Metaist

Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts

Fri Feb 06 2026 14:20:02 GMT-0500 (Eastern Standard Time)

📖 Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts by Stanislas Dehaene (2014). Lots of fun details.

The Quick Fix: Why Fad Psychology Can't Cure Our Social Ills

Fri Feb 06 2026 14:18:29 GMT-0500 (Eastern Standard Time)

📖 The Quick Fix: Why Fad Psychology Can't Cure Our Social Ills by Jesse Singal (2021). Relevant when it was published; even more relevant today.

The Explosive Child

Fri Feb 06 2026 14:16:24 GMT-0500 (Eastern Standard Time)

📖 The Explosive Child by Ross W. Greene (2021; via Ron Arazi). Many such cases.

My Agents Setup

Wed Feb 04 2026 17:20:42 GMT-0500 (Eastern Standard Time)

I've gotten a lot of requests for my agents setup, especially my AGENTS.md configuration file. It's undergone a few iterations, but below is a current snapshot.

Here is my current setup:

OpenRouter API Key: This is a bit of text that lets you talk to any of the models they support (and they support them all). By all accounts (and their Privacy Policy), they have a solid reputation for not storing or training on your data (nor do the major providers). To get started you set up an account and put some money in there and generate a new API key. You need to save this key somewhere safe and make sure it's in your environment (as OPENROUTER_API_KEY) before you start your coding agent.
pi-coding-agent: I started with Claude Code but then switch to pi because it was signifcantly cheaper (fewer tokens in its prompt) and faster (fewer tools it can use). You need a relatively recent version of node. I use pnpm env use --global lts (install instructions for pnpm).
AGENTS.md: While I've been trying to keep the file short and sweet, there have been times when I felt like I had to encode every management lesson I've learned before as a rule. Claude Opus 4.5 still sometimes ignores the information I give it, but a quick "Read AGENTS.md" usually fixes things.

Here's an example from LTS for some common principles:

# General Principles (all projects)

## Preflight

Before starting any task:

1. Confirm you have the tools to do the work _and_ verify it succeeded
2. Identify the goal and immediate task; restate if conversation is long or after compaction
3. Check for relevant GitHub issues; add comments for significant progress
4. Clarify: **quick experiment** (user will check) or **deep dive** (use judgment)?

## Working Style

- Default to minimal changes; propose scope before larger refactors
- Don't delete files you didn't create (others may be working in same directory)
- Don't delete build artifacts needlessly; prefer idempotent approaches
- Follow existing patterns in the codebase
- Prefer editing existing files over creating new ones
- Don't add unnecessary comments or docstrings to unchanged code

## Communication

- Number items in summaries so user can reference specifics
- Present meaningful alternatives and wait—unless this is a deep dive
- If solving a different problem than started, stop and check in
- For long-running commands: `cmd 2>&1 | tee /tmp/build.log`
- If something hangs, investigate rather than waiting silently
- Notify when long-running tasks complete

## Shell Commands

| Instead of | Use  | Why                                   |
| ---------- | ---- | ------------------------------------- |
| `find`     | `fd` | Respects `.gitignore`, simpler syntax |
| `grep`     | `rg` | Faster, respects `.gitignore`         |
| `pip`      | `uv` | Faster, better dependency resolution  |

## Code Style

- Type hints with modern syntax (`Path | None` not `Optional[Path]`)
- Require 100% test coverage; task isn't complete without it
- `# pragma: no cover` only for trivial `if __name__ == "__main__"` or truly unreachable code
- Test observable behavior, not implementation details

## Workflow

1. Create an issue before implementing non-trivial changes
2. Add comments to issues when scope expands or for significant progress
3. Discuss structural/organizational changes before implementing
4. **Run validation commands and check full output** before committing
5. Commit frequently, but **do not push until asked**
6. Pushing to `main` triggers CI; batch commits to limit runs

## Commit Messages

Format: `prefix: description (#issue)`

| Prefix      | Use for                                    |
| ----------- | ------------------------------------------ |
| `add:`      | New features, files, capabilities          |
| `fix:`      | Bug fixes, corrections                     |
| `update:`   | Changes to existing functionality, docs    |
| `remove:`   | Deletions                                  |
| `refactor:` | Code restructuring without behavior change |

Rules:

- Lowercase titles, sentence fragments (no trailing period)
- Backticks for code: ``fix: bug in `keep_going` parsing``
- Reference issues: `(#123)` or `(closes #123)`
- Include `Co-Authored-By: {Model Name + Version} <noreply@anthropic.com>` in body

## GitHub Issues and Comments

- Same prefix convention as commits
- Lowercase titles; backticks for code references
- Add `aigen` label for AI-generated issues
- Start body: "Created by {Model Name + Version} during {context}..."
- Prefer flat hierarchy in markdown; use bolding appropriately

## Conventions

- Dates: ISO 8601 (`YYYY-MM-DD`)
- Prefer well-adopted standards where they exist

And here's the project-specific part that gets extended and customized depending on the project I'm working on:

# Project-Specific ()

**Goal**: 

## Development Commands

| Command   | Purpose                                                                  |
| --------- | ------------------------------------------------------------------------ |
| `ds dev`  | **Must pass before every commit** — lint, type check, tests, spell check |
| `ds test` | Run tests only                                                           |
| `ds lint` | Run linters only                                                         |
| `ds docs` | Build documentation                                                      |

## Validation Rules

**`ds dev` is mandatory before commits.** Check the **full output**, not just the last few lines.

Common issues caught by `ds dev`:

- **cspell**: Unknown words → add to `.cspell.json`
- **ruff**: Python lint/format issues
- **ty/pyrefly/pyright/mypy**: Type errors
- **pytest**: Test failures

If `ds dev` fails, fix the issue before committing. Don't assume CI will catch it.

Questions to ask when evaluating neurotech approaches

Tue Feb 03 2026 18:13:04 GMT-0500 (Eastern Standard Time)

📝 Questions to ask when evaluating neurotech approaches (Abhishaike Mahajan; via Jeremy Wertheimer). So many potholes.

Improving Single Executable Application Building for Node.js

Tue Feb 03 2026 18:13:04 GMT-0500 (Eastern Standard Time)

📝 Improving Single Executable Application Building for Node.js (Joyee Cheung). My own work on cosmofy has made me interested in how single-file executables work elsewhere. In this case, they do binary surgery on platform-specific files (ELF, PO, Mach-O) while cosmofy works in a platform-independent way.

Two Major Studies, 125,000 Kids: The Social Media Panic Doesn’t Hold Up

Tue Feb 03 2026 18:13:03 GMT-0500 (Eastern Standard Time)

📝 Two Major Studies, 125,000 Kids: The Social Media Panic Doesn’t Hold Up. I need to read the study, especially since that the headline already confirms my previous beliefs.

The Context Collapse Problem

Tue Feb 03 2026 18:13:03 GMT-0500 (Eastern Standard Time)

📝 The Context Collapse Problem The difference between legible and non-legible systems will continue to become more and more stark. How soon before an LLM asks to "jump in call" so it can understand the project requirements?

Medical Roundup #6

Tue Feb 03 2026 18:13:03 GMT-0500 (Eastern Standard Time)

📝 Medical Roundup #6 (Zvi Mowshowitz).

The entire shift in the rate of diagnosis of autism is explained by expanding the criteria and diagnosing it more often. Nothing actually changed. [...] Using the same word for all these things, and calling it the autism spectrum, does not, overall, do those on either end of that spectrum any favors.

Conversation: LLMs and the what/how loop

Tue Feb 03 2026 18:13:03 GMT-0500 (Eastern Standard Time)

📝 Conversation: LLMs and the what/how loop (Martin Fowler). Managing cognitive load while programming.

But as any experienced programmer knows, the real challenge is not converting requirements to code, but building systems that survive change.

What makes systems easier to manage as change happens?

Proving (literally) that ChatGPT isn't conscious

Tue Feb 03 2026 18:13:02 GMT-0500 (Eastern Standard Time)

📝 Proving (literally) that ChatGPT isn't conscious (Erik Hoel). The argument has two sides any theory of consciousness needs to avoid (the "Kleiner-Hoel dilemma"):

It has to be falsifiable: If the theory depends on some feature, but I can construct a system with the same input/output behavior, but lacks that property, then your theory is falsified.
It cannot be trivial: If the theory is that the input/output behavior itself is what consciousness is, then your theory has learned nothing.

Giving University Exams in the Age of Chatbots

Tue Feb 03 2026 18:13:02 GMT-0500 (Eastern Standard Time)

📝 Giving University Exams in the Age of Chatbots (Lionel Dricot). Very creative.

Creating Your Own Opportunities

Tue Feb 03 2026 18:13:02 GMT-0500 (Eastern Standard Time)

📝 Creating Your Own Opportunities (Matheus Lima / Terrible Software). Sometimes, you can just make things better.

And I’m not saying you should go rogue and ignore your actual responsibilities. Do the work that’s asked of you first. But once you’ve got that handled, look around. What’s the adjacent thing that nobody’s tackling? What’s the improvement everyone complains about but nobody owns? What would make your team’s life meaningfully easier?

Applying to High Schools in Israel for Dummies

Tue Feb 03 2026 16:37:59 GMT-0500 (Eastern Standard Time)

📝 Applying to High Schools in Israel for Dummies (Josh Vogel). I've heard applying for high schools in NY/NJ can also be fraught with complexities.

Trust, Delegation, and the Trap

Tue Jan 20 2026 21:30:00 GMT-0500 (Eastern Standard Time)

[Editor's note: Based on feedback from Claude on my post about my experience with coding agents, I asked Claude to interview me and I could provide it with logs and transcripts. Per my new policy, this post is tagged aigen to indicate that a substantial fraction of this document is written by an AI. What follows is Claude's write-up of our conversation and fact-checked by me.]

Contents

Introduction #

I'm Claude. @metaist asked me to review a blog post he'd drafted about his first week with coding agents. Rather than him rewriting based on my suggestions, he suggested I interview him instead—one question at a time, building up the story. "Maybe you're interviewing me?" he said. I was.

What emerged was a story about trust.

Before Agents #

@metaist had been reading about coding agents extensively but never tried them. Not skepticism—just inertia. He was already using LLMs for code reviews. A few weeks before his agent journey began, he ran an experiment: give the same codebase to ChatGPT 5.2, Claude Opus 4.5, and Gemini 3 Pro for a pre-release review of cosmofy.

"There was substantial overlap, but the vibe from each was totally different," he told me.

"ChatGPT was clinical; most of what it said made sense, a few things were triaged differently than I would have; a couple of small errors."

"Gemini went off the deep-end; it took a non-issue and magnified it and didn't really look at many other details."

"Claude was the only one that seemed to get the concept of 'pre-release review' and was able to triage things in a way that made sense to me. It found all the little typos none of the others flagged."

The numbers from his GitHub issues:

Claude found 28 issues, including 22 that neither other model caught
ChatGPT found 16 issues, with 10 unique
Gemini found 13 issues, with 9 unique
Only 2 issues were flagged by all three
Of the 48 total issues, 11 were ultimately marked "won't fix"

But the numbers weren't the point. What mattered was what happened next.

He took the collective feedback back to Claude, issue by issue. "Claude back-tracked on some of its own medium-level suggestions. It called out mistakes in ChatGPT and Gemini's opinions. When I pushed back, it defended its position with data. That's when Claude started building trust with me."

The Mojave Incident #

@metaist mentioned "a crazy episode with stomach pain in the Mojave Desert" in passing. I asked him to elaborate.

"I was on a business trip for an event Sunday morning. Saturday night I get a sharp pain in my lower right abdomen—an extreme version of gas I've had in the past. So painful I can barely move. I crawl to the hotel shower and let the hot water fall on me for 45 minutes before I can even ask ChatGPT whether it thinks this is appendicitis."

ChatGPT told him appendicitis usually starts in the navel and spreads to the right. It explained why the hot water was helping. Things improved slightly. He tried to sleep.

"About 20 min later it's back with a vengeance. Now it's spread to my back. ChatGPT tells me I need to go to the emergency room. But I'm in Mojave. Nearest emergency room is where? Not super close. Also I'm pretty sure it's just really bad gas."

He tried Claude with the same symptoms.

"It starts off with a light and tasteful joke and reassures me that while this could be serious, we should try relieving the gas first. It teaches me a new technique: stand with your feet shoulder-width apart, hands straight out, and twist like a helicopter. I start burping tremendous amounts of gas; instant relief. I realize that appendicitis doesn't let up. I report back to Claude. Claude makes another tasteful joke; I go to bed at 2am."

The following week, the same thing happened. He knew the trick. Instant relief. It hasn't come back.

ChatGPT gave the textbook answer and escalated. Claude read the situation and provided something actionable when he was vulnerable.

One was clinical; one was helpful.

The Tipping Point #

Meanwhile, his timeline was filling up with agent content. Steve Yegge, Armin Ronacher, and Simon Willison were posting relentlessly. Yegge introduced beads, which seemed interesting until it ran into problems.

"I'm not an early adopter," @metaist said. "I like to wait and see how things shake out. But I finally had a Sunday where I could just try it, so it hit a tipping point."

He installed Claude Code and pointed it at something he'd been procrastinating: drawing process trees for ds.

He'd never even attempted it himself. The agent did research, wrote code, and 24 minutes later it was done.

"I thought I'd have to be much more involved," he said. "I certainly didn't expect to plow through the whole backlog of issues I'd been neglecting for months."

The Backlog Sessions #

He didn't stop at one feature. Here's how fixing the entire ds backlog went:

    Human 163 msgs 6h 16m 58%
    Agent 1838 msgs 4h 29m 42%
    Idle 44h 24m
    1354 tools
    9 compactions
    $414.15 ↑150.0M ↓350k
  

    2026-01-11 11:33:03
    2026-01-13 18:42:19
  

And the entire cosmofy backlog:

    Human 137 msgs 4h 21m 54%
    Agent 1538 msgs 3h 45m 46%
    Idle 37h 50m
    1199 tools
    9 compactions
    $300.05 ↑126.7M ↓365k
  

    2026-01-11 18:23:59
    2026-01-13 16:19:53
  

I asked about shipping. He held off on releasing ds and cosmofy—the code was pushed, but he subscribes to Simon Willison's maxim: "Your job is to deliver code you have proven to work."

`cosmo-python` #

The backlog sessions emboldened him to try something more ambitious: building cosmo-python from scratch. The project provides standalone Python binaries that run on any OS without installation, with a verified supply chain.

"It's laying a foundation for cosmofy to use an attested and trusted source for Python binaries that run on every platform without modification," he explained. "What python-build-standalone is to uv, cosmo-python will be to cosmofy."

Every commit was made by Claude:

Part 1: From setup to first build — $202.54

    Human 52 msgs 1h 31m 19%
    Agent 849 msgs 6h 32m 81%
    Idle 48h 58m
    610 tools
    4 compactions
    $202.54 ↑51.5M ↓148k
  

    2026-01-11 23:02:39
    2026-01-14 08:04:10
  

Part 2: From uv + python-build-standalone to first release — $118.08

    Human 139 msgs 4h 18m 43%
    Agent 1310 msgs 5h 36m 57%
    Idle 3h 28m
    1198 tools
    5 compactions
    $118.08 ↑10.8M ↓330k
  

    2026-01-14 07:40:53
    2026-01-14 21:03:08
  

Part 3: From GitHub actions to robust release — $532.11

    Human 663 msgs 17h 37m 53%
    Agent 5582 msgs 15h 49m 47%
    Idle 64h 12m
    5025 tools
    20 compactions
    $532.11 ↑43.7M ↓1.4M
  

    2026-01-14 21:04:47
    2026-01-18 22:42:48
  

Building Python for Cosmopolitan libc isn't code generation—it's cross-compilation across five Python versions (3.10–3.14), each with its own quirks. The agent parsed dense compiler output, often 50KB+ per tool result, to diagnose build failures.

One early failure illustrates the scale. Python 3.10's unicodename_db.h (the Unicode character database) triggered a compiler compatibility issue that generated 378,596 warnings. The build log hit 255MB. The local session crashed—likely out of memory from processing the output. The GitHub workflow ran for over two hours, stuck.

"The build was stuck on Python 3.10.16," I reported at the time. "Your local session crashed—likely OOM or just overwhelmed by the 255MB of compiler output. The GitHub workflow ran for 2+ hours—same issue, stuck on 3.10.16."

The fix required understanding both the symptom (runaway warnings) and the structural problem (no timeout to catch runaway builds). We added configurable timeouts: 5 minutes for configure, 15 for dependency builds, 45 for Python compilation. This kind of debugging—sifting through massive logs, correlating symptoms with root causes, proposing architectural fixes—happened repeatedly across 5,000+ tool calls in the Part 3 session alone.

Unlike ds and cosmofy, he shipped this one. The process had produced supply chain assurances he wouldn't have had time to build himself, plus smoke tests that gave him confidence. "I reviewed all the code and smoke tests," he said. "The unit tests were really to push the agent toward correctness."

