Multiview on the same computer through a local Grapevine connection, and while it farms, multiview can sign its Kick players out so they do not take the drop credit. The link is off until you switch it on in both apps, and the first pairing requires an Allow action on the card so another program cannot pair as Stream Multiview by itself.

The connected-app feature is optional and local. It is not a cloud service, and the README says Pear Drops tells the other app what it is farming, never your login. Pairing keys live in grapevine.json and are encrypted with the system keychain. For a new app, that is a thoughtful boundary between convenience and credential sharing.

Local data and encrypted login

The Kick login is stored in the app's own browser session on the computer. It is encrypted with the system keychain, macOS Keychain or Windows DPAPI, through Electron's cookie-encryption fuse and a hardening script that runs at start, per the README. The login token is read inside the kick.com page and never passes through the app's own code or IPC. Settings includes a Forget my login when I quit option, and if the app crashed, it does that at the next start instead.

Everything else stays in the app's data folder: settings.json, a pear-drops.db database holding history, the drop archive, and a startup cache, and drop artwork in drop-images. The database is backed up daily, weekly, or not at all depending on Settings, keeping the last 3, 7, or 14 copies, and before any update that changes storage. If the database is damaged, the app restores the newest good copy and says so in History.

The README does not describe a vendor account or cloud sync. It does describe a local-only socket for connected apps, encrypted pairing keys, and a page guard that limits which windows can send messages to the app. Those are project claims, not an independent security audit, and the app is unsigned, so the usual caution applies.

Installers, source, and zero stars

Windows 10 and 11 64-bit users get a Pear-Drops-Setup exe. Mac users on macOS 13 or newer, Apple silicon or Intel, get a dmg. Linux users run from source with Node.js 24 or newer, using npm install and npm start, or the start-mac-linux.sh script. There is also npm run demo, which runs against a stand-in Kick with made-up campaigns, channels, and a week of history, so you can try the interface without an account.

The app is not signed with a paid Apple or Microsoft certificate yet, so both platforms warn on first launch. The README walks through the Windows SmartScreen More info and Run anyway path, and the macOS Privacy and Security Open Anyway path. It also explains the macOS keychain prompt after an update and the Always Allow option that keeps the saved login decryptable.

The release history is five versions across two calendar days, from v0.1.13 at 04:38 UTC on October 20 to v0.1.17 at 01:16 UTC on October 21. The changes are mostly daily-reward and tray polish: v0.1.13 fixed the chest animation, v0.1.14 added queue rows for channel-limited drops, v0.1.15 added the hopping chest, v0.1.16 matched the macOS tray icon to the menu bar, and v0.1.17 put the chest in a filling ring. Those are polish releases, not feature milestones, which fits a project that is still finding its footing. GitHub reported 0 stars, 0 forks, and 0 open issues on October 21, 2026, and the repository was created on October 17, 2026. The GPL-3.0 license is clear, but adoption is not there yet.

Repository fact	Value on October 21, 2026
Created	October 17, 2026
Latest release	v0.1.17, October 21, 2026
Stars, forks, open issues	0, 0, 0
Language	JavaScript with Electron
License	GNU General Public License v3.0
Installers	Windows 10 or 11 64-bit; macOS 13 or newer
Linux	From source with Node.js 24 or newer
Last push	October 21, 2026

FAQ

Does Pear Drops collect game drops?
It focuses on Kick emote and badge drops. Game drops that need a linked game account are skipped unless you turn them on, per the README.

Does Pear Drops send fake watch events to Kick?
No. The README says the app never sends watch events itself. It runs the real kick.com channel page and lets Kick's own player report the watch time.

Why does Pear Drops keep the player unmuted?
Browsers throttle muted background video, so the app mutes the window while leaving the player unmuted inside the page. That keeps playback running without sending audio to your speakers.

Can I use Pear Drops while watching another Kick stream?
Not on the same account. Kick credits one stream per account at a time, and the app will warn you when another player is taking the credit. Close other Kick players before farming.

Does Pear Drops run on Linux?
Only from source, with Node.js 24 or newer. The prebuilt installers are for Windows 10 or 11 and macOS 13 or newer.

Is Pear Drops safe to install?
It is GPL-3.0 source, but it is less than four days old, has zero stars, and is unsigned, so both Windows and macOS warn on first launch. Read the source or wait for a track record before trusting it with an account.

Our take: Pear Drops has the most interesting technical bet in this group. Running Kick's real player avoids the two weakest parts of a miner: faking watch events and keeping a headless browser from being detected. The conflict alarm and local-only connected-app design show the author thought about failure modes, not just features. But less than four days of age and zero stars is not enough evidence to call it reliable. We would try the demo mode, read the privacy section, and wait for a few weeks of releases before putting a real Kick account behind it.

Bottom Line

Pear Drops is a GPL-3.0 Electron desktop app that collects Kick emote and badge drops by running Kick's real player, muting the window, claiming rewards, and switching channels when a stream ends or stalls. It supports Windows 10 and 11, macOS 13 and newer, and Linux from source, with local data, encrypted login storage, a one-stream conflict alarm, and optional local linking to Stream Multiview. Its latest release is v0.1.17 from October 21, 2026, the repository was created on October 17, 2026, and it had zero stars at check time. The approach is promising, but the project is far too young to treat as proven.

Sources

StreamOrchard/pear-drops repository README, including the install guide, feature list, Kick drop mechanics, privacy and security section, development commands, limitations, disclaimer, and GPL-3.0 license, verified October 21, 2026.

StreamOrchard/pear-drops releases, including v0.1.17 published October 21, 2026 and the v0.1.15, v0.1.16, and v0.1.17 tray icon changes.

GitHub repository metadata for StreamOrchard/pear-drops, verified October 21, 2026: 0 stars, 0 forks, 0 open issues, JavaScript, GPL-3.0, created October 17, 2026, last push October 21, 2026.

Hordes of Azeroth: A Survivors Game Mode That Runs Inside WoW 1.12.1

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:48:44 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Hordes of Azeroth is a survivors-style game mode built on World of Warcraft 1.12.1, with an endless horde, spells that fire on their own, and a level-up choice each time you gain a level. It runs on a server on your own PC, and the first public build arrived on October 20, 2026, followed by v0.1.1 on October 21.

The project is not a private server in the usual sense. It is a single-player conversion of the old client, wrapped in a launcher that starts a local database, login server, and world server before launching a patched copy of the game. You bring your own 1.12.1 client, build 5875, English, and the launcher builds what it needs from those files. That makes it closer to a total conversion mod with its own server stack than to a realm you log into with friends.

Image: The Hordes of Azeroth banner. Image credit: jonhaldors/hordes-of-azeroth, via the repository README.

What Hordes of Azeroth is

The README describes Hordes of Azeroth as a survivors-style game mode for World of Warcraft 1.12.1. An endless horde, spells that fire on their own, and a choice at every level up are the core loop. You play alone, for now, on a server that runs on your own PC.

The feature list covers Normal, Heroic, and Mythic zones, with more zones planned. Mythic has an Endless mode that lasts as long as you do. Raids have a Challenge mode, and Invasions bring raid enemies into the open world. Bosses telegraph their attacks so you can dodge them, and the Dash ability makes you untouchable for a moment, with its range and cooldown upgradeable at the outpost.

Loot includes gear, Darkmoon Cards, and Glyphs, and you extract one item at the end of a stage. The outpost holds account-wide gold, upgrades, fortunes, and class talents. Six classes are playable now, each with its own spells and evolutions, while Druid, Paladin, and Warlock are listed as coming later. Controller support covers Xbox and PlayStation pads, with prompts switching between controller and keyboard depending on what you used last.

Feature	What the README says
Genre	Survivors-style game mode on World of Warcraft 1.12.1
Players	Single player; each player runs their own local server
Zones	Normal, Heroic, and Mythic, with more planned
Modes	Endless on Mythic; Challenge mode in raids; Invasions in the open world
Classes	Six playable now; Druid, Paladin, and Warlock planned
Controls	Keyboard and mouse, plus Xbox and PlayStation controllers
Platform	Windows 10 or 11, 64-bit
Licence	GPL-3.0 for the project; the server is a VMaNGOS fork under GPL-2.0

What you need to install it

The requirements are specific. You need your own copy of World of Warcraft 1.12.1, build 5875, English (enUS). Hordes of Azeroth does not include the game. You also need a copy of that client used only for this project, because the launcher adds its own files to the folder and warns against mixing it with a client you use for other servers.

The machine needs Windows 10 or 11, 64-bit, about 3 GB of free space next to the game, and up-to-date graphics drivers. The launcher itself is a single executable, HoALauncher.exe, which you put next to WoW.exe. You press Install, wait for the first install to download about 460 MB, and then press Play.