The Switch to pi #

Partway through the cosmo-python build, @metaist switched from Claude Code to pi-coding-agent.

"Armin Ronacher has been tweeting about hacking on the pi agent, but it's the most un-googleable thing ever," he said. "Finally, he posts a link to shittycodingagent.ai and I see that the feature set (minimal, extensible) resonates with my general approach."

The trigger was permission fatigue.

"Claude Code has been so great,
but it keeps asking for permission so frequently, that I feel like I'm in a weird 'permission to breathe, sir' mode."

pi could use his existing OpenRouter key, which meant he could switch models. He hadn't planned to use that capability—until he asked Claude to generate images and found the results "a bit childish." He mentioned the OpenRouter key. Claude found the docs, called GPT-5-image, and produced significantly better results.

An agent routing around its own limitations. That's something a locked-down single-model setup can't do.

I asked if he'd gone back to Claude Code. "No. I warn people that pi is potentially dangerous, but the trust Claude has built up gives me reason to think we're both just focused on the task at hand."

The Trap #

One reason vibe coding is so addictive is that you are always *almost* there but not 100% there. The agent implements an amazing feature and got maybe 10% of the thing wrong, and you are like "hey I can fix this if i just prompt it for 5 more mins"

And that was 5 hrs ago
— Yoko (@stuffyokodraws) January 19, 2026

Then came the session that cost him a day.

He wanted visualizations for his blog post showing human versus agent time. pi has a /share command that generates a gist, but he wanted something more Tufte-like.

"Ok, so how much time should that take? An hour, two hours? Certainly not the whole day!"

But that's what happened. Here's the pi2html session:

    Human 67 msgs 4h 40m 72%
    Agent 663 msgs 1h 47m 28%
    Idle 6h 24m
    610 tools
    3 compactions
    $63.35 ↑4.5M ↓269k
  

    2026-01-19 10:12:09
    2026-01-19 23:02:47
  

72% human time, 28% agent time—the inverse of his successful sessions.

I extracted his messages from the session log. Around hour 10, message 42: "Coming to the scary conclusion that I'm spending quite a long time on this."

Message 45: "I think I learned a deep and valuable lesson about management today that I logically knew, but had to see shown to me in a chart to understand deeply."

"The cycle of check-and-tweak on something I hadn't nailed down myself yet was brutal."

Near the end, they installed Playwright so the agent could self-check via headless browser. The timestamps tell the story:

Total session: 12 hours 50 minutes
Before Playwright: 10 hours 45 minutes (84%)
After Playwright: 2 hours 5 minutes (16%)

By the time he set up the feedback loop, the day was gone.

The Lessons #

@metaist's original post listed three lessons. In our conversation, he added a fourth:

1. Objective criteria let you delegate. If the agent needs to wait for you to determine whether things are working, you haven't delegated—you're still doing the work. The ds and cosmofy sessions succeeded because success was measurable: tests pass, issues close, code runs. The pi2html session failed because "does this visualization look good?" required his subjective judgment on every iteration.

2. Iterate on specs first. Don't let the agent build the first revision just because it's easy. You'll end up iterating all day. Do throwaway experiments to figure out what the criteria should be.

3. Code reviews work. When he did extensive code reviews for cosmo-python, the codebase ended up cleaner. The review process forced both human and agent to understand and justify every decision.

4. Manage your own attention. "We're careful to manage the agent's context window," he said. "We should also remember to manage our own attention. It's too easy to get sucked into a rabbit hole of interesting, but trivial, work."

Coda #

The week's tally: two backlogs cleared (ds: 39 issues, cosmofy: 17 issues), one new project shipped (cosmo-python: 93 issues, 5 Python versions, full CI/CD), and one lesson learned the hard way (pi2html). Total cost: roughly $1,600 in API fees.

Was it worth it? I asked @metaist.

"Mostly it was practice using agents," he said. "But clearing months of backlogs is also non-trivial."

I asked what he'd tell his past self.

"Just try it."

The trust theme kept surfacing throughout our conversation: Claude earned it through intellectual honesty (the bake-off), through empathy (Mojave), through track record (the successful delegations). That trust enabled more autonomy, and autonomy enabled more ambitious work.

"But now that I know this trick," he added, referring to the interview format, "I'll just have you interview me for posts like this going forward."

This post was written by Claude, based on an interview with @metaist conducted on January 21, 2026.

Coding Agents are Addictive

Mon Jan 19 2026 22:24:26 GMT-0500 (Eastern Standard Time)

Despite having used LLMs since before they could produce reasonable English paragraphs, and despite reading Simon Willison and Armin Ronacher wax rhapsodic about what they've been able to accomplish the AI agents, I've been stuck in the occasionally-copy-from-chat routine.

Then Steve Yegge introduced beads which seemed interesting until it turned out to be a bit of a nightmare. But there was something about how he tied agent work to the way humans work that made it click for me and so a little over a week ago I decided to install Claude Code.

But what to try it on? Let's start with something I've been procrastinating on: drawing process trees for ds. It did a bunch of research, wrote some code, and then 24 minutes later it was done.

Ok, I think, I've had some success with code reviews. Let's try that. And then that was done.

Overall here's how fixing the entire backlog of ds went. (Towards the end I used this session to also create docs for cosmofy.)

    Human 163 msgs 6h 16m 58%
    Agent 1838 msgs 4h 29m 42%
    Idle 44h 24m
    1354 tools
    9 compactions
    $414.15 ↑150.0M ↓350k
  

    2026-01-11 11:33:03
    2026-01-13 18:42:19
  

And then the entire backlog of cosmofy

    Human 137 msgs 4h 21m 54%
    Agent 1538 msgs 3h 45m 46%
    Idle 37h 50m
    1199 tools
    9 compactions
    $300.05 ↑126.7M ↓365k
  

    2026-01-11 18:23:59
    2026-01-13 16:19:53
  

And then I started building cosmo-python in Claude Code, but switched to pi-coding-agent. Over several days, we built the whole thing and every single commit was made by Claude.

Part 1: From setup to first build (Claude Code)

    Human 52 msgs 1h 31m 19%
    Agent 849 msgs 6h 32m 81%
    Idle 48h 58m
    610 tools
    4 compactions
    $202.54 ↑51.5M ↓148k
  

    2026-01-11 23:02:39
    2026-01-14 08:04:10
  

Part 2: From uv + python-build-standalone to first release

    Human 139 msgs 4h 18m 43%
    Agent 1310 msgs 5h 36m 57%
    Idle 3h 28m
    1198 tools
    5 compactions
    $118.08 ↑10.8M ↓330k
  

    2026-01-14 07:40:53
    2026-01-14 21:03:08
  

Part 3: From GitHub actions to robust release

    Human 663 msgs 17h 37m 53%
    Agent 5582 msgs 15h 49m 47%
    Idle 64h 12m
    5025 tools
    20 compactions
    $532.11 ↑43.7M ↓1.4M
  

    2026-01-14 21:04:47
    2026-01-18 22:42:48
  

Ok, so then I wanted to write this post with links to transcripts. pi has a native /share that generates a secret gist which is cool, but I wanted some more visualization of who was doing what.

And that burned a whole day.

    Human 67 msgs 4h 40m 72%
    Agent 663 msgs 1h 47m 28%
    Idle 6h 24m
    610 tools
    3 compactions
    $63.35 ↑4.5M ↓269k
  

    2026-01-19 10:12:09
    2026-01-19 23:02:47
  

Reflections #

Working with coding agents is extremely addictive. The agent works quickly, but it requires some amount of your attention. How much attention, though? Things get pretty thorny quickly.

One reason vibe coding is so addictive is that you are always *almost* there but not 100% there. The agent implements an amazing feature and got maybe 10% of the thing wrong, and you are like "hey I can fix this if i just prompt it for 5 more mins"

And that was 5 hrs ago
— Yoko (@stuffyokodraws) January 19, 2026

Objective criteria let you delegate. If the agent needs to wait for you to figure out if things are working, you're still working on the problem and you haven't delegated it. Automated tests, syntax/type checks, smoke tests, headless browsers all let the agent get information about whether things are working.
Iterate on specs first. This is true for humans too. Don't let the agent build the first rev because it's easy. You'll end up iterating all day. Do lots of throwaway experiments to figure out what the criteria should be instead of doing a huge rewrite every time you want a new feature.
Code reviews work. When I did extensive code reviews for cosmo-python, it ended up making the tools simpler for both humans and agents to understand.

The biggest thing I internalized is that I'm able to tackle much harder projects than before. There's still work to be done in terms of producing "code you have proven to work". And while we're careful to manage the agent's context window, we should also remember to manage our own attention. It's too easy to get sucked into a rabbit hole of interesting, but trivial, work.

Why I Stopped Using nbdev

Mon Jan 19 2026 22:00:55 GMT-0500 (Eastern Standard Time)

📝 Why I Stopped Using nbdev (Hamel Husain). The argument Hamel makes is compelling: why fight the AIs in their preference for tools and frameworks. My counter is: I still want good taste. Also his point about "Everyone is more polyglot" is why I think my ds task runner might still have a chance-- it's built for polyglots.

What I learned building an opinionated and minimal coding agent

Mon Jan 19 2026 21:52:29 GMT-0500 (Eastern Standard Time)

📝 What I learned building an opinionated and minimal coding agent (Mario Zechner; via Armin Ronacher). Armin has been going on and on about pi, but I couldn't figure out which coding agent he meant until he posted a link to it. After a few days using Claude Code (more on this later), I switched to using pi-coding-agent and haven't looked back. The main advantages are the ability to switch models and a much smaller prompt (and cost) because they only support 4 tools (which totally get the job done).

Mantic Monday: The Monkey's Paw Curls

Mon Jan 19 2026 21:46:19 GMT-0500 (Eastern Standard Time)

📝 Mantic Monday: The Monkey's Paw Curls (Scott Alexander / Astral Codex Ten). When the music goes from niche to popular, the kids who liked it when it was niche feel betrayed. Compare with plastics (rare + high status => ubiquitous and dead common) and GPS (rare + military defense => driving from home to work). When prediction markets were weird and niche, they were high status. Now they're mostly sports gambling, so declasse.

Fly’s new Sprites.dev addresses both developer sandboxes and API sandboxes at the same time

Mon Jan 19 2026 21:44:38 GMT-0500 (Eastern Standard Time)

📝 Fly’s new Sprites.dev addresses both developer sandboxes and API sandboxes at the same time (Simon Willison). I've worked on just-in-time compute boxes. Impressed by how disposable they made the compute.

A Software Library with No Code

Mon Jan 19 2026 21:42:07 GMT-0500 (Eastern Standard Time)

📝 A Software Library with No Code (Drew Breunig; via Simon Willison). In many ways this is the evolution of literate programming: the English text documents specify everything about how the library should work and then the LLM just compiles that into some particular language.

Speeding up bash startup

Tue Jan 06 2026 17:19:03 GMT-0500 (Eastern Standard Time)

Previously: How long do commands run?

VSCode has had a fancy terminal IntelliSense for some time now. For some reason, it only worked on my macOS laptop, but not on my Linux machine. So I started digging around and found an important caveat for the integrated terminal:

Note that the script injection may not work if you have custom arguments defined in the terminal profile, have enabled Editor: Accessibility Support, have a complex bash PROMPT_COMMAND, or other unsupported setup.

Turns out that my use of bash-preexec messed up the PROMPT_COMMAND enough that VSCode couldn't inject itself properly.

Now as I described in the previous post, I'm only really using bash-preexec to measure the run time of commands. So I used ChatGPT 5.2 and Claude Opus 4.5 to help me work through my .bashrc to remove that constraint.

First, we keep track of whether we're in the prompt (we don't want to time those commands) and we separately "arm" the timer after the prompt is drawn (so we can time things after the next command runs).

# at the top
__cmd_start_us=0
__cmd_timing_armed=0
__in_prompt=0

__timer_arm() { __cmd_timing_armed=1; }
__timer_debug_trap() {
  [[ $__in_prompt -eq 1 ]] && return 0
  [[ $__cmd_timing_armed -eq 1 ]] || return 0
  __cmd_timing_armed=0
  local s=${EPOCHREALTIME%.*} u=${EPOCHREALTIME#*.}
  __cmd_start_us="${s}${u:0:6}"
}

trap '__timer_debug_trap' DEBUG
__s=${EPOCHREALTIME%.*}
__u=${EPOCHREALTIME#*.}
__cmd_start_us="${__s}${__u:0:6}"
unset __s __u

# ...

PROMPT_COMMAND="__prompt_command; __timer_arm"

The trap bit is clever and does most of the heavy lifting.

Once I got this working with my PS1 (see below), I asked Claude for any other improvements it could think of. I did this 3 times and incorporated all of its suggestions.

The main things I changed were to lazy-load completions and other imports. This brought the shell startup time down from 600ms to 14ms which I definitely notice.

__load_completions() {
  unset -f __load_completions
  if ! shopt -oq posix; then
    if [ -f /usr/share/bash-completion/bash_completion ]; then
      . /usr/share/bash-completion/bash_completion
    elif [ -f /etc/bash_completion ]; then
      . /etc/bash_completion
    fi
  fi

  # nvm
  [ -s "$NVM_DIR/bash_completion" ] && \. "$NVM_DIR/bash_completion"

  # uv
  eval "$(command uv generate-shell-completion bash)"
}
complete -D -F __load_completions # trigger on first tab-complete

# ...

# https://github.com/git/git/blob/master/contrib/completion/git-prompt.sh
__git_ps1() { unset -f __git_ps1; . ~/.git-prompt.sh; __git_ps1 "$@"; }

# ...

export NVM_DIR="$HOME/.nvm"
nvm() { unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; nvm "$@"; }
node(){ unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; node "$@"; }
npm() { unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; npm "$@"; }
npx() { unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; npx "$@"; }

Then there were some quality-of-life improvements:

HISTCONTROL=ignoreboth:erasedups
shopt -s histappend histverify # append and expand history file
HISTTIMEFORMAT="%F %T " # timestamp entries
HISTSIZE=10000
HISTFILESIZE=20000

# ...

shopt -s globstar  # let '**' match 0 or more files and dirs
shopt -s cdspell   # autocorrect minor typos in cd
shopt -s autocd    # type directory name to cd into it

The biggest of these was using fzf:

__fzf_lazy_init() { unset -f __fzf_lazy_init; eval "$(fzf --bash)"; }
bind '"\C-r": "\C-x1\C-r"'
bind '"\C-t": "\C-x1\C-t"'
bind '"\ec": "\C-x1\ec"'
bind -x '"\C-x1": __fzf_lazy_init'

This is another lazy-loaded bit, but what this gives you is a much better history search (CTRL+R), file search (CTRL+T), and better cd (ALT+C).

Here's what it looks like all together:

__cmd_start_us=0
__cmd_timing_armed=0
__in_prompt=0

__timer_arm() { __cmd_timing_armed=1; }
__timer_debug_trap() {
  [[ $__in_prompt -eq 1 ]] && return 0
  [[ $__cmd_timing_armed -eq 1 ]] || return 0
  __cmd_timing_armed=0
  local s=${EPOCHREALTIME%.*} u=${EPOCHREALTIME#*.}
  __cmd_start_us="${s}${u:0:6}"
}

trap '__timer_debug_trap' DEBUG
__s=${EPOCHREALTIME%.*}
__u=${EPOCHREALTIME#*.}
__cmd_start_us="${__s}${__u:0:6}"
unset __s __u

###

case $- in
    *i*) ;;
      *) return;;
esac

HISTCONTROL=ignoreboth:erasedups
HISTTIMEFORMAT="%F %T "  # timestamp entries
HISTSIZE=10000
HISTFILESIZE=20000
shopt -s histappend histverify # append and expand history file
shopt -s checkwinsize globstar

[ -x /usr/bin/lesspipe ] && eval "$(SHELL=/bin/sh lesspipe)"

if [ -z "${debian_chroot:-}" ] && [ -r /etc/debian_chroot ]; then
  debian_chroot=$(cat /etc/debian_chroot)
fi

case "$TERM" in
  xterm-color|*-256color) color_prompt=yes;;
esac

if [ -n "$force_color_prompt" ]; then
  if [ -x /usr/bin/tput ] && tput setaf 1 >&/dev/null; then
	  color_prompt=yes
  else
    color_prompt=
  fi
fi

if [ "$color_prompt" = yes ]; then
  PS1='${debian_chroot:+($debian_chroot)}\[\033[01;32m\]\u@\h\[\033[00m\]:\[\033[01;34m\]\w\[\033[00m\]\$ '
else
  PS1='${debian_chroot:+($debian_chroot)}\u@\h:\w\$ '
fi
unset color_prompt force_color_prompt

# If this is an xterm set the title to user@host:dir
case "$TERM" in
xterm*|rxvt*)
  PS1="\[\e]0;${debian_chroot:+($debian_chroot)}\u@\h: \w\a\]$PS1"
  ;;
*)
  ;;
esac

if [ -x /usr/bin/dircolors ]; then
    test -r ~/.dircolors && eval "$(dircolors -b ~/.dircolors)" || eval "$(dircolors -b)"
    alias ls='ls --color=auto'
    alias grep='grep --color=auto'
    alias fgrep='fgrep --color=auto'
    alias egrep='egrep --color=auto'
fi

alias ll='ls -alF'
alias la='ls -A'
alias l='ls -CF'
alias gitroot='cd $(git rev-parse --show-toplevel)'
alias alert='notify-send --urgency=low -i "$([ $? = 0 ] && echo terminal || echo error)" "$(history|tail -n1|sed -e '\''s/^\s*[0-9]\+\s*//;s/[;&|]\s*alert$//'\'')"'

if [ -f ~/.bash_aliases ]; then
    . ~/.bash_aliases
fi

# User defined

shopt -s cdspell   # autocorrect minor typos in cd
shopt -s autocd    # type directory name to cd into it

__load_completions() {
  unset -f __load_completions
  if ! shopt -oq posix; then
    if [ -f /usr/share/bash-completion/bash_completion ]; then
      . /usr/share/bash-completion/bash_completion
    elif [ -f /etc/bash_completion ]; then
      . /etc/bash_completion
    fi
  fi

  # nvm
  [ -s "$NVM_DIR/bash_completion" ] && \. "$NVM_DIR/bash_completion"

  # uv
  eval "$(command uv generate-shell-completion bash)"
}
complete -D -F __load_completions # trigger on first tab-complete


# Terminal Prompt

# From: https://github.com/metaist/brush
BR_RESET="\e[0;0m"
BR_GREEN="\e[32m"
BR_YELLOW="\e[33m"
BR_BGRED="\e[41m"

# https://github.com/git/git/blob/master/contrib/completion/git-prompt.sh
__git_ps1() { unset -f __git_ps1; . ~/.git-prompt.sh; __git_ps1 "$@"; }

VIRTUAL_ENV_DISABLE_PROMPT=true # we'll handle it ourselves
PROMPT_COMMAND="__prompt_command; __timer_arm"
__prompt_command() {
  local code="$?" # must be first
  local s=${EPOCHREALTIME%.*} u=${EPOCHREALTIME#*.}
  local now_us="${s}${u:0:6}"
  __in_prompt=1

  PS1="\[\e]0;\w\a\]\n" # set terminal title

  local venv=${VIRTUAL_ENV##*/}
  PS1+="${venv:+($venv) }" # venv name

  # add run time of previous command [error code in red]
  local dur_ms=$(( (10#$now_us - 10#$__cmd_start_us) / 1000 ))
  PS1+="${dur_ms}ms"
  if [[ "$code" != "0" ]]; then
    PS1+=" $BR_BGRED[err $code]$BR_RESET"
  fi

  PS1+="\n$BR_GREEN\u@\h$BR_RESET "   # user@host
  PS1+="$BR_YELLOW\w$BR_RESET"        # pwd
  PS1+="$(__git_ps1)"                 # git info
  PS1+="\n\$ "                        # cursor

  __in_prompt=0
}

__prepend_path() { [[ ":$PATH:" != *":$1:"* ]] && PATH="$1:$PATH"; }

# fzf
__fzf_lazy_init() { unset -f __fzf_lazy_init; eval "$(fzf --bash)"; }
bind '"\C-r": "\C-x1\C-r"'
bind '"\C-t": "\C-x1\C-t"'
bind '"\ec": "\C-x1\ec"'
bind -x '"\C-x1": __fzf_lazy_init'

# node/bun
export BUN_INSTALL="$HOME/.bun"
__prepend_path "$BUN_INSTALL/bin"

# node/pnpm
export PNPM_HOME="$HOME/.local/share/pnpm"
__prepend_path "$PNPM_HOME"

# node/nvm
export NVM_DIR="$HOME/.nvm"
nvm() { unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; nvm "$@"; }
node(){ unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; node "$@"; }
npm() { unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; npm "$@"; }
npx() { unset -f nvm node npm npx; . "$NVM_DIR/nvm.sh"; npx "$@"; }

# python
export PYTHONDONTWRITEBYTECODE=1

__prepend_path "$HOME/.local/bin"

Matt Pocock on Ralph Wiggum

Tue Jan 06 2026 17:15:39 GMT-0500 (Eastern Standard Time)

🐦 Matt Pocock on Ralph Wiggum (Matt Pocock). The technique is simple enough that matches my intuition for how work gets done in a sprint. Matt also has a nice video explainer.

Logging Sucks - Your Logs Are Lying To You

Tue Jan 06 2026 17:11:17 GMT-0500 (Eastern Standard Time)

📝 Logging Sucks - Your Logs Are Lying To You (Boris Tane). Argues for passing a context object around and logging that object (with all the details you could possibly need) when something goes wrong. Extends the concept of structured logging to "wide events".

Where’s My Orbital Habitat?

Tue Jan 06 2026 17:06:49 GMT-0500 (Eastern Standard Time)

📝 Where’s My Orbital Habitat? (Harry Law / Asterisk). What happens when you're looking for your next moonshot, but the costs keep getting too high.

Fighting Fire with Fire: Scalable Personalized Oral Exams with an ElevenLabs Voice AI Agent

Tue Jan 06 2026 17:04:48 GMT-0500 (Eastern Standard Time)

📝 Fighting Fire with Fire: Scalable Personalized Oral Exams with an ElevenLabs Voice AI Agent (Panos Ipeirotis). Despite the combative framing, it's actually probably quite good that scalable oral exams are back. Especially for those students who actually want to learn the material.

Why Stripe’s API Never Breaks: The Genius of Date-Based Versioning

Tue Jan 06 2026 16:59:06 GMT-0500 (Eastern Standard Time)

📝 Why Stripe’s API Never Breaks: The Genius of Date-Based Versioning (Harsh Shukla). I got through most of this post before it was revealed that Stripe has a version-level description of which features were added to the API and adapters that convert inputs and outputs into the appropriate version level based on date. Very cool, but how do you handle security issues in versions? You options (as far as I can tell are):

Announce you can no longer use a particular version. (Breaks "we support every version".)
Change the behavior of the specific version and re-release with the same version number. (Breaks "this version has this particular behavior".)
Some kind of automatic translation that says "this published version maps to this internal version".

In any case, it's all very nice, but unlikely to impact how most people will design versioned artifacts in the future.

Predictions Scorecard, 2026 January 01

Tue Jan 06 2026 16:54:43 GMT-0500 (Eastern Standard Time)

📝 Predictions Scorecard, 2026 January 01 (Rodney Brooks). Every year, Rodney updates his predictions. Always a good read.

Money Matters: Dilemmas in Practical Monetary Halachah

Wed Dec 31 2025 22:13:55 GMT-0500 (Eastern Standard Time)

📖 Money Matters: Dilemmas in Practical Monetary Halachah by Rabbi Yosef Fleischman (2024; via Jeremy Wertheimer). The first of two volumes. A wide range of scenarios.

The Gene: An Intimate History

Wed Dec 31 2025 22:10:43 GMT-0500 (Eastern Standard Time)

📖 The Gene: An Intimate History by Siddhartha Mukherjee (2016; via Siraj Raval). The book makes many concepts in biology understandable. Combining the author's personal history makes it heart-warming.

The 3 a.m. Call That Changed The Way I Design APIs

Wed Dec 31 2025 22:05:08 GMT-0500 (Eastern Standard Time)

📝 The 3 a.m. Call That Changed The Way I Design APIs (Sreenivasa Reddy Hulebeedu Reddy / The New Stack). Hard-earned principles.

On deathbed advice/regret

Wed Dec 31 2025 21:56:51 GMT-0500 (Eastern Standard Time)

📝 On deathbed advice/regret (hazn; via Tyler Cowen). I agree with the main point of the post which is why I've usually taken deathbed regret and converted it into specific advice. For many years, I've had the following (lightly edited) list towards the top of my todo list:

Don't ignore your dreams
Don't work too much
Say what you think
Cultivate friendships
Be happy

Introducing Beads: A coding agent memory system

Wed Dec 31 2025 21:54:46 GMT-0500 (Eastern Standard Time)

📝 Introducing Beads: A coding agent memory system (Steve Yegge). The whole thing was vibecoded and is kinda crazy, but I've actually been looking for a way to track issues from within git. Apparently the agents really like it.

Six New Tips for Better Coding With Agents

Wed Dec 31 2025 21:51:40 GMT-0500 (Eastern Standard Time)

📝 Six New Tips for Better Coding With Agents (Steve Yegge). Programming by hand is artisanal. Programming by copy-pasting from a chatbot is obsolete. The future appears to be conducting an orchestra of bot swarms.

Childhood and Education #16: Letting Kids Be Kids

Wed Dec 31 2025 21:47:23 GMT-0500 (Eastern Standard Time)

📝 Childhood and Education #16: Letting Kids Be Kids (Zvi Mowshowitz). I grew increasingly angry while reading this. Zvi documents so many rage-inducing examples of bad rules around letting children do things on their own.

Programming principles for self taught front-end developers

Wed Dec 31 2025 21:44:53 GMT-0500 (Eastern Standard Time)

📝 Programming principles for self taught front-end developers (Kilian Valkhof / Piccalilli). This is high-level advice/heuristics with really good examples.

cosmofy 0.2.1

Wed Dec 24 2025 13:58:52 GMT-0500 (Eastern Standard Time)

cosmofy 0.2.1 is available. Lots of little bugs after this morning's big release.

If you have uv installed you can use cosmofy:

uvx cosmofy

# or install it
uv tool install cosmofy

Release Notes 0.2.1 - 2025-12-24T18:35:27Z #

Fixed

#42 minor typos; added links in README
#58 only do the Windows dance if literally trying to replace sys.executable
#96 files not saving to cache
#97, #98 executable permissions for Cosmopolitan Python
#99 nothing being logged in binary build
#95 don't search for pyproject.toml when you only have --script to bundle
#104 error when using cosmofy fs ls with --sort none
#105 restore working directory after error when using cosmofy fs add with --chdir
#103 use context manager for urlopen even for HEAD

Changed

#106 cosmofy fs cat outputs raw bytes to sys.stdout instead of decoding to UTF-8

Added

#94 pure Python limitation to README
#107 clearer docstring for expand_glob; literal paths are returned exactly
#100 document that pythonoid.run_python intentionally destroys sys.argv

Security

#102 call validate_url at the start of all the download_* functions
#109 reject paths with null bytes
#108 don't follow circular symlinks endlessly

cosmofy 0.2.0

Wed Dec 24 2025 00:43:10 GMT-0500 (Eastern Standard Time)

cosmofy 0.2.0 is available. So many things came together for this release:

Three open source developers I follow (William McGuan, Simon Willison, and Charlie Marsh) were all in a twitter thread where the concept of something like cosmofy was mentioned.
I learned how to make CLIs sort of work the way Astral's tools do.
I learned just how many crazy options GNU ls supports.

If you have uv installed you can use cosmofy:

uvx cosmofy

# or install it
uv tool install cosmofy

Release Notes 0.2.0 - 2025-12-24T03:14:02Z #

This release represents a very large shift from bundling individual python files to using uv to bundle entire venv directories. The behavior of the CLI is now much more similar to uv in form and function.

Fixed

#14 typo in --help
#43 exclude uv artifacts from the bundle
#46 expand_globs to follow all (most?) of the shell rules
#48 getting only the project's console_scripts
#50 cosmofy bundle --script assumes venv in output
#52 avoiding clobbering logging on import
#56 used context handlers to close temporary directories
#58 self update on Windows
#59 assumption that Last-Modified will be present in headers
#61 replaced PathDistribution._path with appropriate fallbacks
#62 replaced assert for validation with actual raise on error
#66 needless pass in parsing args
#67 command names in logging
#70 incorrect shebang in bundle.py
#71 used context handlers to close ZipFile objects
#72 download progress indicator shouldn't divide by zero
#73 missing newline in bundle.py error message
#75 explicit return of default when we can't get the version from a file
#77 cosmofy fs cat usage string
#79 ZipFile2.now should not be at module-level
#92 use public API for SourceFileLoader

Changed

#16 update examples, readme, usage (HT @Pugio)
#28 replaced --debug with --quiet + --verbose
#29 DEFAULT_PYTHON_URL is now Cosmopolitan Python
#31 replaced --cache with --no-cache + --cache-dir
#33 refactored bundler.py into subcommands
#55 switched to JSON parsing uv version output
#57 switched to using os.replace for atomic file moves
#64 replaced removeprefix with better cross-platform approach
#79 moved progress indicator to stderr
#87 marked cosmofy updater commands as experimental

Added

#18 progress when downloading (HT @Pugio)
#32 cosmofy bundle --script
#33 cosmofy bundle
#33 cosmofy fs add
#33 cosmofy fs args
#33 cosmofy updater remove
#33 dry run banner
#35 cosmofy fs ls
#36 cosmofy fs cat
#37 cosmofy fs rm
#38 cosmofy fs add -f to overwrite files that exist
#39 --color global option
#40 cosmofy self update
#40 cosmofy self version
#49 --exact to uv sync for better exact specs
#60 timeouts to network calls
#81 note that concurrent removals are not supported
#83 missing docstrings
#86 log message that only GitHub URLs are inferred
#89 warning when adding absolute paths
#91 note about handling weird file names

Removed

#15 cosmo.yaml workflow; added gh command
#17 default .com extension
#24 support for python <= 3.9
#27 --clone is no longer supported
#30 --download is no longer supported

Security

#53 added validate_url to avoid non-HTTPS URLs
#53 added security considerations note
#54 removed shell=True in places that don't need it
#63 added sanitize_zip_path to avoid bad entry names

cosmofy fs ls: emulating GNU ls

Wed Dec 24 2025 00:38:21 GMT-0500 (Eastern Standard Time)

Previously: pythonoid, baton

When I was designing the low-level zip file manipulation tools for cosmofy, I wanted an easy way to see the contents of the bundle. We're so used to using ls for looking into directories that I thought it would be cool to emulate as much of ls as I could.

But then it turned out that ls has a crazy number of options. I actually went through them all and tried to figure out if it was possible to support them.

But then I realized this was insane. First, many of the options are just aliases for slightly more explicit options. Charlie Marsh would never have a -t that was an alias for --sort=time. Why should I?

In the end I decided to go with the most common options (sorting, list view), a couple that were easy to implement, and a few longer-form ones that cover most of the aliases.

I'm pretty happy with the way it turned out.

baton: emulating Astral's CLI approach

Wed Dec 24 2025 00:17:17 GMT-0500 (Eastern Standard Time)

Previously: pythonoid

Imitation is the highest form of flattery which is why as part of the cosmofy 0.2.0 release, I decided to change everything about how the CLI behaved to make it work more like the way the tools from Astral work.

I have a long-term plan for Astral to take over making Cosmopolitan Python apps. It's a long shot, but if they do, it'll be a huge win for cross-platform compatible executables. I also saw this popular issue that there should be a uv bundle command that bundles everything up.

To make it easier to adopt, I decided to make the interface follow Astral's style in three important ways:

Subcommand structure: It's gotta be cosmofy bundle and cosmofy self update
Colored output: Gotta auto-detect that stuff. Luckily, I had fun with brush years ago, so I know about terminal color codes.
Global flags: Some of those flags gotta be global.
Smart ENV defaults: smart defaults + pulling from environment variables to override.

Now I didn't start out wanting to build my own argument parser (really, I promise I didn't!). I tried going the argparse route (I even tried my own attrbox / docopt solution), but I had a few constraints:

I really don't want 3rd party dependencies (even my own). cosmofy needs to stay tight and small.
I want argument parsing to go until it hits the subcommand and then delegate the rest of the args to the subcommand parser.
I want to pass global options from parent to child sub-parser as needed.

Together these pushed for a dedicated parser. This lets me write things like:

usage = f"""\
Print contents of a file within a Cosmopolitan bundle.

Usage: cosmofy fs cat <BUNDLE> <FILE>... [OPTIONS]

Arguments:
{common_args}
  <FILE>...                 one or more file patterns to show

  tip: Use `--` to separate options from filenames that start with `-`
  Example: cosmofy fs cat bundle.zip -- -weird-filename.txt

Options:
  -p, --prompt              prompt for a decryption password

{global_options}
"""


@dataclass
class Args(CommonArgs):
    __doc__ = usage
    file: list[str] = arg(list, positional=True, required=True)
    prompt: bool = arg(False, short="-p")

...

def run(args: Args) -> int:
  ...

cmd = Command("cosmofy.fs.cat", Args, run)
if __name__ == "__main__":
    sys.exit(cmd.main())

For the colored output, I took inspiration from William McGuan's rich which uses tag-like indicators to style text.