After that, always start the game through the launcher. It starts the server, then the game, and stops the server when you close the game. While you play, it sits in the system tray and returns when the game exits. Updates arrive through the launcher: Play becomes Update, the patch notes appear, and your characters are backed up before the patch. You do not download the launcher again for each version.

How a run works

The loop is the survivors formula applied to Azeroth. Enemies come at you in a horde, your spells fire automatically, and each level gives you a choice that shapes the build. The difference is the setting and the vocabulary: zones, raids, dungeons, gear, Darkmoon Cards, Glyphs, and an outpost that carries progress across runs.

The outpost is the meta-progression layer. Gold is account-wide, and upgrades, fortunes, and class talents persist. At the end of a stage you extract one item, which gives each run a decision about what to keep. Boss attacks are designed to be read and dodged rather than absorbed, and Dash is the emergency button when a pattern catches you.

The modes change the shape of a run. Normal, Heroic, and Mythic set the difficulty of a zone. Endless mode on Mythic removes the finish line. Challenge mode lives inside raids. Invasions pull raid enemies into the open world. That is a lot of structure for a first public release, and the changelog is clear that balance has not had much work yet: some classes, spells, and builds will be too strong or too weak, and some zones are harder than they should be.

The monster path problem

The most interesting technical detail is the navigation mesh. In World of Warcraft, the server decides where monsters walk, not the client. Your client only has the shape of the world. The launcher turns that shape into walking maps for every zone the game uses, so mobs chase you around trees, up ramps, and through doorways instead of walking through walls.

The same maps keep spawns honest. Every monster spawns on a spot it can actually reach you from, never inside a rock or a tree, on a rooftop, or across a river or cliff. That is the kind of problem a survivors game makes worse, because the horde is constantly pathing toward a moving player.

The first release built those maps on the player's processor, which meant a long wait at 100 percent CPU. v0.1.1 changes that: the paths now come as a download, and installs and updates fetch them in a few minutes at about 330 MB. If the download fails, the launcher falls back to building them locally and asks whether to use all processor cores or half, so the PC stays usable. Local builds now run at a lower priority.

Release	Date	Main change
v0.1.0	October 20, 2026	First public version; six classes; navmeshes built from the player's game files
v0.1.1	October 21, 2026	Navmeshes download as about 330 MB; local builds fall back and run at lower priority

Local server, local saves

Hordes of Azeroth runs a small game server on your PC: a database, a login server, and a world server. The game connects to it, and the README says the connection never leaves your PC. The server only accepts connections from your own computer, so nobody on your network or the internet can reach it, and the launcher says you can answer the Windows Firewall prompt either way without changing how the game works.

The launcher goes online for two things: checking GitHub for updates and downloading them. Once the game is installed, you can play without internet, though the launcher cannot check for updates until you reconnect. Characters live in a local database inside the game folder. Before every update, the launcher saves a backup in HordesOfAzeroth\backups and keeps the three newest.

This is a cleaner privacy story than a hosted private server, because there is no remote account and no shared realm database. It also means your progress is only as safe as your disk. The launcher's backup policy helps with bad patches, not with a dead drive.

Early balance and missing pieces

The project is less than four days old. The repository was created on October 17, 2026, the first public release landed on October 20, and v0.1.1 followed on October 21. It had 5 stars, 1 fork, and 1 open issue at the time of check. Those are launch numbers, not a track record.

The README names the missing pieces, and the changelog warns that balance is early and bugs should be expected. Druid, Paladin, and Warlock are not playable yet. There is no multiplayer: each player runs a separate server, and the launcher only goes online for updates. Mac and Linux support is planned but not available. The launcher is Windows-only for now.

The launcher is also unsigned. Windows may show a SmartScreen warning on first run, and the README tells you to use More info and Run anyway. Every release lists the launcher's SHA-256 checksum, and the source is in the repository if you want to check or build it yourself. New unsigned programs that download files and start other programs are a common false-positive target for antivirus tools, so the checksum matters more than the warning alone.

The asset and licence question

Hordes of Azeroth is an unofficial fan project. The README says Blizzard Entertainment did not make it, does not endorse it, and has no part in it. It also says nothing in the repository or the releases comes from the game: no models, textures, sounds, maps, interface files, or game data. You bring a 1.12.1 client you own, and the launcher builds what it needs from it on your PC, including the patched game copy, the server's map data, and the spell effects patch.

The one added asset is the navigation mesh, which the README describes as custom data that says only where monsters can walk, with no models, textures, or terrain in it. The login screen uses the client's own interface files with the project's files loaded on top, and everything under Interface in the repository is written for the project.

The code is GPL-3.0. The server is a fork of VMaNGOS, published as hordes-of-azeroth-server under GPL-2.0. Other projects and licences are listed in a third-party notices file. That separation is the project's legal argument, not a ruling from Blizzard. Anyone installing it should understand that it requires a client they own and that fan projects can be asked to stop.

FAQ

Do I need World of Warcraft installed to play Hordes of Azeroth?
Yes. The README requires your own copy of World of Warcraft 1.12.1, build 5875, English (enUS), in a separate folder used only for this project. The download does not include the game.

Can I play Hordes of Azeroth with friends?
Not yet. Each player runs their own server on their own PC. The launcher only goes online to check for updates and download them, and the README says multiplayer is not available.

Does Hordes of Azeroth change my existing WoW folder?
It makes a patched copy called HordesOfAzeroth.exe and adds its own server, addon, spell effects patch, and graphics fix. The README tells you to use a separate 1.12.1 client so those files do not mix with another server's setup.

Does Hordes of Azeroth run on Mac or Linux?
No. It is Windows 10 or 11, 64-bit only for now. The README says Mac and Linux support is planned.

Why does the launcher need an internet connection?
It uses the connection to check GitHub for updates and download them. After installation you can play offline, but the launcher cannot check for updates until you are back online.

Why does the first install take so long?
The launcher needs navigation meshes for every zone so monsters can path around the world. v0.1.1 downloads them ready-made at about 330 MB as part of an install that downloads about 460 MB, with a local build as the fallback.

Our take: Hordes of Azeroth is the most technically ambitious project in this batch, and it is not even a drop miner. Turning a 2006-era client into a single-player survivors game means solving pathfinding, server orchestration, patching, and persistence before the first enemy walks toward you. The local-only server and the asset separation are the right calls for a fan project. The caution is the age: under four days, two releases, and an explicit warning that balance is rough. This is worth watching if you keep a 1.12.1 client around. It is not yet worth planning a weekend around.

Bottom Line

Hordes of Azeroth is a GPL-3.0 survivors-style game mode for World of Warcraft 1.12.1, played alone on a local server that the launcher starts and stops with the game. It adds Normal, Heroic, and Mythic zones, Endless and Challenge modes, Invasions, six classes, an outpost with account-wide progression, and controller support. v0.1.1 arrived on October 21, 2026, with navigation meshes delivered as an about 330 MB download instead of a long local build. It requires your own 1.12.1 build 5875 client, runs only on Windows for now, has no multiplayer, and is still early enough that balance and bugs are part of the package.

Sources

jonhaldors/hordes-of-azeroth repository README, including the requirements, install steps, feature list, questions, local server and firewall explanation, asset statement, and licence notes, verified October 21, 2026.

jonhaldors/hordes-of-azeroth releases, including v0.1.0 published October 20, 2026 and v0.1.1 published October 21, 2026, with the navigation mesh download change and launcher checksum.

Hordes of Azeroth CHANGELOG, confirming the six playable classes, the planned Druid, Paladin, and Warlock, the early balance warning, and the v0.1.1 navigation mesh change.

GitHub repository metadata for jonhaldors/hordes-of-azeroth, verified October 21, 2026: 5 stars, 1 fork, 1 open issue, Lua, GPL-3.0, created October 17, 2026, last push October 21, 2026.

LiberNT: A ReactOS Fork Built Around Windows 11

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:23:21 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: LiberNT is a ReactOS-derived operating system that targets Windows 11 compatibility instead of the older NT5-era behavior that has long defined ReactOS. Its latest debug build, published October 14, runs on amd64, arm64 and riscv64, with experimental Windows 11 driver-model work, NTFS updates and a WoW64 path that uses FEX on Arm.

The LiberNT desktop on arm64 in QEMU, build nt10.0-043-2555. Credit: LiberNT/GitHub.

What LiberNT is

LiberNT is a free operating system built on the Windows NT architecture. Its README calls it a fork of ReactOS focused on modern hardware support, modern platform support and modern Windows NT compatibility work that can later be submitted upstream. It is not a Linux distribution and does not use the Linux kernel.

The project targets Windows 11 compatibility with room for experimental features. It focuses on 64-bit platforms, with prebuilt images for amd64, arm64 and riscv64. The repository is licensed under GPL-3.0, was created on May 4, 2026, and showed 456 stars and 30 forks when checked on October 21, 2026.