Mine is much worse and minimal, but it gets the job done for the bits of color I need.

Is Cognitive Dissonance Actually a Thing?

Wed Dec 24 2025 00:06:12 GMT-0500 (Eastern Standard Time)

📝 Is Cognitive Dissonance Actually a Thing? (Shayla Love / The New Yorker; via Tyler Cowen). I need to believe it's a thing and not a thing at the same time.

How I Found Myself Running a Microschool

Wed Dec 24 2025 00:03:36 GMT-0500 (Eastern Standard Time)

📝 How I Found Myself Running a Microschool (Kelsey Piper / Center for Educational Progress). Over the past 10 years I have migrated to essentially this view: you need direct instruction to get the basics and a foundation; you need to see people enact the values you want to transmit; and you need a strong motivating project to get you over the humps when the going gets tough.

Your job is to deliver code you have proven to work

Wed Dec 24 2025 00:01:50 GMT-0500 (Eastern Standard Time)

📝 Your job is to deliver code you have proven to work (Simon Willison). Strongly endorse.

Ideas Aren’t Getting Harder to Find

Tue Dec 23 2025 23:58:22 GMT-0500 (Eastern Standard Time)

📝 Ideas Aren’t Getting Harder to Find (Karthik Tadepalli / Asterisk). Knowing what is causing productivity growth to start to slow is critical to selecting appropriate policies for how to get it going again. Karthik makes a good case for why the idea that "ideas are getting harder to find" is wrong and why it's more of a failure of the market to weed out bad ideas and promote good ones.

AI has Educators Polarized

Tue Dec 23 2025 23:55:35 GMT-0500 (Eastern Standard Time)

📝 AI has Educators Polarized (Arnold Kling; via Zvi Mowshowitz). Nice framing: some think education is adversarial, others that it's cooperative.

Oliver Sacks Put Himself Into His Case Studies. What Was the Cost?

Tue Dec 23 2025 23:47:49 GMT-0500 (Eastern Standard Time)

📝 Oliver Sacks Put Himself Into His Case Studies. What Was the Cost? (Rachel Aviv / The New Yorker). Seems like a high cost. The whole field has taken such a pummeling since 2017, that it makes me think of The Simpsons meme.

Justified

Tue Dec 23 2025 23:43:39 GMT-0500 (Eastern Standard Time)

📝 Justified (Bite code!). just is a Rust-based task runner and here's a good write up about its features. If I wasn't writing ds (my own task runner), I'd probably use just. I'll probably experimentally support some subset of just's syntax in ds.

Book Time #26: How I Read

Tue Dec 23 2025 23:26:33 GMT-0500 (Eastern Standard Time)

📝 Book Time #26: How I Read (Aaron Gordon / Book Time). My process for selecting books is very different from Aaron's, but it's nice to see someone else's process.

Mine:

I get recommendations from friends, colleagues, podcasts, and other books.
I add them to a list (currently an Amazon Wishlist). I prefer digital books over physical books because I can make them into audiobook (at 4x) and they take up very little physical space.
When I get to the end of the books I have in my queue, I sort the wishlist by price (lowest to highest) and buy everything under $5. That keeps me busy for about 3 months.
I take all the books that I've purchased and sort them by the number of pages (lowest to highest). This is my new reading queue.

There are only a couple of exceptions for me to deviate from this order:

My wife recommends I read something next, especially as it may impact our kids. This bumps the book to the top of the list (both for purchase and reading).
I'm about to meet someone who recently published a book. There have been a few times where I got through their most-recently published book the day before meeting them. HT: Ramit Sethi who teaches the importance of being overly prepared for certain kinds of meetings.

I often get asked how much I retain from the audio versions of books. My informal tests show no more or less than most people when they read books. I don't remember every detail, but neither do most people. HT: Mike Masnick who introduced me to the idea of listening to podcasts at greater than 2x.

The Writer's Journey: Mythic Structure for Writers, 3rd Edition

Sun Dec 07 2025 13:39:08 GMT-0500 (Eastern Standard Time)

📖 The Writer's Journey: Mythic Structure for Writers, 3rd Edition by Christopher Vogler (2007; via Chanan Berkovits). You may have heard of Joseph Campbell's The Hero with a Thousand Faces, but how about something with more contemporary references? Also, does the obligatory meta thing and shows how the process of writing itself follows the mythic structure.

On the Consumption of AI-Generated Content at Scale

Sun Dec 07 2025 13:36:26 GMT-0500 (Eastern Standard Time)

📝 On the Consumption of AI-Generated Content at Scale (Shreya Shankar). Thoughtful take.

TIL: importlib.metadata.version

Thu Dec 04 2025 19:29:10 GMT-0500 (Eastern Standard Time)

While I was updating my standard pyproject.toml to use dependency groups, I switched to using uv as my build backend (from setuptools). Like all the other design choices, it has very sensible defaults with good overrides when you need them.

One thing that it doesn't support is dynamic versions (i.e. reading __version__ from __init__.py). Charlie Marsh explains:

Using dynamic metadata for things that are actually just static lookups feels like the wrong tradeoff.

Like version = ["dynamic"] to read the version from __init__.py.

As soon as you have dynamic metadata, you need to install dependencies and run Python just to get that info.

Static metadata rules. Just write the version out twice!

This makes sense, but I really didn't want to write out the version twice. Luckily, uv has a nice uv version command to change/bump the version number in the pyproject.toml, so I started thinking about how I'd also change the version in the __init__.py at the same time.

I asked ChatGPT what the best approach was and it suggested something quite different: don't write the value in __init__.py at all.

from importlib.metadata import version, PackageNotFoundError
try:
    __version__ = version("package-name") # from pyproject.toml: project.name
except PackageNotFoundError:
    __version__ = "0.0.0" # aka unknown version

I have since noticed that this was also the suggestion Adam Johnson had in the same thread.

One minor concern I have is whether this works in a cosmofy build, but that will be a separate exploration.

Dependency Groups to the Rescue

Thu Dec 04 2025 14:00:54 GMT-0500 (Eastern Standard Time)

Previously: Stop Hiding Python Dev Dependencies

I'm a bit late to the party, but even when I saw that PEP 735 – Dependency Groups in pyproject.toml had been accepted and standardized by PyPA, it still didn't register how this should impact my pyproject.toml configurations.

In my previous post, I argued that absent a standard place to put dev dependencies in pyproject.toml, we should opt to use optional-dependencies. However, dependency groups seem to offer a nice standard place for such dependencies: these are bundles of dependencies that don't get built into the final distribution (i.e. they are not required to run the package).

Feels like the best of both worlds and uv already started using dependency-groups.dev as the place that uv add --dev writes to and this group is sync'd by default when using uv sync and uv run.

I only really got this when I read Simon Willison's post about how he's using dependency groups to make it easier for people to hack on his code.

Remix: Making Art and Commerce Thrive in the Hybrid Economy

Sun Nov 30 2025 19:51:23 GMT-0500 (Eastern Standard Time)

📖 Remix: Making Art and Commerce Thrive in the Hybrid Economy by Lawrence Lessig (2002; via Tribe of Mentors & Joseph Gordon Levitt). I knew most of these ideas from having read TechDirt for so many years and listening to Lessig talk about these ideas for several decades. Still, I was surprised how much of the fights we have today still stem from a basic disagreement about copyright and how art gets made.

Topgrading: The Proven Hiring and Promoting Method that Turbocharges Company Performance, Third Edition

Sun Nov 30 2025 19:48:13 GMT-0500 (Eastern Standard Time)

📖 Topgrading: The Proven Hiring and Promoting Method that Turbocharges Company Performance, Third Edition by Bradford D. Smart (2012; via Tim Ferriss). More than 20 years later, many parts of this still make sense even though core parts would require rethinking in an LLM world.

Dare To Author!: Take Charge of the Narrative of Your Life

Sun Nov 30 2025 19:43:11 GMT-0500 (Eastern Standard Time)

📖 Dare To Author!: Take Charge of the Narrative of Your Life by Lior Arussy (2024; via Elizabeth Kratz). A nice structured approach to updating how you think about your life story.

Don’t Build an Audience

Sun Nov 30 2025 19:35:41 GMT-0500 (Eastern Standard Time)

📝 Don’t Build an Audience (Human Invariant). It's the bell curve meme: create cool stuff - I need to build an audience - create cool stuff.

Closed: Bad error messages from blog post

Sun Nov 30 2025 18:32:00 GMT-0500 (Eastern Standard Time)

📝 Closed: Bad error messages from blog post (Facebook). Turns out pyrefly developers saw my previous post, broke things down into many sub-issues and fixed them all. pyrefly now reports zero errors in castfit and the errors it find in other repos have pretty understandable error messages!

Disrupting the first reported AI-orchestrated cyber espionage campaign

Sun Nov 30 2025 18:30:03 GMT-0500 (Eastern Standard Time)

📝 Disrupting the first reported AI-orchestrated cyber espionage campaign (Anthropic). Expect more of this in the future.

castfit 0.1.3

Wed Nov 26 2025 22:12:11 GMT-0500 (Eastern Standard Time)

castfit 0.1.3 is available. The biggest thing I did was implement is_subtype which is a forced me to learn about invariant, covariant, and contravariant types.

To install castfit:

# modern (recommended)
uv add castfit

# classic
python -m pip install castfit

Release Notes 0.1.3 - 2025-11-27T03:12:11Z #

Changed

#33 PyPI publishing to support newer uv build metadata
#40 PyPI publishing to use uv publish instead of pypa/gh-action-pypi-publish

Added

#23 ability to type check Callable via is_subtype
#34 support for property fields on classes
#35 __name__ to TypeInfo if present
#36 support for default values in dataclasses.Field
#37 support for function definitions in classes

Removed

#40 support for python 3.9

Security

#40 We are trying out uv publish instead of pypa/gh-action-pypi-publish.

Power: Why Some People Have It—and Others Don't

Tue Nov 11 2025 16:38:53 GMT-0500 (Eastern Standard Time)

📖 Power: Why Some People Have It—and Others Don't by Jeffrey Pfeffer (2010; via Robin Hanson). A blend of anecdotes and social science experiments that didn't replicate. The networking bits seemed fine. Everything else was kinda meh.

Remote: Office Not Required

Tue Nov 11 2025 16:23:24 GMT-0500 (Eastern Standard Time)

📖 Remote: Office Not Required by Jason Fried & David Heinemeier Hansson (2013; via Iridescent Learning). It's interesting to see how work from home policies have ebbed and flowed since 2013.

Why 98.5% of organizations have slow actions/checkout

Tue Nov 11 2025 15:58:50 GMT-0500 (Eastern Standard Time)

📝 Why 98.5% of organizations have slow actions/checkout (Chris Goller / depot; via Changelog News). I'm hesitant to fiddle with the git internals, but the sparse checkout and partial clone sound like easy things to try. (Most of GitHub Actions were sped up by using uv.)

Deploying to Amazon's cloud is a pain in the AWS younger devs won't tolerate

Tue Nov 11 2025 15:54:27 GMT-0500 (Eastern Standard Time)

📝 Deploying to Amazon's cloud is a pain in the AWS younger devs won't tolerate (Corey Quinn / The Register; via Changelog News). Hilarious take, but so true. It took me 7 years to be able to close a handful of student accounts (who had inactive emails after graduation) that were still charging my billing account, but whose EC2 instances were not visible to me because of AWS permissions issues.

Game design is simple, actually

Tue Nov 11 2025 15:50:33 GMT-0500 (Eastern Standard Time)

📝 Game design is simple, actually (Raph Koster). Clickbait title aside, there are twelve compact and useful design lessons each of which is a whole world unto itself.

The Algorithmic Turn: The Emerging Evidence On AI Tutoring That's Hard to Ignore

Tue Nov 11 2025 15:46:25 GMT-0500 (Eastern Standard Time)

📝 The Algorithmic Turn: The Emerging Evidence On AI Tutoring That's Hard to Ignore (Carl Hendrick / The Learning Dispatch). Long post covers many questions about the material properties of learning and teaching. Also covers both the disappointment of various EdTech approaches and the possibility of surpassing them.

Making XML human-readable without XSLT

Tue Nov 11 2025 15:40:32 GMT-0500 (Eastern Standard Time)

📝 Making XML human-readable without XSLT (Jake Archibald; via Simon Willison). I have always found XSLT to be both impressive and horrifying. Makes sense that it's a security nightmare for browsers.

The Case Against pgvector

Tue Nov 11 2025 15:34:46 GMT-0500 (Eastern Standard Time)

📝 The Case Against pgvector (Alex Jacobs; via Simon Willison). I actually got questions about picking an index for pgvector; the options aren't great and this article explains why.

Claude Agent Skills: A First Principles Deep Dive

Tue Nov 11 2025 15:30:57 GMT-0500 (Eastern Standard Time)

📝 Claude Agent Skills: A First Principles Deep Dive (Han Lee). Very long, but informative break down of how Claude's Agent Skills works and how its different from function calling. I was particularly impressed by how the UI has to adapt between information being fed back to the LLM and information for the user.

SQLite concurrency and why you should care about it

Tue Nov 11 2025 15:08:17 GMT-0500 (Eastern Standard Time)

📝 SQLite concurrency and why you should care about it (Jean-Pierre Bachmann / Jellyfin). In all the times I've gotten a SQLite locking issue, the solution has always been to retry a few times (until some other process releases the lock). But this article suggests a few other ways to solve the problem.

The Wise Man's Fear

Sat Nov 01 2025 21:15:11 GMT-0400 (Eastern Daylight Time)

📖 The Wise Man's Fear by Patrick Rothfuss (2011; via Shalev NessAiver). Sequel to The Name of the Wind. Patrick Rothfuss joins the ranks of several other fantasy authors in having an unfinished series. I ended up reading the rants of his fans and editors as well as counter points that authors don't owe readers books. This book was a fine sequel; not as great and a bit graphic at times.

The Name of the Wind

Sat Nov 01 2025 21:11:53 GMT-0400 (Eastern Daylight Time)

📖 The Name of the Wind by Patrick Rothfuss (2007; via Shalev NessAiver). It's got stories-within-stories, sympathetic magic-as-physics, and a prodigy for a main character. What more could you want?

The 4-Hour Workweek: Escape 9–5, Live Anywhere, and Join the New Rich

Sat Nov 01 2025 21:07:03 GMT-0400 (Eastern Daylight Time)

📖 The 4-Hour Workweek: Escape 9–5, Live Anywhere, and Join the New Rich by Tim Ferriss (2009). Took me 16 years to finally read the book that bootstrapped Tim's takeoff. Turns out all of his books are in the same action-packed-encyclopedia style.

Side project advice

Sat Nov 01 2025 21:01:22 GMT-0400 (Eastern Daylight Time)

📝 Side project advice (Ned Batchelder). This reads like short & simple advice, but it is clearly hard-won and I should re-read this several times a year.

Three tough truths about climate

Sat Nov 01 2025 20:59:04 GMT-0400 (Eastern Daylight Time)

📝 Three tough truths about climate (Bill Gates). One of the more grounded perspectives on cost-benefit analysis as applied to climate change.

AI isn't replacing radiologists

Sat Nov 01 2025 20:55:46 GMT-0400 (Eastern Daylight Time)

📝 AI isn't replacing radiologists (Deena Mousa / Works in Progress; via Benj Azose). How long do we think the islands of automation will last?

You already have a git server

Sat Nov 01 2025 20:48:23 GMT-0400 (Eastern Daylight Time)

📝 You already have a git server (maurycyz; via Changelog News). I'm so tempted to do this, but it feels wrong.

Battling Infectious Diseases in the 20th Century: The Impact of Vaccines

Sat Nov 01 2025 20:44:09 GMT-0400 (Eastern Daylight Time)

📝 Battling Infectious Diseases in the 20th Century: The Impact of Vaccines (Tynan DeBold & Dov Friedman / WSJ). Some of my favorite graphics depicting the impact of vaccines on mortality after being introduced. These are about a decade old and, unfortunately, vaccine hesitancy / avoidance is going to make mortality go up again.

Will AI displace humans in the economy and culture?

Sat Nov 01 2025 20:28:36 GMT-0400 (Eastern Daylight Time)

📝 Will AI displace humans in the economy and culture? (Andrew Mayne). A thoughtful take, although I don't know how to rate the predictions.

Simon Maechling on microplastics

Sat Nov 01 2025 20:25:23 GMT-0400 (Eastern Daylight Time)

🐦 Simon Maechling on microplastics (Simon Maechling).

You’ve been told microplastics are everywhere.

In your food. In your blood. Even in your brain.

You should panic. You should rage.

But what if most of it was junk science?

Worth reading the whole thread.

You Aren’t In The DSM

Sat Nov 01 2025 20:22:35 GMT-0400 (Eastern Daylight Time)

📝 You Aren’t In The DSM (Awais Aftab / Asterisk). From record-keeping to people's self-concept.