The legal footing is explicit. Per the README, LiberNT follows a clean-room approach: it studies Windows behavior through Microsoft's public documentation, SDK and WDK headers, public symbols and black-box testing. The project says it does not copy or transcribe proprietary Windows code, and it does not copy GPL code from Linux.

LiberNT at a glance	Status per project
Base	Fork of ReactOS, GPL-3.0
Compatibility reference	Windows 11 (NT10); Linuxiac reports Windows 11 24H2
Prebuilt architectures	amd64, arm64, riscv64; PowerPC release images
Boot	UEFI and BIOS; GOP framebuffer and early display
Storage and buses	NTFS, NVMe, StorPort, SD/eMMC, PCIe, USB xHCI
Graphics	WDDM with Mesa; Direct3D 10, 11 and 12 work listed in release notes
x86 on Arm	WoW64 with FEX, a build option

Windows 11, not NT5

ReactOS has historically targeted the Windows NT5 era, the architecture behind Windows XP and Server 2003. LiberNT's main branch is named main-nt10, and the project's release notes say the Windows 11 platform is now the default target. The October 5 build added the Windows 11 kernel export set on 64-bit targets, a native Direct3D 12 runtime and Direct3D 11.2 through 11.4 support.

That is a deeper change than a theme or a start menu. Windows 11 compatibility touches the kernel exports, the driver model, the graphics stack and the user-mode contracts that applications depend on. The October 14 release notes list a new console stack built on condrv.sys and conhost, Direct3D 10 on the native runtime, a Direct3D 11.1 display driver interface, and NTFS support for replaying the Windows 11 LFS 2.0 log so volumes mount read-write.

Linuxiac reported that LiberNT uses Windows 11 version 24H2 as its compatibility reference. The README does not name the exact build, so treat 24H2 as Linuxiac's reporting rather than a specification published by the project. The broader claim, that Windows 11 is the reference instead of NT5, is stated in the README.

The platform list

LiberNT's platform scope is wider than a typical ReactOS fork. The project ships prebuilt images for amd64, arm64 and riscv64, and includes PowerPC images in its release builds. The ARM64 work includes Raspberry Pi support and the RP1 I/O controller, which is the kind of board bring-up that matters for embedded and single-board use.

The boot path supports UEFI and BIOS, with a GOP framebuffer and early display support. Storage and bus work covers NVMe, StorPort, SD/eMMC, PCIe, Plug and Play and USB xHCI. Graphics use the Windows Display Driver Model with Mesa, and the ARM64 build can run x86 applications through WoW64 with FEX when that option is enabled.

LiberNT Setup is designed to behave like Windows Setup. Per the README, it lists drives, creates, deletes, formats, shrinks and extends partitions, and installs the system on NTFS with the EFI system partition it needs. Release images come as a preinstalled disk image, a VHD for virtual machines and a bootable installation ISO.

What works and what does not

The release notes show steady progress across the stack. The October 8 build added NVMe controller installation, NTFS compression and short-name tunneling, Windows 11-style driver installation and a writable LiveCD RAM drive. The October 14 build added NTFS replay for Windows 11 volumes, Direct3D 10 on the native runtime, xHCI fixes and more WoW64 handle handling.

Application compatibility is much narrower than those platform notes suggest. Linuxiac reported that the project's compatibility database had only five entries, all tested against a specific development build. Google Chrome, the open-source strategy game 0 A.D. and Command & Conquer: Yuri's Revenge were reported to work. Steam was listed as usable. The RPCS3 PlayStation 3 emulator started but was not classified as usable.

Linuxiac also reported that LiberNT replaces several ReactOS core components, including a new NVS memory manager and kernel pool implementation, a new cache manager, a new NTFS driver and an updated StorPort implementation with NVMe support. Those are Linuxiac's characterizations of the project; the README and release notes confirm the NTFS, NVMe and graphics work but do not use the same component names.

Experimental status

The README is blunt about readiness. The builds on the releases page are debug pre-releases for testing, and users are told to run them in a virtual machine or on spare hardware rather than a computer they depend on. That warning applies to the current October 14 build, not just an early alpha.

LiberNT also discloses its use of AI. Per the README, AI helps review and write code, and everything it writes is reviewed by a human. Before anything reaches the main branch, a person runs and validates the end-to-end tests. Contributors are asked to say in the pull request which parts AI helped with and how they tested them.

The project is active. The repository was pushed on October 21, 2026, and the release history shows multiple builds in October alone. That pace is encouraging for a young fork, but it also means the compatibility picture can change from build to build. A feature listed in one release note is not a promise that every application will work.

Our take: ReactOS has spent decades chasing Windows XP and Server 2003 compatibility. LiberNT is interesting because it aims at the modern NT10 contract, where 64-bit kernel exports, the driver model and the graphics stack matter more than running 20-year-old applications. The risk is that Windows 11 compatibility is a much larger surface, and five compatibility entries is not evidence that it has been covered. The project's own warning is the right frame: test it in a VM, not on your main PC.

FAQ

What is LiberNT?
LiberNT is a free operating system built on the Windows NT architecture and based on ReactOS. It targets Windows 11 compatibility, with prebuilt images for amd64, arm64 and riscv64.

Is LiberNT a Linux distribution?
No. It does not use the Linux kernel. It implements the Windows NT architecture and is a fork of ReactOS, which is itself an open-source Windows-compatible operating system.

Which architectures does it support?
The project ships prebuilt images for amd64, arm64 and riscv64, and includes PowerPC images in its release builds. ARM64 includes Raspberry Pi support.

Can it run Windows applications?
Some. Linuxiac reported that Chrome, 0 A.D. and Command & Conquer: Yuri's Revenge worked in the project's compatibility database, while Steam was usable and RPCS3 started but was not classified as usable. The database had only five entries.

Is it ready for daily use?
No. The current builds are debug pre-releases, and the README says to run them in a virtual machine or on spare hardware rather than a computer you depend on.

Does LiberNT copy Windows code?
No, per the README. The project says it uses public documentation, SDK and WDK headers, public symbols and black-box testing, and that it does not copy or transcribe proprietary Windows code or GPL code from Linux.

Bottom Line

LiberNT is a GPL-3.0 ReactOS fork targeting Windows 11 (NT10), with prebuilt amd64, arm64 and riscv64 images, UEFI and BIOS boot, NTFS and NVMe support, WDDM graphics and a WoW64 path that uses FEX on Arm. Its compatibility list is tiny and its builds are debug pre-releases, so it is a project to test in a VM, not a Windows replacement. The interesting part is the target: a clean-room NT implementation aiming at Windows 11 rather than the NT5 era.

Sources: LiberNT repository and README; LiberNT nt10.0-043-2555 release notes, October 14, 2026; Linuxiac, October 21, 2026; GitHub API metadata fetched October 21, 2026. All sources checked October 21, 2026.

12V-2x6: Seven Melted RTX GPUs in Two Days

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:23:19 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: A Chinese repair channel says it received seven Nvidia GPUs in two days with burned or melted 12+4-pin power connections, mostly RTX 5090s. The report lands as Super Flower lists a $38 inline OCP dongle that monitors current across the 12V lines and triggers a black screen when the load stays too high.

Super Flower VGA 12+4Pin OCP Dongle. Credit: Super Flower.

The repair report

PC Gamer reported on October 21 that the channel 修电脑的张哥, roughly Zhang the Computer Repairman, posted a video after receiving seven cards in two days with damage at the 12+4-pin power interface. The cards were mostly RTX 5090s, with Asus and MSI RTX 4090s, a Gainward card and a Founders Edition RTX 5090 also in the mix.

The damage was not cosmetic. Some cards had a burned socket and a blackened PCB. On others the connector had melted into the socket, forcing the owner to cut the cable before sending the card in. Zhang removed the damaged interface, cleaned the contacts, soldered on a new 12+4-pin socket and tested each card with 3DMark. Some cables at the power supply still ran warm, but he did not flag them as dangerous after the repair.

One repair shop is not a failure rate. The channel is popular on Bilibili and YouTube, so it may attract more damaged cards than a typical shop. The users were not all running weak power supplies, and some cards came from lab computers that may have been worked harder than a home gaming rig. Even with those caveats, seven cards from several manufacturers in two days is a pattern worth watching.

Reported failures at one Chinese repair channel	Detail
Cards received	Seven in two days
Models	Mostly RTX 5090; Asus and MSI RTX 4090; Gainward; Founders Edition RTX 5090
Failure point	12+4-pin cable and GPU socket interface
Damage	Burned socket, blackened PCB, or connector melted into the socket
Repair	Remove the old interface, clean contacts, solder a new socket, test the card

Source: PC Gamer, October 21, 2026, citing the repair channel's video.

Why the connector fails

The 12V-2x6 connector, often called 12+4-pin, is the power interface on high-end Nvidia cards. It carries up to 600W through a small set of contacts. When one pin makes poor contact, the current that should spread across the bundle concentrates in fewer paths. Resistance rises, heat rises, and the plastic housing or the PCB around the socket can burn before the card reports a fault.

PC Gamer's report is careful about the cause. The failures are not purely down to cheap adapters, old power supplies or cables that were not fully seated. The amount of power running through these cards is enough to damage the connector when the contact is marginal. That is why the same failure shows up across different board partners and different card models.

The connector can also fail slowly. A card may boot and run games while the socket is already discolored or the cable is running hot. By the time the user notices a crash or a smell, the damage may have moved from the cable to the card. The repair is possible, but it is a board-level job, not a cable swap.

Super Flower's OCP dongle

Super Flower has listed the VGA 12+4PIN OCP Dongle, a device it first showed at Computex 2026. It sits between the power supply cable and the GPU and monitors current across the individual 12V lines. An ARGB light and a buzzer provide the warning. The company published the thresholds on its product page.

Current condition	Alert	What happens
9.5A to 12A, or more than 6A difference between any two 12V lines, held for 10 seconds	Orange LED and rapid short beeps	Warning clears if current returns to normal
More than 12A, held for 10 seconds	Red flashing LED and continuous alarm for 180 seconds, then solid red for 15 seconds	VGA Black Screen Protection triggers; shut down and inspect the GPU power connections
Price and availability	NT$1,190, about $38	Currently bundled with LEADEX VIII PLATINUM PRO PREMIUM power supplies; not sold separately yet

Sources: Super Flower product page and VideoCardz, October 21, 2026.

VideoCardz reported that the dongle is listed at NT$1,190 in Taiwan, or about $38. It is currently offered as a promotional gift with Super Flower's LEADEX VIII PLATINUM PRO PREMIUM power supplies in 850W, 1000W and 1200W versions. The company has not said when it will be sold separately or whether it will reach markets outside Taiwan.

The category is growing. VideoCardz notes that Cooler Master's $50 GPU Shield Adapter and Thermal Grizzly's WireView Pro II also monitor the GPU power connection. The Super Flower unit is notable for its low price and for the black-screen trigger, which is more aggressive than a light or a log entry.

What it can and cannot do

The dongle can catch a sustained overload or a large current imbalance and force the display to cut out before the connector cooks. It cannot fix a loose connection, a damaged cable or a socket that is already worn. It also cannot tell you why the current is high. It is a warning and shutdown trigger, not a repair for a bad contact.

The 180 second alarm and 15 second solid red period are deliberate. They give the user time to see and hear the warning before the black screen. The protection state remains active even if the current returns to normal, so the system does not silently resume after a fault. Super Flower's instructions are to shut down and inspect the connections before restarting.

The price is easier to justify than a new GPU, but the bundle-only launch limits who can buy one. If the dongle becomes a standalone product at $38, it will be one of the cheaper ways to add per-line monitoring to an existing card. Until then, most buyers will have to rely on the card's own sensors and a visual check.

What owners should check

Zhang's field advice, as reported by PC Gamer, is to touch the cable about an inch above the GPU connector during use. Warm is normal. Notably too hot to touch is a warning sign. He also suggests watching the 12+4-pin voltage readout in GPU-Z. If the voltage drops well below 12V, to 11.3 or 11.4V, there is likely a problem and the connector should be checked.

The touch test is a rough field check, not a measurement. A better routine is to power down, reseat the cable at both ends, listen for a positive click, avoid tight bends near the connector and inspect the housing for discoloration. If the cable or socket is already damaged, stop using the card and have it repaired.

A lower power limit can reduce the current through the connector, but it does not fix a bad contact. The same goes for a new cable: it may restore a solid connection, but it will not help if the card's socket is worn. The failure is a system problem, not a single accessory problem.

Our take: the most useful part of the repair report is the mix of vendors. Seven cards from Asus, MSI, Gainward and Nvidia's own Founders Edition suggests this is not one partner's cable or one bad batch. The least useful part is treating seven cards as a rate. We do not have a denominator, and a popular repair channel is not a random sample. The right response is to check the connection, not to panic about every RTX 5090.

FAQ

What happened at the repair shop?
A Chinese repair channel reported receiving seven Nvidia GPUs in two days with burned or melted 12+4-pin power connections. Most were RTX 5090s, with some RTX 4090s and cards from several manufacturers.

Is this a widespread RTX 5090 failure?
The report does not establish a failure rate. It is one repair channel's two-day intake, and the channel's popularity may bring in more damaged cards than a typical shop. The mix of vendors is notable, but the sample is small.

What is the 12V-2x6 connector?
It is the 12+4-pin power connector used on high-end Nvidia cards, rated for up to 600W. The 12V-2x6 name refers to the 12V rails and two rows of six power contacts. The extra four sense pins are why the same connector is often called 12+4-pin.

What does Super Flower's dongle do?
It monitors current across the individual 12V lines and warns at 9.5A to 12A or a 6A imbalance. Above 12A for 10 seconds, it alarms and eventually triggers a black screen so the user shuts down and inspects the connection.

Can I buy the dongle separately?
Not yet. VideoCardz reports that it is currently bundled with Super Flower's LEADEX VIII PLATINUM PRO PREMIUM power supplies at NT$1,190, about $38, with no standalone sale date or international availability announced.

What should I check on my own GPU?
Check that the cable is fully seated at both ends, inspect for discoloration and feel the cable about an inch above the GPU connector during use. If it is too hot to touch or the 12+4-pin voltage drops to 11.3 or 11.4V, shut down and inspect the connection.

Bottom Line

Seven melted or burned connectors at one repair shop in two days is a warning, not a verdict on every RTX 5090. The 12V-2x6 connector runs close to its limit on a 600W card, and a marginal contact can heat up long before the card reports a fault. Super Flower's $38 OCP dongle adds per-line current monitoring and a black-screen trigger, but it is bundled with a power supply for now and cannot repair a bad connection. Check the seating, the heat and the voltage before the damage reaches the board.

Sources: PC Gamer, October 21, 2026; Super Flower VGA 12+4PIN OCP Dongle product page; VideoCardz, October 21, 2026. All sources fetched October 21, 2026.

RTX PRO 5500: China's 3 Million GPU Order Nvidia Cannot Fill

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:23:17 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: ByteDance, Alibaba and Tencent have reportedly ordered more than 3 million of Nvidia's RTX PRO 5500 workstation GPUs, a package worth over $35 billion, according to Reuters. Nvidia has told customers it cannot fill the orders, and sources told Reuters it may deliver only 5% to 10% of the demand.

NVIDIA RTX PRO 5500 Blackwell workstation GPU. Credit: NVIDIA.

The order

Reuters reported on October 20 that ByteDance, Alibaba and Tencent had placed orders for more than 3 million RTX PRO 5500 cards, with a combined value above $35 billion. Two sources put the per-card price at around $11,400. Neither Nvidia nor the three Chinese companies confirmed the order on the record.

The size is the story. Three companies have ordered more cards than Nvidia is reportedly able to produce for the entire world in 2027. Tom's Hardware reported that Nvidia is in a position to make only about 3 million RTX PRO 5500 cards globally next year, which means the Chinese order is not a normal purchase request. It is a claim on a constrained allocation.

An Nvidia spokesman told Reuters that demand from China has seen "unprecedented growth since 2022," when export restrictions began to reshape the market. The spokesman also said U.S. firms remain restricted by a combination of outdated U.S. export controls and China's own limits on U.S. imports. That is the backdrop for a card that is legal to sell in China but still scarce.

RTX PRO 5500 order book	Figure	Source
Cards ordered by ByteDance, Alibaba and Tencent	More than 3 million	Reuters
Combined order value	More than $35 billion	Reuters
Implied average price	About $11,400 per card	Reuters via Tom's Hardware
Share Nvidia may deliver	5% to 10% of demand	Reuters sources
Nvidia's 2027 global production	About 3 million cards	Tom's Hardware
Memory	84GB GDDR7 with ECC	NVIDIA
Memory bandwidth	1,398 GB/s	NVIDIA
Maximum power	Up to 600W	NVIDIA

Why this GPU

The RTX PRO 5500 is a Blackwell workstation card with 84GB of GDDR7 memory, 1,398 GB/s of bandwidth and a 600W power ceiling, according to NVIDIA's product page. It has 21,760 CUDA cores, the same count as the RTX 5090, per Wccftech, but 84GB of memory in a single card. That combination is what makes it useful for AI inference: more model weights and more working memory fit on one card.