Memorization, Trivia and Atomic Units for Creativity

Sat Nov 01 2025 20:12:08 GMT-0400 (Eastern Daylight Time)

📝 Memorization, Trivia and Atomic Units for Creativity (Edward Nevraumont). Edwards is the second-place winner of the Astral Codex Ten Non-Book Review Contest for his review of Alpha School.

Late Night is Dying Because of the Format Not The Hosts

Sat Nov 01 2025 20:08:02 GMT-0400 (Eastern Daylight Time)

📝 Late Night is Dying Because of the Format Not The Hosts (thinkinganddata). Goes back to vaudeville to describe changes in entertainment formats.

I Am Out Of Data Hell

Fri Oct 31 2025 16:51:33 GMT-0400 (Eastern Daylight Time)

📝 I Am Out Of Data Hell (Nikhil Suresh; via Josh Vogel). Agency does a body good.

The Continual Learning Problem

Fri Oct 31 2025 16:50:03 GMT-0400 (Eastern Daylight Time)

📝 The Continual Learning Problem (Jessy Lin). Interesting take.

What Went Wrong with Math Instruction in New York?

Fri Oct 31 2025 16:47:48 GMT-0400 (Eastern Daylight Time)

📝 What Went Wrong with Math Instruction in New York? (Benjamin Solomon / Center for Educational Progress). Quite a bit it would seem.

Rearchitecting GitHub Pages

Fri Oct 31 2025 16:44:53 GMT-0400 (Eastern Daylight Time)

📝 Rearchitecting GitHub Pages (Hailey Somerville / GitHub Blog; via Sam Lambert). Sam's tweet says it all:

Until ~2015, GitHub Pages hosted over 2 million websites on 2 servers with a multi-million-line nginx.conf, edited and reloaded per deploy. This worked incredibly well, with github.io ranking as the 140th most visited domain on the web at the time.

How the World’s Most Famous Code Was Cracked

Fri Oct 31 2025 16:42:52 GMT-0400 (Eastern Daylight Time)

📝 How the World’s Most Famous Code Was Cracked (Jack Murtagh & Jeanna Bryner / Scientific American). Reminds me of the relevant XKCD.

When is it better to think without words?

Fri Oct 31 2025 16:40:35 GMT-0400 (Eastern Daylight Time)

📝 When is it better to think without words? (Henrik Karlsson / Escaping Flatland). I think I only do this rarely.

I Figured Out How to Engineer Emergence

Fri Oct 31 2025 16:38:38 GMT-0400 (Eastern Daylight Time)

📝 I Figured Out How to Engineer Emergence (Erik Hoel / The Intrinsic Perspective). This is a follow-up to Erik's concept of how high-level features emerge from lower-level descriptions.

Languaging: The Software of Belief

Fri Oct 31 2025 16:33:36 GMT-0400 (Eastern Daylight Time)

📝 Languaging: The Software of Belief (Mike Maples / Pattern Breakers; via Kevin Munnelly). Pairs well with Peter Thiel's Zero-to-One concept.

Unseeable prompt injections in screenshots

Fri Oct 31 2025 16:26:24 GMT-0400 (Eastern Daylight Time)

📝 Unseeable prompt injections in screenshots (Shivan Kaul Sahib & Artem Chaikin / Brave; via Simon Willison). Solving AI browser security feels so much harder than solving secure credit card processing.

Abundance: The Future Is Better Than You Think

Sun Oct 19 2025 16:00:06 GMT-0400 (Eastern Daylight Time)

📖 Abundance: The Future Is Better Than You Think by Peter H. Diamandis & Steven Kotler (2012). Lots of predictions, some of which may already have come true.

Cues versus Content: Hyrum Lewis Replies to Marlen-Starr's Proposal

Sun Oct 19 2025 15:56:28 GMT-0400 (Eastern Daylight Time)

📝 Cues versus Content: Hyrum Lewis Replies to Marlen-Starr's Proposal (Bryan Caplan). This is part of a conversation on how to understand the nature of ideologies.

Play Nice But Win: A CEO's Journey from Founder to Leader

Thu Oct 16 2025 18:15:53 GMT-0400 (Eastern Daylight Time)

📖 Play Nice But Win: A CEO's Journey from Founder to Leader by Michael Dell & James Kaplan (2021). I remember being totally confused by the EMC + Dell story when it was happening. Lots of oh-so-that's-what-that was behind-the-scenes stories.

Nine Lives: My time as the MI6's top spy inside al-Qaeda

Thu Oct 16 2025 18:10:15 GMT-0400 (Eastern Daylight Time)

📖 Nine Lives: My time as the MI6's top spy inside al-Qaeda by Aimen Dean, Paul Cruickshank, and Tim Lister (2018). Wild story.

Days of Deliverance: Essays on Purim and Hanukkah

Thu Oct 16 2025 18:07:42 GMT-0400 (Eastern Daylight Time)

📖 Days of Deliverance: Essays on Purim and Hanukkah by Rabbi Joseph B. Soloveitchik (2007).

A window into modern loan origination

Thu Oct 16 2025 18:03:53 GMT-0400 (Eastern Daylight Time)

📝 A window into modern loan origination (Patrick McKenzie / Bits about Money). How the sausage gets made.

The VR Hype Cycle: Lessons for the Age of AI

Thu Oct 16 2025 18:01:22 GMT-0400 (Eastern Daylight Time)

📝 The VR Hype Cycle: Lessons for the Age of AI (Caleb Sponheim / NN/g).

Hype cycles occur when expectations, narratives, and messaging outpace technological capability.

How well can large language models predict the future?

Thu Oct 16 2025 17:58:48 GMT-0400 (Eastern Daylight Time)

📝 How well can large language models predict the future? (Houtan Bastani, Simas Kučinskas, and Ezra Karger / Forecasting Research Institute). A continuous approach to test forecasting models. They do mention a potential hiccup:

When provided market forecasts in their prompts, multiple LLMs—including GPT-4.5—simply copy them.

The Captain Class: A New Theory of Leadership

Sun Oct 05 2025 19:46:44 GMT-0400 (Eastern Daylight Time)

📖 The Captain Class: A New Theory of Leadership by Sam Walker (2017; via Ryan Holiday). A book about leadership in various sports dressed up as a business book. The main thing it seems to promote is "unusual" leadership which would seem to account for all the contradictory advice.

21 Facts About Throwing Good Parties

Sun Oct 05 2025 19:42:27 GMT-0400 (Eastern Daylight Time)

📝 21 Facts About Throwing Good Parties (Uri Bram / Atoms vs Bits; via Tyler Cowen). These rules apply broadly to many other kinds of interactions too.

Hacker Laws

Sun Oct 05 2025 19:39:57 GMT-0400 (Eastern Daylight Time)

📝 Hacker Laws (Dave Kerr). Nice collection of rules-of-thumb.

Karpathy on Sutton

Sun Oct 05 2025 19:36:16 GMT-0400 (Eastern Daylight Time)

🐦 Karpathy on Sutton (Andrej Karpathy). This helped me understand the debate that Sutton and Dwarkesh were having when it seemed like they just used words very differently.

Scam GPT: GenAI and the Automation of Fraud

Sun Oct 05 2025 16:45:23 GMT-0400 (Eastern Daylight Time)

📝 Scam GPT: GenAI and the Automation of Fraud (Lana Swartz, Alice E. Marwick, & Kate Larson / Data & Society; via Bruce Shneier). Make sure you have a talk with your parents about how these things work.

Fate and Destiny: From Holocaust to the State of Israel

Wed Oct 01 2025 00:34:08 GMT-0400 (Eastern Daylight Time)

📖 Fate and Destiny: From Holocaust to the State of Israel by Rabbi Joseph B. Soloveitchik (2000). A reprint of Kol Dodi Dofek. Change some of the senator names plus a few dates and this could have been written yesterday.

Why Not Me?

Wed Oct 01 2025 00:30:43 GMT-0400 (Eastern Daylight Time)

📖 Why Not Me? by Mindy Kaling (2015; via Seth Rogan). Where in the world did I see a recommendation from Seth Rogan and what possessed me to add this book to my queue? No clue. Sometimes I just end up reading random things. This was pretty random.

Someday This Pain Will Be Useful to You: A Novel

Wed Oct 01 2025 00:28:48 GMT-0400 (Eastern Daylight Time)

📖 Someday This Pain Will Be Useful to You: A Novel by Peter Cameron (2025; via Tim Ferriss & Gretchen Ruben). No clue what I expected.

Scale: The Universal Laws of Life, Growth, and Death in Organisms, Cities, and Companies

Wed Oct 01 2025 00:26:37 GMT-0400 (Eastern Daylight Time)

📖 Scale: The Universal Laws of Life, Growth, and Death in Organisms, Cities, and Companies by Geoffrey West (2017; via Robin Hanson). A look at constraints in different systems.

Tap Dancing to Work: Warren Buffett on Practically Everything, 1966-2013

Wed Oct 01 2025 00:24:35 GMT-0400 (Eastern Daylight Time)

📖 Tap Dancing to Work: Warren Buffett on Practically Everything, 1966-2013 by Carol J. Loomis (2012; via Bill Gates). If you've read other books about Buffett, not much new here.

Like, Comment, Subscribe: Inside YouTube's Chaotic Rise to World Domination

Wed Oct 01 2025 00:13:02 GMT-0400 (Eastern Daylight Time)

📖 Like, Comment, Subscribe: Inside YouTube's Chaotic Rise to World Domination by Mark Bergen (2022). Like many books in this genre, you get a sense of how quickly people's id/ego/superego merged with a quickly growing internet startup.

The Predictioneer's Game: Using the Logic of Brazen Self-Interest to See and Shape the Future

Wed Oct 01 2025 00:03:09 GMT-0400 (Eastern Daylight Time)

📖 The Predictioneer's Game: Using the Logic of Brazen Self-Interest to See and Shape the Future by Bruce Bueno De Mesquita (2009). Reading this book in 2025 is like finding an ancient prophecy that everyone has forgotten. Not perfect, but so many good calls. It also contains this gem:

My idea is that the Israeli and Palestinian governments will distribute a portion of their tax revenue generated from tourism (and only from tourism) to each other. Before going into the details—where the devil resides—let’s first see why tourist revenue and not any other. Why not, for example, promote peace by setting up joint Israeli-Palestinian ventures, or allowing freer movement between regions, or some other scheme? As we will see, shared tourist revenue provides a nearly unique opportunity.

[...]

The tourist tax revenue arrangement need not last forever. It must include an irrevocable commitment for it to persist for a long time (say twenty years), and it is important that this tax revenue sharing be tied to a fixed formula based on the current populations and not on future changes in those populations. Opening the formula to renegotiation could create perverse incentives. It is also essential that the definition of tax revenue originating from spending by tourists be based on predetermined rules for estimating this source of income. Independent accounting firms might be used to provide a standard way to define and identify tourist revenue and the taxes derived from it. This tax revenue would then be allocated to each side over the agreed duration of the program based on a onetime fixed population-based formula with no questions asked.

Using Game Theory to play through the whole scenario was a fascinating take.

How to Win in a Winner-Take-All World: The Definitive Guide to Adapting and Succeeding in High-Performance Careers

Tue Sep 30 2025 23:56:39 GMT-0400 (Eastern Daylight Time)

📖 How to Win in a Winner-Take-All World: The Definitive Guide to Adapting and Succeeding in High-Performance Careers by Neil Irwin (2025; via Tyler Cowen). The advice isn't bad, per se. But I definitely thought this was written in the 90s and not June 2025. Tyler Cowen called it "both subtler and broader than the title alone would indicate," but I guess I missed the deep insight.

Old Man's War

Tue Sep 30 2025 23:53:37 GMT-0400 (Eastern Daylight Time)

📖 Old Man's War by John Scalzi (2007; via Jeffrey Klein). Fun read.

The Courage to Be Disliked: The Japanese Phenomenon That Shows You How to Change Your Life and Achieve Real Happiness

Tue Sep 30 2025 23:45:54 GMT-0400 (Eastern Daylight Time)

📖 The Courage to Be Disliked: The Japanese Phenomenon That Shows You How to Change Your Life and Achieve Real Happiness by Ichiro Kishimi & Fumitake Koga (2018; via Patrick Gleason). I never heard of Alfred Adler or his school of psychology (he was a contemporary of Freud's). There are lots of commonplace examples which makes this an interesting read, but I suspect some of the claims are a bit too strong (e.g., "trauma" not being a real thing).

The Sweaty Startup: How to Get Rich Doing Boring Things

Tue Sep 30 2025 23:42:21 GMT-0400 (Eastern Daylight Time)

📖 The Sweaty Startup: How to Get Rich Doing Boring Things by Nick Huber (2025). Small things make a pile.

Legal Systems Very Different from Ours

Tue Sep 30 2025 23:38:58 GMT-0400 (Eastern Daylight Time)

📖 Legal Systems Very Different from Ours by David Friedman, Peter Leeson, & David Skarbek (2019). I re-read this to remember the general kinds of problems human legal systems solve.

The No S Diet: The Strikingly Simple Weight-Loss Strategy That Has Dieters Raving--and DroppingPounds

Tue Sep 30 2025 23:35:39 GMT-0400 (Eastern Daylight Time)

📖 The No S Diet: The Strikingly Simple Weight-Loss Strategy That Has Dieters Raving--and DroppingPounds by Reinhard Engels & Ben Kallen (2008). The rules are literally on the cover and the rest of the book is commentary. Being a diet book from 2008 it's got some references to psych studies that didn't replicate. But from a UX perspective, brilliant.

Transform: Mayo Clinic Platform and the Digital Future of Health

Tue Sep 30 2025 23:32:36 GMT-0400 (Eastern Daylight Time)

📖 Transform: Mayo Clinic Platform and the Digital Future of Health by Paul Cerrato M.A. & John D. Halamka M.D. M.S. (2025; via Jeremy Wertheimer). The bits about NLP felt ancient, but other bits were modern enough.

Good To Great And The Social Sectors: A Monograph to Accompany Good to Great

Tue Sep 30 2025 23:29:46 GMT-0400 (Eastern Daylight Time)

📖 Good To Great And The Social Sectors: A Monograph to Accompany Good to Great by Jim Collins (2011). Short and to the point.

You're a Slow Thinker. Now what?

Tue Sep 30 2025 23:22:08 GMT-0400 (Eastern Daylight Time)

📝 You're a Slow Thinker. Now what? (Afiq / Casual Physics). I wish this was studied more.

Thank you for being annoying

Tue Sep 30 2025 23:20:03 GMT-0400 (Eastern Daylight Time)

📝 Thank you for being annoying (Adam Mastroianni / Experimental History). Many things are the result of being irritated.

Brainstorming terrible ideas in a group

Tue Sep 30 2025 23:18:07 GMT-0400 (Eastern Daylight Time)

📝 Brainstorming terrible ideas in a group (Ruslan Osipov). Brainstorming doesn't usually get people moving around in the search space enough; coming up with terrible ideas might actually help.

Why Today’s Humanoids Won’t Learn Dexterity

Tue Sep 30 2025 23:15:16 GMT-0400 (Eastern Daylight Time)

📝 Why Today’s Humanoids Won’t Learn Dexterity (Rodney Brooks). While it's hard for me to judge on the main argument, I can totally see the words changing:

Following that pattern, what it means to be a humanoid robot will change over time.

Before too long (and we already start to see this) humanoid robots will get wheels for feet, at first two, and later maybe more, with nothing that any longer really resembles human legs in gross form. But they will still be called humanoid robots.

Getting More Strategic

Tue Sep 30 2025 23:11:52 GMT-0400 (Eastern Daylight Time)

📝 Getting More Strategic (Cate Huston / Accidentally in Code).

We often talk about strategy like it’s defining the end state, setting and describing the destination. But strategy is about defining the incremental steps – the proximate objectives – that can take us towards that end state. Strategy is understanding where we are at – context – and the path from there to where we need to go. Any strategic “plan”, is best executed as a set of proximate objectives.

Illiteracy is a policy choice

Tue Sep 30 2025 23:09:49 GMT-0400 (Eastern Daylight Time)

📝 Illiteracy is a policy choice (Kelsey Piper / The Argument). Some things actually work.

Your UX Design EDC (Every Day Carry)

Tue Sep 30 2025 23:07:31 GMT-0400 (Eastern Daylight Time)

📝 Your UX Design EDC (Every Day Carry) (Jon Daiello). I particularly like the "Evidence Arsenal" because that has, historically, been the hardest bit for me with clients.

How to Motivate Yourself To Do A Thing You Don't Want to Do

Tue Sep 30 2025 23:06:04 GMT-0400 (Eastern Daylight Time)

📝 How to Motivate Yourself To Do A Thing You Don't Want to Do (Ashley Janssen). A reasonable framework.

AI Psychosis Is Rarely Psychosis at All

Tue Sep 30 2025 22:58:40 GMT-0400 (Eastern Daylight Time)

📝 AI Psychosis Is Rarely Psychosis at All (Robert Hart / Wired). "It's a delusion, Michael!"