It is not an H200 or a B300. Those accelerators remain banned or heavily restricted for China, which is why buyers are routing demand to a workstation GPU that is not currently restricted. Tom's Hardware reported that the RTX PRO 5500 is fully compliant with the U.S. export rules that still allow it to be sold in China. The card is slower than the top AI accelerators, but a legal card that ships is worth more than a faster card that does not.

The software side matters as much as the memory. Nvidia's CUDA stack is the default for many AI frameworks, and Chinese teams have years of code written against it. A domestic accelerator can look competitive on a spec sheet and still lose on porting time, compiler maturity and library support. That is the gap the RTX PRO 5500 exploits.

The 5% to 10% gap

Nvidia has told partners it may be able to meet only 5% to 10% of the demand, according to sources who spoke to Reuters. At the low end, a 3 million card order becomes 150,000 delivered cards. At the high end, it becomes 300,000. Either way, the overwhelming majority of the order is a queue position rather than a shipment.

The numbers do not describe a simple wafer shortage. Nvidia's global production is already allocated across customers and regions, and the RTX PRO 5500 is not the only product using its Blackwell silicon and GDDR7 memory. The Chinese order is competing with every other order in the same queue. That is why a 3 million card production plan can coexist with a 5% to 10% fill rate for one group of buyers.

For the three Chinese companies, the calculation is still rational. If inference demand grows faster than they can secure domestic alternatives, paying $11,400 per card for a fraction of the order keeps their AI services running. The order also reserves a place in line for later allocations. A purchase order that ships slowly is still a supply agreement.

Why Huawei is not the default

Huawei's Ascend accelerators are the obvious domestic alternative, and Wccftech framed the Reuters order as a rejection of those chips in favor of Nvidia. Wccftech reported that the Ascend 950PR looks competitive on paper, with 128GB of memory and 1.4 to 1.6 TB/s of bandwidth, but argued that Nvidia's CUDA support and raw performance keep Chinese buyers on its platform.

That is a claim from Wccftech's report, not a benchmark we ran. The broader point is easier to defend: export controls have not removed demand for Nvidia hardware in China. They have pushed it toward the most capable Nvidia product that can still be sold legally. The RTX PRO 5500 sits in that narrow band, and buyers are treating it as a strategic input rather than a workstation upgrade.

Huawei is not out of the picture. Domestic AI accelerators are receiving heavy investment, and the supply gap gives them a captive market. But the reported order shows that software compatibility and delivery certainty still carry a premium, even when the supplier cannot promise full fulfillment.

What it means

For Nvidia, the order is more than $35 billion of demand it cannot fully monetize in 2027. For ByteDance, Alibaba and Tencent, it is a bet that inference capacity is worth paying for before the supply gets tighter. For the rest of the market, it is another signal that high-memory GPUs are being pulled into AI deployments before they reach other buyers.

The export controls are also doing something more subtle than blocking sales. They are shaping which products get designed, which customers get priority and which legal workarounds become mass-market parts. The RTX PRO 5500 is a workstation card that happens to fit the rules. At 3 million units of demand, it has become an AI product by another name.

None of this means the order will arrive intact. Nvidia has not confirmed the figures, the Chinese companies have not confirmed the figures, and the 5% to 10% fill rate is sourced to people familiar with the discussions. Treat the order as a demand signal, not as delivered capacity.

Our take: the interesting part is not that China wants Nvidia. It is that export rules have created a gray-zone product with enough memory to be useful and enough legality to be bought at scale. The 5% to 10% fill rate makes the order partly a queue position, not a purchase that will arrive on schedule. That is a strange outcome for a policy designed to limit access.

FAQ

How many RTX PRO 5500 cards did ByteDance, Alibaba and Tencent order?
More than 3 million cards, according to Reuters. The reported order is worth more than $35 billion at around $11,400 per card.

How much of the order can Nvidia fill?
Nvidia has told partners it may meet only 5% to 10% of demand, according to sources who spoke to Reuters. That works out to roughly 150,000 to 300,000 cards from a 3 million card order.

Why not buy Huawei's Ascend chips instead?
Huawei's accelerators are the domestic alternative, but Nvidia's CUDA software stack and performance remain a major draw, according to Wccftech's report. The order does not prove Huawei is out of the market.

Is the RTX PRO 5500 legal to sell in China?
It is not currently restricted by the U.S. export rules, per Tom's Hardware. Export policy can change, and the card's status is one reason buyers are moving quickly.

What are the RTX PRO 5500's key specs?
NVIDIA lists 84GB of GDDR7 memory with ECC, 1,398 GB/s of bandwidth and up to 600W of power. Wccftech reports 21,760 CUDA cores, the same count as the RTX 5090.

Have Nvidia or the Chinese companies confirmed the order?
No. The figures come from Reuters sources, and neither Nvidia nor ByteDance, Alibaba or Tencent has confirmed the order on the record.

Bottom Line

ByteDance, Alibaba and Tencent have reportedly ordered more than 3 million RTX PRO 5500 cards worth over $35 billion, while Nvidia may be able to deliver only 5% to 10% of the demand. The card is not an H200 or B300, but it has 84GB of GDDR7, runs CUDA and is not currently restricted for China. The order is a demand signal and a queue position, not a confirmed shipment. It shows that export controls have redirected Chinese AI demand rather than eliminated it.

Sources: Reuters report, October 20, 2026, as summarized by Tom's Hardware; NVIDIA RTX PRO 5500 product page; Wccftech, October 21, 2026. All sources fetched October 21, 2026.

Samsung 2027 Memory: 80% Already Booked Before the Year Starts

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 08:23:15 -0400

By Indie Kings | October 21, 2026

Updated October 21, 2026: Samsung has reportedly booked about 80% of its planned 2027 memory output through long-term agreements, according to KB Securities, leaving hyperscalers to compete for the last 20%. The estimate lands as Micron faces a one-day strike at its Taoyuan plant in Taiwan on October 22 and Dell's CEO warns the shortage could last four years.

Samsung Electronics' semiconductor booth at SedEx2026 in Seoul. Credit: Newsis via The Korea Herald.

The 80% number

KB Securities estimates that Samsung has already secured roughly 80% of its planned 2027 memory production through long-term agreements. Kim Dong-won, head of research at the brokerage, said the figure covers five-year contracts for 70% of Samsung's planned conventional DRAM and NAND output, plus supply commitments for all of its projected high-bandwidth memory production.

"Samsung is estimated to have already secured approximately 80 percent of its planned memory production for 2027 through long-term supply agreements," Kim said in a report released Tuesday, per The Korea Herald. The remaining 20% is not sitting on a shelf. Kim said competition among hyperscalers for that uncontracted output is still going.

The brokerage also expects Samsung's 2027 operating profit to reach 555 trillion won, or about $417.6 billion, up 51% from an estimated 367 trillion won this year. That is a forecast from KB Securities, not company guidance. The useful part for buyers is the contract coverage, because it shows how little 2027 memory is still available at spot prices.

Samsung's 2027 memory segment	Covered by long-term agreements	Contract detail
Total planned memory output	About 80%	Remaining 20% still contested by hyperscalers
Conventional DRAM and NAND	70%	Five-year contracts with global technology companies
High-bandwidth memory	100%	Entire projected HBM output under supply commitments

Source: KB Securities estimate reported by The Korea Herald, October 20, 2026.

HBM takes the rest

HBM is the part of the memory market that AI accelerators have pulled away from the PC supply chain. KB Securities expects Samsung's HBM average selling prices to rise 100% to 150% year over year in 2027, with HBM4 reaching 80% of the company's HBM revenue, per The Korea Herald.

The industry pattern is the same at Samsung's rivals. Micron CEO Sanjay Mehrotra said on the company's September 30 earnings call that more than 75% of Micron's planned 2027 output had already been committed, according to The Korea Herald. SK hynix said in July that it had finalized long-term agreements with around 10 major customers.

That matters to gaming PCs even though HBM does not go into a GeForce card. HBM and conventional DRAM compete for wafer starts, packaging capacity and the same limited pool of memory test equipment. When data center buyers lock up HBM and a large share of DRAM, the consumer market gets what is left, and it pays the spot price for it.

PC Gamer's report draws a consumer consequence from the same note: a double-digit quarter-over-quarter price increase for consumer memory after the new year. That is PC Gamer's forecast, not a shelf tag. It does mean the start of 2027 is the window to watch, because that is when the 2027 contracts and the consumer allocation start to show up in retail pricing.

Micron's one-day strike

Micron's Taoyuan union will hold a one-day strike in Taiwan on October 22, Reuters reported via Tom's Hardware. Workers are asking for concrete terms for this year's bonus and a transparent formula for sharing future profits. Talks between the company and the union ended without an agreement in September, and employees approved industrial action in early October.

The timing is awkward because fabs are built to run around the clock. Tom's Hardware reported that the strike will have a minor impact on Micron's memory output, because a one-day warning strike does not necessarily stop all manufacturing if management prepares for it. The Taichung union, which represents Micron's largest Taiwan site, is scheduled to meet the company for mediation on the same day, but it has not authorized a strike.

Micron employs about 15,000 people in Taiwan, and its plants there make commodity DDR5 and LPDDR as well as HBM. A single day at one site is small next to Samsung's 2027 contract book. It is still a labor risk sitting on top of a supply chain that has already sold most of next year forward.

Dell's four-year warning

Michael Dell told CNBC's David Faber that the mismatch between demand and supply gets worse next year. "I do think it gets worse next year in terms of demand and supply," Dell said, according to Benzinga's report on the interview published by Yahoo Finance on October 16. When Faber asked specifically about memory, Dell added: "Maybe it's a four-year problem because there's just not enough supply."

Dell's reasoning is about construction time. New fabs take about four years to build, so plants planned today may not begin production until 2029 or 2030. Dell is also carrying a record $95 billion backlog of AI server orders, which gives the company a direct financial reason to care about every DRAM and HBM allocation it can secure.

That does not mean every memory product stays scarce for four years. It means the capacity being added now is aimed at the AI buildout, and the consumer market should expect to compete for the marginal supply. The 80% contract figure and Dell's four-year comment are two versions of the same signal: buyers with multiyear agreements are locking in supply before it reaches the spot market.

What buyers should watch

Three dates matter more than the daily spot-price headlines. The first is October 22, when the Micron strike and the Taichung mediation happen. The second is Samsung's detailed third-quarter earnings on October 29, which will show how much of the HBM optimism is already in the numbers. The third is the start of 2027, when the new consumer memory pricing is expected to land.

For PC builders, the practical read is narrower than "buy everything now." DDR5 kit prices can move independently of HBM contract prices, and availability varies by capacity and speed bin. The contract data points to higher prices after the new year, not to an instant shortage of every kit on a retailer's shelf.

The bigger risk is a repeat of 2026: a headline about one factory or one strike gets treated as the whole story, while the actual constraint is multiyear allocation. Watch the contract coverage, the HBM4 ramp and the fab timelines. Those are the numbers that set the floor for consumer prices.

Our take: the memory market has moved from spot pricing to forward contracts, and that changes how these stories should be read. One repair shop or one labor action is noise next to 80% of 2027 output being booked before 2026 ends. The caveat is that KB Securities is making an estimate, not publishing Samsung's order book, and the uncontracted 20% could still be allocated in ways that surprise the spot market.

FAQ

How much of Samsung's 2027 memory output is already booked?
About 80%, according to a KB Securities estimate reported by The Korea Herald on October 20, 2026. The figure covers long-term supply agreements, not confirmed revenue.

Who is buying the remaining 20%?
Hyperscalers and other large buyers are still competing for the uncontracted output, according to KB Securities. The remaining supply is the part that will be priced closest to the spot market.

Will consumer RAM prices rise again?
PC Gamer projects a double-digit quarter-over-quarter increase for consumer memory after the new year, based on the KB Securities note. That is a forecast, and retail pricing can lag or diverge from contract pricing.

Why does HBM matter to gaming PCs?
HBM does not go into gaming GPUs, but it uses DRAM wafers and advanced packaging capacity that also serve conventional memory. When HBM is fully committed, it tightens the overall memory supply chain.

What is the Micron strike about?
Micron's Taoyuan union is holding a one-day strike on October 22 over 2026 bonus terms and a clearer profit-sharing formula. Reuters reported that the stoppage would have a minor impact on Micron's memory output.

Is the shortage really a four-year problem?
Michael Dell said it could be, because new fabs take about four years to build and may not produce until 2029 or 2030. That is his assessment, not a fixed timeline for every memory product.

Bottom Line

Samsung's 2027 memory book is about 80% full, HBM is fully committed and the last 20% is being fought over by hyperscalers. Micron's one-day Taoyuan strike is small next to that contract math, but it adds labor risk on October 22. Dell's four-year warning explains why the capacity being built now will not rescue 2027. For PC builders, the signal is higher consumer memory prices after the new year, with availability depending on the capacity and speed bin.

Sources: KB Securities estimate reported by The Korea Herald on October 20, 2026; PC Gamer on October 21, 2026; Tom's Hardware on October 21, 2026; Benzinga via Yahoo Finance on October 16, 2026. All sources fetched October 21, 2026.

Xbox Helix: $1,200 to $1,800 and Holiday 2028

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 07:47:14 -0400

By Indie Kings | October 20, 2026

Updated October 20, 2026: A Fortune profile of Xbox CEO Asha Sharma reports that Microsoft has internally projected the Project Helix console at $1,200 to $1,800 to consumers, unless the company decides to subsidize it. The report also says the next-gen Xbox handheld is targeted for the 2027 holiday season, with the console following in time for the 2028 holidays. Microsoft has not confirmed any pricing or launch date.

The one-sentence version: the next Xbox is reportedly being planned as a $1,200 to $1,800 console for holiday 2028, with a handheld arriving a year earlier, and neither number is official.

Image credit: Microsoft Xbox, via TechPowerUp. This is an Xbox brand image, not a Project Helix product photo.

What Fortune actually reported

Fortune published a profile of Asha Sharma, the Xbox CEO, on October 20, 2026. According to TechPowerUp's summary of the report, the profile says the Project Helix console has been internally projected to cost consumers between $1,200 and $1,800, unless Microsoft decides to subsidize it. The same report says the next-gen handheld is planned for the 2027 holiday season and the console for the 2028 holidays.

Those are reported internal projections, not announced prices. Microsoft has not confirmed the numbers, the launch windows or the final product mix. The report cites people familiar with the matter, which is the standard sourcing for pre-announcement hardware stories. Readers should treat the figures as a credible leak, not a price tag.

The report also says Helix engineering deadlines have been moved up by at least several months, putting extra pressure on development teams. That detail matters because it explains why the pricing and timing are being discussed together. A console that is being pulled forward is a console whose cost structure is still moving.

Item	Reported figure	Status
Helix console price	$1,200 to $1,800	Internal projection, per Fortune; not confirmed by Microsoft
Handheld launch	Holiday 2027	Reported target; not confirmed
Console launch	Holiday 2028	Reported target; not confirmed
Current Xbox Series X with disc drive	$799.99	Current retail price after recent increases
Xbox profit margin	About 3%	Fortune estimate, per TechPowerUp

Source: Fortune profile of Asha Sharma, October 20, 2026, via TechPowerUp.

Why $1,200 to $1,800 is plausible

The first reaction to a $1,200 console is sticker shock. The current Xbox Series X with a disc drive sells for $799.99 after recent price increases, and that is already a high price for a console. A $1,200 to $1,800 Helix would be a different category of product, closer to a gaming PC than to a traditional console.

There are three reasons the range is not fantasy. The first is memory. Micron's CEO has said memory supply will be much tighter in 2027 and 2028 than in 2026. A console launching in holiday 2028 will be buying memory in the middle of that squeeze, and memory is one of the largest cost components in a modern console.

The second is the chip. Helix is expected to use a custom AMD system-on-chip with a large GPU and a CPU complex that can run console games and PC software. Our earlier coverage of the Magnus SoC and the reported 56 TFLOPS figure put Helix in a different performance class from the Series X. A bigger chip, more memory and a more complex cooling system all raise the bill of materials.

The third is the handheld. Microsoft is reportedly planning a separate portable device for 2027, followed by the console in 2028. Two devices mean two development programs, two supply chains and two marketing launches. The handheld may share an architecture with the console, but it does not share the same chassis, display, battery or thermal design.

None of that makes $1,800 likely. It makes the range understandable. A $1,200 console is expensive; a $1,800 console is premium PC territory. The final number will depend on how much Microsoft is willing to lose on each unit and how much memory and silicon it can secure at a fixed price.

The handheld-first, console-second timeline

The reported schedule is handheld in holiday 2027, console in holiday 2028. That is a reversal of the usual console launch pattern. Microsoft would put a portable device in the market first, then follow with the living-room box a year later.

The strategy has logic. A handheld can launch with a lower price, a smaller chip and a library that includes existing Xbox games. It can also test the market for a portable Xbox before the company commits to a full console generation. If the handheld succeeds, the console arrives with an installed base and a proven software stack. If it struggles, Microsoft has learned something before the expensive launch.

The risk is that the handheld defines the generation before the console arrives. A 2027 handheld will be judged against the Steam Deck, the ROG Ally and whatever Sony or Nintendo ships in the same window. If it runs a Windows-based OS with a console interface, it will be compared to Windows handhelds. If it runs a custom Xbox OS, it will be compared to the Steam Deck's SteamOS. Either comparison is difficult.