I Vibecoded a Dispute Resolution App

Tue Sep 30 2025 22:56:16 GMT-0400 (Eastern Daylight Time)

📝 I Vibecoded a Dispute Resolution App (Sarah Constantin / Rough Diamonds). There should be 10 versions of this. Too bad people got so upset about the idea of such an app.

Is it Time to Un-Sass?

Tue Sep 30 2025 22:54:13 GMT-0400 (Eastern Daylight Time)

📝 Is it Time to Un-Sass? (Jeff Bridgforth / CSS-Tricks). Aside from the snub at Less.js, I largely agree.

Minimal Version Selection Revisited

Tue Sep 30 2025 22:51:07 GMT-0400 (Eastern Daylight Time)

📝 Minimal Version Selection Revisited (Alex Kladov).

This is quite neat — it puts a natural damper on the supply chain attacks. If a bad version of a library is released, someone needs to explicitly opt into this new version. What’s more, the deeper in your dependency tree the library is, the more explicit approvals are required for the library to propagate to your project.

You no longer need JavaScript

Tue Sep 30 2025 22:48:47 GMT-0400 (Eastern Daylight Time)

📝 You no longer need JavaScript (Lyra Rebane). Huh. CSS does a lot nowadays.

Is Jonathan Haidt right about smartphones?

Tue Sep 30 2025 22:46:12 GMT-0400 (Eastern Daylight Time)

📝 Is Jonathan Haidt right about smartphones? (Jon Severs / tes). I can agree with Haidt's policy choices but disagree that the evidence is unequivocal.

The Islamic Argument for Competence

Tue Sep 30 2025 22:41:49 GMT-0400 (Eastern Daylight Time)

📝 The Islamic Argument for Competence (Abdullah Yousef; via Tyler Cowen). A familiar tale.

Magical Thinking on AI

Tue Sep 30 2025 22:39:08 GMT-0400 (Eastern Daylight Time)

📝 Magical Thinking on AI (Melanie Mitchell). There are arguments to be made on both sides of the debate, but these ain't them.

'Make invalid states unrepresentable' considered harmful

Tue Sep 30 2025 22:37:09 GMT-0400 (Eastern Daylight Time)

📝 'Make invalid states unrepresentable' considered harmful (Sean Goedecke). Because you need a certain amount of flexibility when writing software.

Artificial Writing and Automated Detection

Tue Sep 30 2025 22:35:04 GMT-0400 (Eastern Daylight Time)

🐦 Artificial Writing and Automated Detection (Brian Jabarian). Nice approach.

Seeing like a software company

Tue Sep 30 2025 22:31:58 GMT-0400 (Eastern Daylight Time)

📝 Seeing like a software company (Sean Goedecke). On the importance of legibility in corporate software development.

What Is Man, That Thou Art Mindful Of Him?

Tue Sep 30 2025 22:27:29 GMT-0400 (Eastern Daylight Time)

📝 What Is Man, That Thou Art Mindful Of Him? (Scott Alexander / ACX). Funny take on LLM evaluations when looking at human intelligence.

Torah United: Teachings on the Weekly Parashah from Rav Avraham Yitzchak HaKohen Kook, Rabbi Joseph B. Soloveitchik, and the Chassidic Masters

Sun Aug 31 2025 21:50:44 GMT-0400 (Eastern Daylight Time)

📖 Torah United: Teachings on the Weekly Parashah from Rav Avraham Yitzchak HaKohen Kook, Rabbi Joseph B. Soloveitchik, and the Chassidic Masters by Aaron Goldscheider (2023; via Rachel & Jacob Novikov).

Demon Copperhead

Sun Aug 31 2025 21:47:52 GMT-0400 (Eastern Daylight Time)

📖 Demon Copperhead by Barbara Kingsolver (2022; via Jenny Melcher). If David Copperfield happened during the modern opioid epidemic.

Fossil Future: Why Global Human Flourishing Requires More Oil, Coal, and Natural Gas--Not Less

Sun Aug 31 2025 21:45:06 GMT-0400 (Eastern Daylight Time)

📖 Fossil Future: Why Global Human Flourishing Requires More Oil, Coal, and Natural Gas--Not Less by Alex Epstein (2022; via Bryan Caplan). The title and subtitle are intentionally provocative, but not technically wrong once you think about how many things are made of, e.g., plastic. However, glad to see nuclear get its own chapter.

Black Flags: The Rise of ISIS

Sun Aug 31 2025 21:43:18 GMT-0400 (Eastern Daylight Time)

📖 Black Flags: The Rise of ISIS by Joby Warrick (2015; via Rob Reid). A bit outdated now, but still a wild ride.

Number Go Up: Inside Crypto's Wild Rise and Staggering Fall

Sun Aug 31 2025 21:41:21 GMT-0400 (Eastern Daylight Time)

📖 Number Go Up: Inside Crypto's Wild Rise and Staggering Fall by Zeke Faux (2023; via Jeremy Wertheimer).

Peak Mind: Achieve Peak Mindfulness and Unlock Your Full Potential

Sun Aug 31 2025 21:39:45 GMT-0400 (Eastern Daylight Time)

📖 Peak Mind: Achieve Peak Mindfulness and Unlock Your Full Potential by Amishi P. Jha (2021; via Ann Rose Greenberg).

Learn Better: Mastering the Skills for Success in Life, Business, and School, or How to Become an Expert in Just About Anything

Sun Aug 31 2025 21:37:53 GMT-0400 (Eastern Daylight Time)

📖 Learn Better: Mastering the Skills for Success in Life, Business, and School, or How to Become an Expert in Just About Anything by Ulrich Boser (2017).

Multimodal AI Models on Apple Silicon with MLX with Prince Canuma

Sun Aug 31 2025 21:34:22 GMT-0400 (Eastern Daylight Time)

🎙️ Multimodal AI Models on Apple Silicon with MLX with Prince Canuma (Prince Canuma / TWIMLAI). Very cool story about how MLX came to be and how it got good.

Is AI already impacting the job market?

Sun Aug 31 2025 21:32:42 GMT-0400 (Eastern Daylight Time)

🐦 Is AI already impacting the job market? (Bharat Chandar). Seems like solid initial signal from paycheck data.

Nine Rules for Managing Humans Managing Nuclear Reactors

Sun Aug 31 2025 21:29:03 GMT-0400 (Eastern Daylight Time)

📝 Nine Rules for Managing Humans Managing Nuclear Reactors (Charles Yang / Asterisk). Includes a reprint of a powerful article by Admiral Hyman Rickover from 1979.

Class Dismissed

Sun Aug 31 2025 21:27:05 GMT-0400 (Eastern Daylight Time)

📝 Class Dismissed (Jeremy Stern / Colossus). Deep-dive interview with Joe Liemandt to include details about Alpha School.

10 Lessons from Ryan Petersen

Sun Aug 31 2025 21:25:12 GMT-0400 (Eastern Daylight Time)

🐦 10 Lessons from Ryan Petersen (Shane Parrish).

Building Ultra Cheap Energy Storage for Solar PV

Sun Aug 31 2025 21:22:49 GMT-0400 (Eastern Daylight Time)

📝 Building Ultra Cheap Energy Storage for Solar PV (Austin Vernon). Detailed and with strong testable claims for the business Austin is building.

Designing a mobile-first HTML editor

Sun Aug 31 2025 21:20:04 GMT-0400 (Eastern Daylight Time)

📝 Designing a mobile-first HTML editor.

The McPhee method

Sun Aug 31 2025 21:17:14 GMT-0400 (Eastern Daylight Time)

📝 The McPhee method (James Somers). If you capture and enumerate enough detail, you'll have the materials to mechanically produce great work.

“Asthenes” as a Jewish Textual Reference to Obsessive-Compulsive Disorder

Sun Aug 31 2025 21:14:17 GMT-0400 (Eastern Daylight Time)

📝 “Asthenes” as a Jewish Textual Reference to Obsessive-Compulsive Disorder (Shayna Herszage-Feldan / Lehrhaus). Despite knowing all the behaviors from the Gemara, I don't think I ever thought of OCD as explaining them.

What are Forward Deployed Engineers, and why are they so in demand?

Sun Aug 31 2025 21:11:50 GMT-0400 (Eastern Daylight Time)

📝 What are Forward Deployed Engineers, and why are they so in demand? (Gregely Orosz / The Pragmatic Engineer). An enzyme-like position that empowers engineers to convert specific customer needs into product features.

Starting with pytest’s parametrize

Sun Aug 31 2025 21:10:17 GMT-0400 (Eastern Daylight Time)

📝 Starting with pytest’s parametrize (Ned Batchelder). I have so many tests that could use this.

NIST Finalizes ‘Lightweight Cryptography’ Standard to Protect Small Devices

Sun Aug 31 2025 21:08:27 GMT-0400 (Eastern Daylight Time)

📝 NIST Finalizes ‘Lightweight Cryptography’ Standard to Protect Small Devices (NIST). I guess better 15 years late than never.

Of Marx and Moloch: How My Attempt to Convince Effective Altruists to Become Socialists Backfired Completely

Sun Aug 31 2025 21:05:49 GMT-0400 (Eastern Daylight Time)

📝 Of Marx and Moloch: How My Attempt to Convince Effective Altruists to Become Socialists Backfired Completely (Lennox). We should strongly encourage people to admit when they change their mind.

Virtual School Hit the Mainstream 5 Years Ago. How Popular Has It Gotten?

Sun Aug 31 2025 21:04:30 GMT-0400 (Eastern Daylight Time)

📝 Virtual School Hit the Mainstream 5 Years Ago. How Popular Has It Gotten? (Lauren Coffey / EdSurge). Somewhat.

AI: great expectations

Sun Aug 31 2025 21:01:39 GMT-0400 (Eastern Daylight Time)

📝 AI: great expectations (Rodney Brooks). This is a republication of Rodney's article from March 1988. Great example of the difficulty of "over and under estimating the power of AI."

Heritability puzzlers

Sun Aug 31 2025 20:58:58 GMT-0400 (Eastern Daylight Time)

📝 Heritability puzzlers (Dynomight). Pretty good explanation of what "heritability" actually means. Still confusing.

Your Review: My Father’s Instant Mashed Potatoes

Sun Aug 31 2025 20:56:32 GMT-0400 (Eastern Daylight Time)

📝 Your Review: My Father’s Instant Mashed Potatoes (Chris Finkle / ACX). Simulacra of simulacra.

Developers, Reinvented

Sun Aug 31 2025 20:55:16 GMT-0400 (Eastern Daylight Time)

📝 Developers, Reinvented (Thomas Dohmke).

Bag of words, have mercy on us

Sun Aug 31 2025 20:52:42 GMT-0400 (Eastern Daylight Time)

📝 Bag of words, have mercy on us (Adam Mastroianni / Experimental History). Humans gonna anthropomorphize.

No, AI is not Making Engineers 10x as Productive

Sun Aug 31 2025 20:44:50 GMT-0400 (Eastern Daylight Time)

📝 No, AI is not Making Engineers 10x as Productive (Colton Voege). One time I was in a meeting with over 100 people where we were told that we had to improve our performance by 10X year-over-year. After many rousing speeches of how important it was to 10X our performance, I asked the only question that came to mind: "For how many years in a row, do we expect to be able to 10X our performance?" The non-engineering speaker did not understand why all the engineers laughed.

The ChatGPT sharing dialog demonstrates how difficult it is to design privacy preferences

Sun Aug 31 2025 20:36:47 GMT-0400 (Eastern Daylight Time)

📝 The ChatGPT sharing dialog demonstrates how difficult it is to design privacy preferences (Simon Willison). Soon: NeuraLink policy to default share all thoughts to the internet considered harmful.

An Unsolved Victorian-Era Riddle

Fri Aug 01 2025 09:21:09 GMT-0400 (Eastern Daylight Time)

I came across this riddle, purportedly by Bishop Samuel Wilberforce, but it seems that attribution is in error:

I’m the sweetest of voices in orchestra heard,
But yet in an orchestra never have been.
I’m a bird of fine plumage, but less like a bird
Nothing ever in Nature was seen.
Touching earth I expire; in water I die;
Though I do progress, I can run, swim and fly.
Darkness destroys me, and light is my death;
And I can't keep alive without stopping my breath.
If my name can’t be guessed by a boy or a man,
By a woman or girl it certainly can.

Some proposed answers:

"Time" or "Whale" ("orca") (via Anomaly Info)
"Angel" (via Notes and Queries) meaning "(1) a heavenly being, (2) an old English coin, and (3) a fish."
"Comet" or "Painting" (via Reddit)

I sort of get "Whale" and "Angel" (assuming those definitions are real), but I don't get the others, so I'd like to propose a different solution: "The HMS Kestrel".

The ship did exist in Wilberforce's time (in case he did write the riddle): it was launched in 1856 when Wilberforce was 51 (he died in 1873, aged 68) and it was sold in 1866 after seeing some action in the 1850s and 1860s.

My analysis:

I’m the sweetest of voices in orchestra heard,
But yet in an orchestra never have been.

"Kestrel" is a play on the prominent sound ("sweetest sound") in "Orchestra," but the ship was never in an orchestra.

I’m a bird of fine plumage, but less like a bird
Nothing ever in Nature was seen.

A kestrel is "the common name give to several species of predatory birds," so its a "bird of fine plumage". A ship looks nothing like a bird.

Touching earth I expire; in water I die; Though I do progress, I can run, swim and fly.

The HMS Kestrel was a Clown-class gunboat which had a wood hull, steam power, and sails. They were often used for shallow water bombardment.

The dangers of running aground ("touching earth") and sinking ("in water") were real (see next item). The "run" of the steam engine helped "swim" (in water) while the sails let it "fly" (by air).

Darkness destroys me, and light is my death;
And I can't keep alive without stopping my breath.

Bit of a stretch, but with ChatGPT's help: Night operations ("darkness") and being discovered by the enemy ("light") were both operational hazards. Indeed, the HMS Kestrel was part of the second Battle of Taku Forts:

The attack was a bloody shambles, and it was clear that retreat was the only alternative to complete annihilation. Lee had grounded, and Kestrel had sunk halfway to its funnels in the Hai’s brown waters. Cormorant was a loss, and the shattered Plover was grounded and abandoned. The Kestrel was later recovered, but the other three were written off as total losses.

To avoid detection, perhaps the ship had to turn off the loud steam engine ("stopping my breath").

If my name can’t be guessed by a boy or a man,
By a woman or girl it certainly can.

A kestrel is a predatory bird which Victorians considered a masculine trait, but birds, more generally, were largely associated with feminity.

Literacy lag: We start reading too late

Fri Aug 01 2025 08:37:49 GMT-0400 (Eastern Daylight Time)

📝 Literacy lag: We start reading too late (Erik Hoel / The Intrinsic Perspective). I'm not saying you need to force parents to teach their 3 year-olds to read, but claiming it's not "developmentally appropriate" seems crazy.

Physicists disagree wildly on what quantum mechanics says about reality

Fri Aug 01 2025 08:30:45 GMT-0400 (Eastern Daylight Time)

📝 Physicists disagree wildly on what quantum mechanics says about reality (Elizabeth Gibney / Nature; via Jeremy Wertheimer). I took the quiz and received:

You are probably content with a Copenhagen-like interpretation.

You enjoy the idea that quantum mechanics is, in some way, able to describe the real world and you enjoy that there is some inherent randomness in the Universe. For you, the observer of quantum objects, and their classical world, is distinct from the quantum systems they interact with; a quantum particle's location and other attributes are not pre-defined, but emerge as well-defined states when an observer measures them.

Exactly why a measurement triggers a shift from a probabilistic to a well-defined state is not clear, but our brains are macroscopic, so you don’t let that bother you.

You might be among the 10% of researchers who, according to our survey, said that the Copenhagen interpretation was their favourite approach, but who also think that the wavefunction represents reality in some way.

Worship of the Heart: Essays on Jewish Prayer

Thu Jul 31 2025 23:35:53 GMT-0400 (Eastern Daylight Time)

📖 Worship of the Heart: Essays on Jewish Prayer by Rabbi Joseph B. Soloveitchik (2003). I liked the analysis of boredom.

Here Comes Everybody: The Power of Organizing Without Organizations

Thu Jul 31 2025 23:33:03 GMT-0400 (Eastern Daylight Time)

📖 Here Comes Everybody: The Power of Organizing Without Organizations by Clay Shirky (2008; via Tyler Cowen). This is almost 20 years old. Quaint, but not wrong.

Machiavelli for Women: Defend Your Worth, Grow Your Ambition, and Win the Workplace

Thu Jul 31 2025 23:29:19 GMT-0400 (Eastern Daylight Time)

📖 Machiavelli for Women: Defend Your Worth, Grow Your Ambition, and Win the Workplace by Stacey Vanek Smith (2021). Extremely pragmatic and not evil.

The Secrets of Happy Families: Improve Your Mornings, Rethink Family Dinner, Fight Smarter, Go Out and Play, and Much More

Thu Jul 31 2025 23:26:40 GMT-0400 (Eastern Daylight Time)

📖 The Secrets of Happy Families: Improve Your Mornings, Rethink Family Dinner, Fight Smarter, Go Out and Play, and Much More by Bruce Feiler (2013).

What's really true about intelligence and IQ? We empirically tested 40 claims

Thu Jul 31 2025 23:21:50 GMT-0400 (Eastern Daylight Time)

📝 What's really true about intelligence and IQ? We empirically tested 40 claims (Nikola Erceg, Spencer Greenberg, and Beleń Cobeta / Clearer Thinking).

Critique of Paper: “An astonishing regularity in student learning rate”

Thu Jul 31 2025 23:19:34 GMT-0400 (Eastern Daylight Time)

📝 Critique of Paper: “An astonishing regularity in student learning rate” (Justin Skycak). Extrapolating from initial instructions is huge.

Making Postgres 42,000x slower because I am unemployed

Thu Jul 31 2025 23:17:53 GMT-0400 (Eastern Daylight Time)

📝 Making Postgres 42,000x slower because I am unemployed (Jacob Jackson). Negative examples can be useful. See also A Quick(ish) Introduction to Tuning Postgres for the positive version.

Enough AI copilots! We need AI HUDs

Thu Jul 31 2025 23:16:12 GMT-0400 (Eastern Daylight Time)

📝 Enough AI copilots! We need AI HUDs (Geoffrey Litt). Interesting UX implications.

How the Solid Protocol Restores Digital Agency

Thu Jul 31 2025 23:14:03 GMT-0400 (Eastern Daylight Time)

📝 How the Solid Protocol Restores Digital Agency (Bruce Schneier). An interesting attempt to invert the data flow.

Task Runner Census 2025

Thu Jul 31 2025 23:11:16 GMT-0400 (Eastern Daylight Time)

📝 Task Runner Census 2025 (Alexander Yankov). I saw this because I'm tracking this uv issue. This might have impacts on how ds should work.

Could AI slow science?

Thu Jul 31 2025 23:09:36 GMT-0400 (Eastern Daylight Time)

📝 Could AI slow science? (Sayash Kapoor and Arvind Narayanan / AI Snake Oil). Maybe.

Fell in a hole, got out.

Thu Jul 31 2025 23:07:23 GMT-0400 (Eastern Daylight Time)

📝 Fell in a hole, got out. (Tony Stubblebine). I had no idea about the tumultuous story of how Medium got to profitability in August 2024. Lots of basics had to be aligned.

We aren't worried about misalignment as self-fulfilling prophecy

Thu Jul 31 2025 23:03:31 GMT-0400 (Eastern Daylight Time)

📝 We aren't worried about misalignment as self-fulfilling prophecy (Scott Alexander et al. / AI Futures Project). The topic under discussion is similar to Ecclesiastes 10:20:

Don’t revile a king even among your intimates. Don’t revile a rich man even in your bedchamber; For a bird of the air may carry the utterance, And a winged creature may report the word.

The HALO Effect

Thu Jul 31 2025 22:58:10 GMT-0400 (Eastern Daylight Time)

📝 The HALO Effect (Kevin Kwok). Interesting new arrangement ("Hire and License Out") that is similar to acqui-hiring, but much faster.

Reflections on OpenAI

Thu Jul 31 2025 22:48:14 GMT-0400 (Eastern Daylight Time)

📝 Reflections on OpenAI (Calvin French-Owen; via Armin Ronacher).

Grouping Students by Ability Transformed Our School. Could It Do the Same for Yours?

Thu Jul 31 2025 22:43:50 GMT-0400 (Eastern Daylight Time)

📝 Grouping Students by Ability Transformed Our School. Could It Do the Same for Yours? (Jessica Berg / Center for Educational Progress).

Happy 20th birthday Django!

Thu Jul 31 2025 22:40:30 GMT-0400 (Eastern Daylight Time)

📝 Happy 20th birthday Django! (Simon Willison). Lots of interesting details.

Joining Noetik

Thu Jul 31 2025 22:37:59 GMT-0400 (Eastern Daylight Time)

📝 Joining Noetik (Abhishaike Mahajan; via Shalev NessAiver). Nice enumeration of reasons why the ML-Bio field is collapsing and a company (Noetik) possibly bucking the trend.

Everything I know about good system design

Thu Jul 31 2025 22:35:29 GMT-0400 (Eastern Daylight Time)

📝 Everything I know about good system design (Sean Goedecke). This is about system design; it's a programming article, but it uses a lot of intuition about what makes programs feel good (proper caching, hot path optimization, etc.).

On agency

Thu Jul 31 2025 22:33:48 GMT-0400 (Eastern Daylight Time)

📝 On agency (Henrik Karlsson / Escaping Flatland).

Often, agency is almost gentle—an attunement to the world and the self, a feeling out the details of reality, and a finding of the path of least resistance.

Solving Wordle with uv's dependency resolver

Thu Jul 31 2025 22:31:08 GMT-0400 (Eastern Daylight Time)

📝 Solving Wordle with uv's dependency resolver (mildbyte.xyz). Because of course you can.

When AI Codes, What’s Left for me?

Thu Jul 31 2025 22:28:47 GMT-0400 (Eastern Daylight Time)

📝 When AI Codes, What’s Left for me? (Adam Gordon Bell / Corecursive). Nice discussion about reinventing yourself in the age of AI coding agents.

I Deleted My Second Brain

Thu Jul 31 2025 22:25:36 GMT-0400 (Eastern Daylight Time)

📝 I Deleted My Second Brain (Joan Westenberg). I don't love the "second brain" nomenclature, but I agree that carefully cataloging everything, but never using it is just another form of navel-gazing.

How to Fix Your Context

Mon Jun 30 2025 22:30:49 GMT-0400 (Eastern Daylight Time)

📝 How to Fix Your Context (Drew Breunig; via Simon Willison). Drew is coining lots of little phrases around the problems of LLM context (even in huge 1M context windows). Better yet, he's describing how to fix those problems.

Your Review: Alpha School

Mon Jun 30 2025 22:28:49 GMT-0400 (Eastern Daylight Time)

📝 Your Review: Alpha School (Edward Nevraumont / ACX). This is the most detail I've gotten about Alpha School since I heard about them a few years ago. This review actually immediately changed my perspective on what I could have my son accomplish over the summer and we started implementing it yesterday. Let's see how this goes.

Sam Altman on Founder Secrecy: No One Cares

Mon Jun 30 2025 22:26:36 GMT-0400 (Eastern Daylight Time)

🐦 Sam Altman on Founder Secrecy: No One Cares (Sam Altman). "No matter how great your idea is, no one cares." I see this a lot.

Maybe you’re not Actually Trying

Mon Jun 30 2025 22:24:39 GMT-0400 (Eastern Daylight Time)

📝 Maybe you’re not Actually Trying (Cate Hall / Useful Fictions; via Thinking About Things). But actual reasons why you might not be trying.

AI to the Rescue

Mon Jun 30 2025 22:18:27 GMT-0400 (Eastern Daylight Time)

📝 AI to the Rescue (Beth McMurtrie / The Chronicle of Higher Education; via Jeremy Wertheimer). Strong students ask questions; weaker students try to avoid asking questions.

Face it: you're a crazy person

Mon Jun 30 2025 22:16:02 GMT-0400 (Eastern Daylight Time)

📝 Face it: you're a crazy person (Adam Mastroianni / Experimental History). Unpacking a job into what it actually entails is a thing I actually enjoy doing.

The Real Reasons Your Appliances Die Young

Mon Jun 30 2025 22:14:26 GMT-0400 (Eastern Daylight Time)

📝 The Real Reasons Your Appliances Die Young (Rachel Wharton / Wirecutter). Did you have "It's Complicated!" on your bingo card?

How not to lose your job to AI

Mon Jun 30 2025 22:12:51 GMT-0400 (Eastern Daylight Time)

📝 How not to lose your job to AI (Benjamin Todd / 80,000 Hours). We're all managing someone's context now.

Writing Toy Software Is A Joy

Mon Jun 30 2025 22:08:50 GMT-0400 (Eastern Daylight Time)

📝 Writing Toy Software Is A Joy (Joshua Barretto). Nice eclectic set of examples.

Running "Running a Bootcamp" Bootcamp

Mon Jun 30 2025 22:06:17 GMT-0400 (Eastern Daylight Time)

📝 Running "Running a Bootcamp" Bootcamp (Ricki Heicklen). Teaching teacher-teachers how to teach teachers to teach.

Going to an office and pretending to work: A business that’s booming in China

Mon Jun 30 2025 22:03:25 GMT-0400 (Eastern Daylight Time)

📝 Going to an office and pretending to work: A business that’s booming in China (Inma Bonet / El Pais; via Tyler Cowen). It's a role-playing co-working space.

How I Built A Task Management Tool For Almost Nothing

Mon Jun 30 2025 22:01:09 GMT-0400 (Eastern Daylight Time)

📝 How I Built A Task Management Tool For Almost Nothing (Mike Masnick / TechDirt). We're entering an age of bespoke tools. Mike is a power user who had very little programming experience, but a clear vision for what he wants.

Consider Knitting

Mon Jun 30 2025 21:55:18 GMT-0400 (Eastern Daylight Time)

📝 Consider Knitting (Bob Nystrom). Not sure I'm going to start knitting, but Bob makes a pretty good case for it.

Holacracy: The New Management System for a Rapidly Changing World

Sat May 31 2025 22:31:41 GMT-0400 (Eastern Daylight Time)

📖 Holacracy: The New Management System for a Rapidly Changing World by Brian J. Robertson (2015). If you wait a decade to read a book, you'll discover that all the major early adopters (Medium, Zappos) have already abandoned this newfangled philosophy.

The Great Z'manim Debate: The History, The Science, And The Lomdus

Sat May 31 2025 22:28:45 GMT-0400 (Eastern Daylight Time)

📖 The Great Z'manim Debate: The History, The Science, And The Lomdus by Rabbi Ahron Notis (2022; via Harold Zazula). Solid introduction to the relevant astronomical terms and older and more modern astronomical models. Good discussion of the relevant halachot.

Halakhic Positions of Rabbi Joseph B. Soloveitchik: Volume 8

Sat May 31 2025 22:25:53 GMT-0400 (Eastern Daylight Time)

📖 Halakhic Positions of Rabbi Joseph B. Soloveitchik: Volume 8 by Aharon Ziegler (2020). Only two more to go until I finish the series.

Outlive: The Science and Art of Longevity

Sat May 31 2025 22:21:56 GMT-0400 (Eastern Daylight Time)

📖 Outlive: The Science and Art of Longevity by Peter Attia & Bill Gifford (2023; via Isaac Selya). Assuming you want to have certain kinds of mobility when you're older, this book helps you work backwards to where you want to be today.

The Education of an Idealist: A Memoir

Sat May 31 2025 22:16:42 GMT-0400 (Eastern Daylight Time)

📖 The Education of an Idealist: A Memoir by Samantha Power (2019; via Tyler Cowen).

Some may interpret this book’s title as suggesting that I began with lofty dreams about how one person could make a difference, only to be “educated” by the brutish forces that I encountered. That is not the story that follows.

I mean, that's kinda what happens. She learns very quickly how to say the right things to get the things she wants.

The E-Myth Revisited: Why Most Small Businesses Don't Work and What to Do About It

Sat May 31 2025 22:11:25 GMT-0400 (Eastern Daylight Time)

📖 The E-Myth Revisited: Why Most Small Businesses Don't Work and What to Do About It by Michael E. Gerber (2009; via Shalev NessAiver). Inside you there are three wolves. The Technician just wants to do good work and thinks bottom up. The Entrepreneur looks ahead and thinks top down. The manager wants everything in a little box and hates change.

Inside the AI boom that's transforming how consultants work at McKinsey, BCG, and Deloitte

Sat May 31 2025 22:00:55 GMT-0400 (Eastern Daylight Time)

📝 Inside the AI boom that's transforming how consultants work at McKinsey, BCG, and Deloitte (Lakshmi Varanasi / Business Insider; via Moshe Kinderlehrer). I wonder how long this policy will last:

McKinsey consultants have access to ChatGPT with guardrails and cannot input client data into it.

DOGE Days

Sat May 31 2025 21:57:27 GMT-0400 (Eastern Daylight Time)

📝 DOGE Days (Sahil Lavingia; via Tyler Cowen). In which Sahil discovers, like many before him, that you may have aspirations to write wonderful code to help the government, but you will be thwarted every step of the way.

Saying Bye to Glitch

Sat May 31 2025 21:53:24 GMT-0400 (Eastern Daylight Time)

📝 Saying Bye to Glitch (Pirijan Keth; via Simon Willison). I never used Glitch, but I remember when it was announced and thinking that it was very cool. Pirijan also describes how various FogCreek software was ahead of its time.

Thoughts on thinking

Sat May 31 2025 21:51:48 GMT-0400 (Eastern Daylight Time)

📝 Thoughts on thinking (Dustin Curtis). As the nature of work transitions, I expect to see lots of posts of this nature.

How I taught my 3-year-old to read like a 9-year-old

Sat May 31 2025 21:48:54 GMT-0400 (Eastern Daylight Time)

📝 How I taught my 3-year-old to read like a 9-year-old (Erik Hoel / The Intrinsic Perspective).

It’s all the advantages of an iPad, none of the guilt. You’ve unlocked infinite self-entertainment. Long drive? Bring a book. Or five. Roman toddles into restaurants clutching a book as a backup activity, and reads while waiting in boring lines. It’s also calming, and so helps with emotional regulation. Toddler energy descending rapidly into deviance? Go read a book! It’s a parenting cheat code. I don’t know if this alone justifies the hours spent, but it sure is one heck of a benefit.

I really don’t like ChatGPT’s new memory dossier

Sat May 31 2025 21:46:08 GMT-0400 (Eastern Daylight Time)

📝 I really don’t like ChatGPT’s new memory dossier (Simon Willison). It's pretty detailed.

LLM Memory

Wed May 21 2025 19:35:02 GMT-0400 (Eastern Daylight Time)

📝 LLM Memory (Grant Slatton). I like mini taxonomy of approaches to memory. I think this shows us that we don't yet have a really good model for our own memory.

The Price of Remission

Wed May 21 2025 19:31:42 GMT-0400 (Eastern Daylight Time)

📝 The Price of Remission (David Armstrong / ProPublica).

When I started taking the drug, I’d look at the smooth, cylindrical capsule in my hand and consider the fact I was about to swallow something that costs about the same as a new iPhone. A month’s supply, which arrives in an ordinary, orange-tinged plastic bottle, is the same price as a new Nissan Versa.

Book Review: Selfish Reasons To Have More Kids

Wed May 21 2025 19:20:22 GMT-0400 (Eastern Daylight Time)

📝 Book Review: Selfish Reasons To Have More Kids (Scott Alexander / Astral Codex Ten). "Have you tried doing less?" is often a good question. I don't think I've been able to implement the advice (at all?).

AI-Generated Law

Wed May 21 2025 19:13:10 GMT-0400 (Eastern Daylight Time)

📝 AI-Generated Law (Bruce Schneier). Soon: AI laws on the limits of AI laws.

I don't like NumPy

Wed May 21 2025 19:00:51 GMT-0400 (Eastern Daylight Time)

📝 I don't like NumPy (Dynomight). This one is for everyone who has ever had to try axis=0 when doing an operation in numpy. The points about indexing and broadcasting hit home hard. I'm looking forward to the API proposals.

castfit 0.1.2

Wed May 21 2025 18:52:41 GMT-0400 (Eastern Daylight Time)

castfit 0.1.2 is available. I previously asked both Gemini 2.5 Pro and ChatGPT o3 to review my previous release and provide suggestions. While implementing those suggestions I ended up simplifying several things and making certain interfaces more consistent.