The console then has to justify a year of waiting and a much higher price. A $1,200 to $1,800 box in 2028 will be competing with gaming PCs, the PS6 and whatever remains of the current console generation. Microsoft will need a clear answer to the question: why buy this instead of a PC or a PS6?

The operating-system fight inside Xbox

The Fortune report describes internal friction over the handheld's operating system. Windows leaders reportedly wanted the device to run a trimmed-down version of Windows Home. Sharma's team wanted its own software for better gaming performance. The disagreement got heated enough that a senior Xbox executive reportedly threatened to scrap the handheld rather than ship the Windows build, though Xbox now appears to be steering the OS.

Sharma's public comment in the report is telling: "We can't build Helix on our own." That is a statement about partnership, not surrender. Microsoft needs Windows, Xbox and silicon partners to make Helix work, but it also needs a software experience that feels like a console rather than a desktop with a controller.

The OS fight is the most important detail in the report because it shapes the product. A Windows handheld can run PC games, mods and storefronts, but it carries Windows overhead, driver complexity and a desktop interface that is not designed for a controller. A custom Xbox OS can be faster and simpler, but it may not run the full PC library and it may alienate players who expect Windows compatibility.

Microsoft has been moving toward a hybrid model for years: Xbox mode on Windows, PC Game Pass, and a console-like shell on top of a PC operating system. Helix is the next step in that strategy. The reported disagreement is not a sign that the strategy is failing. It is a sign that the company is still deciding how much Windows to put inside the box.

The subsidy question

Console pricing is usually a subsidy game. Microsoft, Sony and Nintendo have historically sold hardware at or below cost and made money on games, services and accessories. The $1,200 to $1,800 projection assumes Microsoft does not subsidize the console, or subsidizes it less than it did in previous generations.

Fortune puts Xbox's profit margin at around 3%, according to TechPowerUp. That is a thin margin for a hardware business, and it leaves little room for a large per-unit loss. If Microsoft wants to hit a $799 or $999 price, it will have to absorb a significant loss on every console or find savings elsewhere in the bill of materials.

There is also the question of who pays for the subsidy. Game Pass subscriptions, first-party software, accessories and storefront fees all contribute. A $1,200 console with a strong Game Pass attach rate can still be a good business. A $1,800 console with a weak attach rate is a hard sell, especially if the same games are available on PC and the handheld.

Microsoft has not said whether it will subsidize Helix. The report only says the projection assumes no subsidy unless the company decides otherwise. That leaves the final price open, and it leaves room for a launch price that is lower than the leaked range if Microsoft chooses to compete aggressively.

What it means for the console market

The console market is already moving upward in price. The Xbox Series X with a disc drive is $799.99. The PS5 Pro launched at a premium. Memory prices are rising, and both Sony and Microsoft are reportedly dealing with higher component costs. A $1,200 to $1,800 Helix would not be an outlier in that environment. It would be the logical endpoint.

That is bad news for anyone hoping for a cheap next-gen console. The era of the $399 console may be over. The next generation looks more like a family of devices: a handheld at one price, a premium console at another, and a PC-like box for players who want the highest performance. The console is becoming a premium product, not the default way to play.

The shift also changes the console war. If Helix costs $1,200 and the PS6 costs less, Sony can compete on price. If the PS6 also costs more than $1,000, the two platforms will compete on performance, ecosystem and exclusives rather than sticker price. The winner may be the company that can deliver the best experience per dollar, not the one with the lowest price.

Nintendo is the wild card. The Switch 2 is a different kind of product with a different price and a different audience. A $1,200 Xbox does not compete directly with a $450 Nintendo handheld. It competes with gaming PCs and with the idea that a console should be the cheapest way to play big-budget games.

Our take

The $1,200 to $1,800 range is a leak, not a launch price, and it should be reported that way. But the range is plausible for a 2028 console with a large AMD chip, a PC-compatible software stack and a memory market that is getting tighter. The days of the loss-leading $399 console are fading, and Helix may be the product that makes the shift official.

The handheld-first strategy is the more interesting bet. A 2027 handheld gives Microsoft a lower-priced entry point, a way to test its software stack and a year of feedback before the console arrives. If the handheld works, the console has a built-in audience. If it does not, Microsoft can adjust the console before launch.

The OS fight is the detail to watch. The difference between a Windows handheld and a custom Xbox handheld is the difference between a PC with a controller and a console you can carry. Microsoft has to choose, and the report suggests the company is still choosing. That decision will matter more than the price leak in the long run.

FAQ

How much will the Xbox Helix cost?
Fortune reports that Microsoft has internally projected $1,200 to $1,800 to consumers, unless the company subsidizes the console. Microsoft has not confirmed a price. Treat the range as a reported internal projection, not an official price.

When is the Xbox Helix coming out?
The report says the next-gen Xbox handheld is targeted for holiday 2027, with the Helix console following in time for the 2028 holidays. Microsoft has not confirmed either date.

Is the Helix a console or a PC?
Microsoft has described Helix as a family of devices and a console-first product with portable play in mind. The reported hardware uses a custom AMD system-on-chip and a PC-compatible software stack. The final product mix has not been announced.

Why would the Helix cost more than the Xbox Series X?
The reported price reflects a larger chip, more memory, a more complex cooling system and a tighter memory market in 2027 and 2028. The current Xbox Series X with a disc drive sells for $799.99 after recent price increases.

Will Microsoft subsidize the Helix?
Microsoft has not said. Fortune puts Xbox's profit margin at around 3%, which leaves little room for a large per-unit loss. The reported $1,200 to $1,800 range assumes no subsidy unless Microsoft decides otherwise.

What is the handheld's operating system?
The report says Windows leaders wanted a trimmed-down Windows Home build, while Sharma's team wanted its own software for better gaming performance. Xbox now appears to be steering the OS, but the final software has not been announced.

Bottom Line

Fortune reports that Microsoft has internally projected the Project Helix console at $1,200 to $1,800, with a handheld targeted for holiday 2027 and the console for holiday 2028. The figures are not confirmed by Microsoft, and the company could subsidize the console to bring the price down. The report also describes an internal fight over whether the handheld should run Windows or a custom Xbox OS. The price range is plausible in a market with rising memory costs and premium consoles, but it would mark a clear break from the $399 console era. The handheld-first strategy and the software decision will matter as much as the final number.

Sources

Fortune, "The non-gamer CEO racing to save Microsoft's Xbox," October 20, 2026: fortune.com
TechPowerUp, "Xbox Helix Console Reportedly Headed for Holiday 2028 at $1,200 to $1,800," October 20, 2026: techpowerup.com
TechPowerUp, "Micron CEO Says Memory Supply Will Be Much Tighter in 2027 and 2028 Than in 2026," October 1, 2026: techpowerup.com
TechPowerUp, "Xbox Helix Will Be 'a Great Console First' With Portable Play in Mind, Says CEO," October 6, 2026: techpowerup.com

GTA 6 Pre-Orders: 6.45 Million Digital Copies Before Launch

noreply@blogger.com (Unknown) — Wed, 21 Oct 2026 07:47:09 -0400

By Indie Kings | October 20, 2026

Updated October 20, 2026: Sensor Tower estimates that Grand Theft Auto VI has reached 6.45 million digital pre-orders worldwide, worth about $630 million net of returns. PlayStation accounts for 74% of the total, or roughly 4.80 million copies and $468 million, while Xbox accounts for 1.66 million copies and $162 million. Take-Two has not confirmed any of these figures, and the tracker covers digital copies only.

The one-sentence version: a third-party sales tracker says GTA 6 added about 370,000 pre-orders in a week and is now running far ahead of the biggest games of the past six years in the United States.

Image credit: Rockstar Games, via TechPowerUp.

What the latest estimate says

Sensor Tower's weekly update puts GTA 6 at 6.45 million digital pre-orders worldwide. The firm values those orders at roughly $630 million net of returns. That is an estimate of digital revenue, not a Take-Two earnings figure, and it excludes physical copies.

The previous update, dated October 13, had the game at 6.08 million pre-orders and $594 million. The new total represents an increase of about 370,000 copies in roughly a week. That pace is the story: GTA 6 is still more than four weeks from launch, and the pre-order curve is still climbing.

Metric	October 13 estimate	October 20 estimate	Change
Digital pre-orders	6.08 million	6.45 million	+370,000
Estimated net value	$594 million	$630 million	+$36 million
PlayStation share	Not restated	74%, about 4.80 million	About $468 million
Xbox share	Not restated	26%, about 1.66 million	About $162 million

Source: Sensor Tower sales tracker, via TechPowerUp, October 20, 2026. Figures are third-party estimates, not confirmed by Take-Two.

The PlayStation and Xbox splits do not add up to exactly 100% after rounding, and the dollar figures are rounded to the nearest million. Those are normal artifacts of a tracker that models a market rather than counting every receipt. The direction is what matters: PlayStation is carrying nearly three quarters of the digital pre-order volume.

PlayStation's 74% share

PlayStation's 74% share is not a surprise, but the size of the gap is. Sensor Tower puts PS5 pre-orders at about 4.80 million copies and $468 million, compared with 1.66 million copies and $162 million on Xbox. That is roughly a 2.9-to-1 advantage in units.

The split has been consistent. In late August, Sensor Tower had PlayStation at about 77% of 4.86 million pre-orders. The firm changed its estimation model on October 5, so the older percentage is not a perfect like-for-like comparison, but the direction has not flipped. PlayStation has led since tracking began in June.

There are two obvious explanations. The first is install base: the PS5 has sold more units than the Xbox Series X and Series S, so a game with GTA 6's reach will naturally skew toward the larger platform. The second is hardware momentum. US Xbox hardware sales are at an all-time low this year, according to the report, while average console prices are at record highs. A $100 game is easier to justify on a console people already own and still feel good about.

The share also matters for Sony. A 74% split on a game this large is a platform statement, not just a sales statistic. It means the default place to play the biggest launch of the year is a PlayStation console, and that is the kind of positioning Sony can use in marketing for the rest of the generation.

The Ultimate Edition is doing the heavy lifting

Take-Two CEO Strauss Zelnick said in August that pre-orders were skewing toward the $100 Ultimate Edition. Sensor Tower now says the vast majority of buyers went for that version. That is the detail that turns a unit-count story into a revenue story.

A higher average selling price means the $630 million net estimate is not simply 6.45 million copies at the standard edition price. It reflects a mix weighted toward the most expensive version. That is good for Take-Two's launch revenue, and it is good for the argument that GTA 6 can set records without discounting.

It also raises the stakes for the standard edition. If most early buyers are paying $100, the remaining pre-order window is where the cheaper version does its work. The split between editions will be one of the more interesting numbers to watch after launch, because it tells us whether GTA 6's audience is made up of committed fans buying the premium package or a broader group waiting for the base game.

The Ultimate Edition's lead is not a guarantee that every buyer will keep the order. Digital pre-orders can be canceled, and the net-of-returns figure is an attempt to account for that. But the direction is clear: the premium version is not a niche tier for collectors. It is the main pre-order product.

148% ahead of Call of Duty

In the United States, Sensor Tower says GTA 6's digital pre-orders are 148% ahead of Call of Duty: Modern Warfare 2 at the same point before launch. That makes GTA 6 the best performer among the big releases the firm has tracked over the past six years.

The comparison is useful because Call of Duty is the closest thing the industry has to a reliable annual blockbuster. Modern Warfare 2 launched in 2022 and was a massive seller. Being 148% ahead of that pace does not mean GTA 6 will sell 148% more copies over its lifetime. It means the pre-order curve is steeper at this stage, which is a leading indicator of launch-week demand.

The comparison also has limits. Call of Duty's audience is more annualized and more heavily weighted toward digital purchases on multiple platforms. GTA 6 is a once-a-generation release with a longer marketing runway and a much larger cultural footprint. The two games are not the same kind of product, even if they compete for the same entertainment dollars.

What the comparison does show is that GTA 6 is not relying on a narrow core audience. A 148% lead over Call of Duty in the US suggests demand is broad enough to outperform the most consistent AAA franchise in the market, at least at the pre-order stage.

Why these numbers are estimates, not sales

Sensor Tower's figures are estimates. The firm models digital sales from a combination of retailer data, panel data and its own tracking. It does not have access to Take-Two's internal sales ledger. The numbers are useful for direction, not for accounting.

There are three specific caveats. First, the tracker covers digital copies only. Physical pre-orders are excluded, which matters more in markets where disc sales remain strong. Second, the figures are net of returns, but returns can change between now and launch. Third, Sensor Tower changed its estimation model on October 5, so comparisons with the late-August numbers are not exact.

Take-Two has not confirmed the 6.45 million figure, the $630 million value or the platform split. The company's public comments have focused on the Ultimate Edition's performance, not on a specific pre-order total. Readers should treat the numbers as a credible third-party estimate, not as an official sales announcement.

The estimate is still worth reporting because it is the best public signal available before launch. Pre-order data is one of the few ways to measure demand before a game ships, and the trajectory matters more than any single week's total. A 370,000-unit weekly increase four weeks before launch is a strong pace by any measure.

What it means for November 19

GTA 6 launches on November 19 for PlayStation 5 and Xbox Series X|S. The pre-order numbers point to a launch that will strain digital storefronts, console supply and retail logistics at the same time.

The PlayStation skew is the biggest operational signal. If 74% of digital buyers are on PS5, Sony's storefront and network will carry most of the day-one download traffic. That does not guarantee a smooth launch, but it tells us where the pressure will be. Xbox's smaller share does not mean an easy launch on that platform; it means fewer copies, not no copies.

The Ultimate Edition mix matters for physical retail too. A premium digital version reduces the number of disc buyers, which is consistent with the industry's shift toward digital. It also means fewer used copies in circulation after launch, because digital purchases cannot be resold. That is good for Take-Two and for the game's long-tail revenue.

The console supply story is the wild card. PS5 Pro stock disappeared after the GTA 6 reveal, and average console prices are at record highs. If the game sells as many copies as the pre-order curve suggests, the limiting factor may not be demand. It may be how many consoles Sony and Microsoft can put on shelves in the weeks before and after November 19.

Our take

The 6.45 million figure is impressive, but the more important number is 370,000. That is the weekly increase, and it shows the pre-order curve has not flattened four weeks before launch. A game that was already the biggest entertainment launch of the year is still accelerating.

The PlayStation split is the second important signal. A 74% share is not just a reflection of install base; it is a reflection of momentum. Xbox hardware sales are at an all-time low, and GTA 6 is the kind of game that can make a console purchase feel urgent. If PlayStation captures three quarters of the digital pre-orders, it also captures three quarters of the launch conversation.

The caveats matter, but they do not cancel the story. Sensor Tower's numbers are estimates, Take-Two has not confirmed them, and digital-only tracking misses physical copies. Even with those limits, a 148% lead over Call of Duty at the same point is a remarkable data point. It suggests GTA 6 is not just a big game. It is the game the market has been waiting for since the last generation.

FAQ

How many GTA 6 pre-orders are there?
Sensor Tower estimates 6.45 million digital pre-orders worldwide as of October 20, 2026, worth about $630 million net of returns. Take-Two has not confirmed the figure. The estimate excludes physical copies.

How many pre-orders are on PlayStation and Xbox?
Sensor Tower estimates PlayStation at 74% of the total, or about 4.80 million copies and $468 million. Xbox accounts for about 1.66 million copies and $162 million. The shares are rounded and do not add up to exactly 100%.

Is GTA 6 beating Call of Duty?
In the US, Sensor Tower says GTA 6's digital pre-orders are 148% ahead of Call of Duty: Modern Warfare 2 at the same point before launch. That is a pre-order comparison, not a lifetime sales forecast.

Which edition are most buyers choosing?
Sensor Tower says the vast majority of buyers chose the $100 Ultimate Edition. Take-Two CEO Strauss Zelnick said in August that pre-orders were skewing toward that version.

Are these official sales figures?
No. They are third-party estimates from Sensor Tower, covering digital copies only and net of returns. Take-Two has not confirmed the numbers, and Sensor Tower changed its estimation model on October 5, so comparisons with earlier figures are not exact.

When does GTA 6 launch?
GTA 6 launches on November 19, 2026, for PlayStation 5 and Xbox Series X|S. A PC version has not been announced for launch.

Bottom Line

Sensor Tower estimates 6.45 million digital GTA 6 pre-orders worldwide, worth about $630 million net of returns, with PlayStation holding 74% of the total. The figure is up about 370,000 copies from the October 13 update, and the Ultimate Edition is carrying most of the volume. Take-Two has not confirmed the numbers, the tracker covers digital copies only, and the model changed on October 5, so the exact totals are estimates. The direction is still clear: four weeks before launch, GTA 6 is running 148% ahead of Call of Duty's pre-order pace in the US and is on track to be the biggest entertainment launch of the year.

Sources

TechPowerUp, "GTA 6 Digital Pre-Orders Climb to 6.45 Million, PlayStation Holds 74% Share," October 20, 2026: techpowerup.com
Sensor Tower, GTA VI sales tracker: sensortower.com
Games Industry, "Sensor Tower data shows the yawning gap between GTA 6 and everything else," October 2026: gamesindustry.biz
TechPowerUp, "GTA 6 Pre-Orders Jumped By 436% in a Single Day After Netflix Reveal, As PS5 Dominates Xbox," August 31, 2026: techpowerup.com