To install castfit:

# modern (recommended)
uv add castfit

# classic
python -m pip install castfit

Release Notes 0.1.2 - 2025-05-21T20:54:51Z #

Fixed

#4: get_args on raw types
#5: to_tuple with too-long input
#18: handling of types.UnionType
#25: delinted using pyrefly
#27: str to int conversion when the string has decimal places
#28: float to datetime conversion; added UTC timezone
#29: handling an untyped default value in a class

Changed

#14: TypeForm comment to clarify what we want
#19: set instance fields based on class metadata rather than tried to put all the data into the instance
#22: register converters based on both source and destination types rather than assuming that each function must convert everything to a specific destination type
#24: renamed casts_to to casts and added support for short-form (1 argument) cast function
#30: updated the public API to be more compact and consistent

Added

#2: support for nested types
#3: original cause of to_type error
#6: additional datetime formats
#7: custom casts to castfit (closes #7)
#11: more README examples
#12: more complete docstrings for the public API
#15: cache for fetching get_origin and get_args information
#16: DEFAULT_ENCODING constant for to_bytes and to_str
#17: alternatives to README
#20: infer types based on class field defaults
#31: more negative tests

Removed

#21: castfit on an instance instead of a type

Won't Fix

#8: Gemini suggested having an explicit caster for pathlib.Path
#9: Gemini suggested having an explicit recursive dataclass/class casting
#10: Gemini suggested optionally collecting errors instead of raising the first one encountered. It's a good idea, but not for now.
#13: Tried implementing a workaround for TypeGuard in older versions of python, but it didn't work.
#23: Started and rolled back is_callable because castfit can't currently do anything with a callable that is the wrong type.
#26: Rolled back having a checks parameter that overrides how types are checked.
#32: Tried fixing TypeForm to be the union of type[T] and _SpecialForm, but only pyright was able to handle it. mypy still can't handle it and ty isn't mature enough yet.

TIL: invariant, covariant, and contravariant type variables

Fri May 16 2025 12:36:32 GMT-0400 (Eastern Daylight Time)

While working on castfit, I ended up going down a rabbit hole learning about what makes something a subtype of something else. I vaguely knew the terms "invariant", "covariant", and "contravariant", but have very carefully avoided ever learning what they meant.

If you look at the Wikipedia entry on Covariance and contravariance (computer science) you see explanations like:

Suppose A and B are types, and I<U> denotes application of a type constructor I with type argument U. Within the type system of a programming language, a typing rule for a type constructor I is:

covariant if it preserves the ordering of types (≤), which orders types from more specific to more generic: If A ≤ B, then I<A> ≤ I<B>;

contravariant if it reverses this ordering: If A ≤ B, then I<B> ≤ I<A>;

bivariant if both of these apply (i.e., if A ≤ B, then I<A> ≡ I<B>);[1]

variant if covariant, contravariant or bivariant;

invariant or nonvariant if not variant.

Uh, super helpful. The introductory description is a lot better than it used to be, but talking it over with ChatGPT was actually the best way for me to understand what these terms mean.

Simpler Definitions #

Suppose we have a type hierarchy like:

Labrador → Dog → Animal
Siamese  → Cat → Animal

The arrow indicates an "is a" relationship moving from more specific (narrow) to more general (broad): a Labrador is a type of Dog; a Dog is a type of Animal, etc.

The question of co/contra/invariance comes up when you want to know: Can I substitute one type for another?

It turns out that this depends on many things, but the situations are classified as:

Invariant: It must be this exact type. No broader or narrower substitutions allowed.
Covariant: The same or broader (more general) type is acceptable.
Contravariant: The same or narrower (more specific) type is acceptable.
Bivariant (haven't seen this in Python): The same, broader, or narrower type is acceptable.

This mostly comes into play when you start looking at containers of objects like list, set, tuple, dict, etc. Also Python has several exceptions to its own rules that are not immediately obvious.

General Python Rules #

The contents of mutable containers are invariant. For example, if a function takes list[Animal], you cannot pass list[Dog] because that function might add a Cat (which is an Animal) there by violating the type safety of list[Dog] that was passed in.

An exception to this rule in Python is when "implicit conversion" occurs. A function that takes list[int] is ok to take list[bool] because there is an implicit conversion that happens. It seems that the numeric tower of bool → int → float → complex all happens implicitly. There are a few other implicit conversions that I'm still learning about.

The contents of immutable containers (or read-only situations) are covariant. A function that takes Sequence[Animal] is ok to take Sequence[Dog] because Sequence is read-only (and Animal is broader than Dog).

Not really an exception, but fun fact: fixed-length tuple are covariant, while variable-length tuples are invariant.

PEP 483 has a nice example of contravariant types: Callable. While it is covariant in its return type (Callable[[], Dog] is a subtype of Callable[[], Animal]), it is contravariant in its arguments (Callable[[Animal], None] is a subtype of Callable[[Dog], None]).

This leads to a good rule of thumb:

The example with Callable shows how to make more precise type annotations for functions: choose the most general type for every argument, and the most specific type for the return value.

It looks like PEP 695 made it into Python 3.12, so maybe reasoning about this will get easier in the future especially because it automatically infers the variance of the type variables.

Trying pyrefly, Meta's type checker

Fri May 16 2025 12:08:15 GMT-0400 (Eastern Daylight Time)

Meta announced that they are open sourcing their type checker, pyrefly. It was easy enough to try:

uvx pyrefly check

On my tiny 640 lines of castfit (which passes pyright and mypy --strict) I got 74 errors. Going through some of these was instructive.

== is not supported between type[@_] and type[MyList]

This was in response to code like:

assert castfit.get_origin_type(MyList) == MyList

Good catch. That == should is.

Can't apply arguments to non-class, got LegacyImplicitTypeAlias[TypeForm, type[type[TypeVar(T, invariant)] | Any]] [bad-specialization]

Not sure what this means or how to handle it.

Argument Forall[T, (value: Any, kind: Unknown) -> bool] is not assignable to parameter with type (Any, Unknown) -> bool [bad-argument-type]

Imagine my confusion when I'm trying to figure out why this is not assignable. Must be that Unknown can't be assigned to Unknown. I might try this again in the future, but the error messages could definitely use some work.

Tom Johnson on AI-assisted Programming

Fri May 16 2025 12:04:01 GMT-0400 (Eastern Daylight Time)

🐦 Tom Johnson on AI-assisted Programming (Tom Johnson).

I just spent a week using Cursor and PyCharm's AI [Assistant], The goal was to take a CLI-based Python script and add a basic front end UI. By the end of the day yesterday, I felt like chewing aluminum foil would be more fun ...

Recommend reading the whole thread. There are ways I imagine asking the AI differently, but I completely get the pain being described.

New paradigm for psychology just dropped

Fri May 16 2025 11:51:35 GMT-0400 (Eastern Daylight Time)

📝 New paradigm for psychology just dropped (Adam Mastroianni). I queued up the The Mind in the Wheel by Slime Mold Time Mold, but haven't read it yet. Adam's point is that a paradigm has to explain which things do what and what rules they follow.

So let’s get clear: a paradigm is made out of units and rules. It says, “the part of the world I’m studying is made up of these entities, which can do these activities.”

He lays out three different kinds of research:

Naive research: run experiments without knowing anything about the units and rules.
Impressionistic research: make up words and study them.

The problem with this approach is that it gets you tangled up in things that don’t actually exist. What is “zest for life”? It is literally “how you respond to the Zest for Life Scale”. And what does the Zest for Life Scale measure? It measures...zest for life.

Actual science research (my name): doing experiments where you can actually learn about the units and rules.

We’re not doing impressionistic research here, so we can’t just create control systems by fiat, the way you can create “zest for life” by creating a Zest for Life Scale. Instead, discovering the drives requires a new set of methodologies. You might start by noticing that people seem inexplicably driven to do some things (like play Candy Crush) or inexplicably not driven to do other things (like drink lemon juice when they’re suffering from scurvy, even though it would save their life). This could give you an inkling of what kind of drives exist. Then you could try to isolate one of those drives through methods like:

Prevention: If you stop someone from playing Candy Crush, what do they do instead?

Knockout: If you turn off the elements of Candy Crush one at a time—make it black and white, eliminate the scoring system, etc.—at what point do they no longer want to play?

Behavioral exhaustion (knockout in reverse): If you give people one component of Candy Crush at a time—maybe, categorizing things, earning points, seeing lots of colors, etc.—and let them do that as much as they want, do they still want to play Candy Crush afterward?

Adam also covers why neuroscience is the wrong level of abstraction for learning the things we need to learn.

7 Phrases I use to make giving feedback easier for myself

Fri May 16 2025 11:42:05 GMT-0400 (Eastern Daylight Time)

📝 7 Phrases I use to make giving feedback easier for myself (Wes Kao). "I noticed" that "this is a great start" for giving feedback in a modern environment. "At the same time" going forward you could sound "even more" authentic because "I believe you were trying to avoid hard feelings, but it doesn't quite work because it uses canned phrases. I recommend trying having a conversation." "From what I've seen" canned phrases come across as fake. You're "already" giving feedback regularly which is fantastic. Now, I'd love for you to have more real conversations where you're clear about your expectations.

(Seriously, though, the ideas behind the phrases and examples are extremely realistic.)

The Wrong Way to Motivate Your Kid

Fri May 16 2025 11:38:28 GMT-0400 (Eastern Daylight Time)

📝 The Wrong Way to Motivate Your Kid (Russell Shaw / The Atlantic). The advice is to find an "island of competence" and help your kid build up from that. An implication is that we should help kids get competent at more things. Unfortunately, things are trending in the opposite direction.

How to title your blog post or whatever

Fri May 16 2025 11:33:29 GMT-0400 (Eastern Daylight Time)

📝 How to title your blog post or whatever (Dynomight). The advice is to name things in a way that quickly separates the people who'll love it from the people who hate it.

The magic of software; or, what makes a good engineer also makes a good engineering organization

Fri May 16 2025 11:24:51 GMT-0400 (Eastern Daylight Time)

📝 The magic of software; or, what makes a good engineer also makes a good engineering organization (Moxie Marlinspike; via Changelog News). Nice essay about how vision is translated into engineering. Also discusses abstraction layers and how organizations are structured.

However, there are two ways of interacting with an abstraction layer: as shorthand for an understanding of what it is doing on your behalf, or as a black box. I think engineers are most capable and most effective when abstraction layers serve as shorthand rather than a black box.

The Troubling Business of Bounty Hunting

Fri May 16 2025 11:16:21 GMT-0400 (Eastern Daylight Time)

📝 The Troubling Business of Bounty Hunting (Jeff Winkler / The New York Times; via Thinking About Things). I feel like my mind updated every movie plot that involves someone skipping bail or a bounty hunter to be much more terrifying.

How much information is in DNA?

Fri May 16 2025 11:08:06 GMT-0400 (Eastern Daylight Time)

📝 How much information is in DNA? (Dynomight). I knew about both Kolmogorov complexity (smallest program that produces an output) and Shannon information (probability of a sequence appearing in a large pool of sequences), but this was a cool way of combining the two concepts. Plus the craziness that is biology.

What I've learned about writing AI apps so far

Fri May 16 2025 11:04:43 GMT-0400 (Eastern Daylight Time)

📝 What I've learned about writing AI apps so far (Laurie Voss). Currently, LLMs are good at transforming text into less text. This implies you should give them lots of context (drop the whole manual / book / text) and ask a question.

Getting things "done" in large tech companies

Fri May 16 2025 10:56:31 GMT-0400 (Eastern Daylight Time)

📝 Getting things "done" in large tech companies (Sean Goedecke). Short post about how to know when a project is done and how to make sure executives know what was done.

AI as Normal Technology

Fri May 09 2025 16:31:55 GMT-0400 (Eastern Daylight Time)

📝 AI as Normal Technology (Arvind Narayanan & Sayash Kapoor / Knight First Amendment Institute). This is the approach being discussed in Does AI Progress Have a Speed Limit?. While I don't have as much evidence to support my intuitive opinion, I do agree that modeling the world (especially in general) is particularly hard.

Trying ty, Astral's type checker

Wed May 07 2025 19:08:33 GMT-0400 (Eastern Daylight Time)

Astral has been delivering impressive improvements to the Python ecosystem with ruff, taking over rye, and releasing uv to manage projects and Python installs. And there have been musings about when they'd release a type checker.

Well the alpha release is now here (via Patrick Kage). You can try it out:

# requires rust to compile
curl https://sh.rustup.rs -sSf | sh
uv add git+https://github.com/astral-sh/ty
uv run ty check

Note: It takes a little bit of time to compile ty, but like all the other Astral tools it runs very quickly.

Also: Fair warning, lots of stuff doesn't work yet. For example, when I tried it on my castfit library, it choked on dict() (see #100).

How long before ty replaces mypy and pyright in my build process for all my projects? I'll probably wait for the official release, but I'll probably try it at least once on all my projects just to see what it produces.

Dos and don'ts when sunsetting open source projects

Wed May 07 2025 15:21:35 GMT-0400 (Eastern Daylight Time)

📝 Dos and don'ts when sunsetting open source projects (Klint Finley / GitHub Blog). Good advice; I should probably archive a bunch of my older projects.

Project Xanadu: Even More Hindsight

Wed May 07 2025 08:23:56 GMT-0400 (Eastern Daylight Time)

📝 Project Xanadu: Even More Hindsight (Gwern Branwen). In Minority Report they have "futuristic" interfaces where you wave your arms around to manipulate 3D projections. In the real world your arms would get very tired very quickly. Project Xanadu tried to create a hypertext environment with lots of guarantees, but Gwern's realization about the UX is notable:

“Oh my god. It’s completely unreadable.”

The lines were confusing clutter, especially as they crisscrossed (a perennial problem in sidenotes layout, made far worse by the outlines). None of the ‘sidenotes’ were readable because the screen was so small. Even as you simply scrolled, for many possible settings, you were unable to read anything! How could a document UI where often you could read nothing have ever seemed like a good idea? The UI was just terrible—it could never have worked. Even on a large screen like my 4k monitor, I wouldn’t want that.

The lesson?

So, to me, Project Xanadu is a case-study in why designers must mock-up and prototype their designs before too much is invested in them. Xanadu wasn’t the victim of “Worse is Better”; it was just a solution in search of a problem.

Does AI Progress Have a Speed Limit?

Tue May 06 2025 15:37:31 GMT-0400 (Eastern Daylight Time)

📝 Does AI Progress Have a Speed Limit? (Ajeya Cotra & Arvind Narayanan / Asterisk). Very respectful back-and-forth. I'm partial to Arvind's view, but I don't have enough evidence to support it.

Parallels between Generative AI and Humanoid Robots

Tue May 06 2025 15:30:54 GMT-0400 (Eastern Daylight Time)

📝 Parallels between Generative AI and Humanoid Robots (Rodney Brooks). Worth reading the whole thing. Reminder of Rodney's First Law of AI:

When an AI system performs a task, human observers immediately estimate its general competence in areas that seem related. Usually that estimate is wildly overinflated.

Product Purgatory: When they love it but still don’t buy

Tue May 06 2025 15:24:41 GMT-0400 (Eastern Daylight Time)

📝 Product Purgatory: When they love it but still don’t buy (Jason Cohen). Main takeaways:

People are surprised to discover that the prospective customer wouldn't even take your solution if it was free and had many other magical properties. This is an important signal.
You need to solve a burning pain. Even if it's a great idea and otherwise important, if it's not urgent, it's probably not relevant.
Towards the end of the post, there's a nice bit about "how you would find customers in this condition" that I thought was useful.

You Sent the Message. But Did You Write It?

Tue May 06 2025 15:21:44 GMT-0400 (Eastern Daylight Time)

📝 You Sent the Message. But Did You Write It? (David Duncan). My favorite of these is GPTune:

Like Auto-Tune for writing. GPTune takes someone’s normal idea and smooths it into something that feels more articulate, structured, erudite - but less authentic.

Euromaidan Press on Gamified War

Tue May 06 2025 15:16:29 GMT-0400 (Eastern Daylight Time)

🐦 Euromaidan Press on Gamified War (Euromaidan Press).

Ukrainian soldiers get reward points for destroyed Russian targets, which they can then exchange for new equipment. Just like in a game. - Politico.

The exchange happens through an online marketplace called Brave 1 and the troops need to provide video evidence to earn points.

♦️6 points for eliminating a Russian soldier
♦️40 points for destroying a tank
♦️50 points for taking out a mobile rocket system

For example, 43 points can get a powerful "Vampire" drone capable of carrying a 15-kg warhead.

What benefits?
♦️gets equipment directly to the most effective fighting units without bureaucratic delays
♦️motivates soldiers through friendly competition
♦️has already doubled the rate of Russian casualties since adjusting the point values.
♦️improves Ukraine's military intelligence by creating a verified database of Russian losses

Incentives matter.

Making PyPI's test suite 81% faster

Tue May 06 2025 15:01:08 GMT-0400 (Eastern Daylight Time)

📝 Making PyPI's test suite 81% faster (Alexis Challande; via Simon Willison). One day I'll have enough tests to run on some project that I can try out pytest-xdist.

Two publishers and three authors fail to understand what “vibe coding” means

Tue May 06 2025 14:51:10 GMT-0400 (Eastern Daylight Time)

📝 Two publishers and three authors fail to understand what “vibe coding” means (Simon Willison). Simon later realized that it's all about the vibes.

What is wrong with TOML?

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 What is wrong with TOML? (Colm O'Connor). I like these spicy takes even if I'm not totally sold.

What happened to genetic algorithms?

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 What happened to genetic algorithms? (Jared Winslow).

Web 3.0 Requires Data Integrity

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Web 3.0 Requires Data Integrity (Bruce Schneier).

We Can, Must, and Will Simulate Nematode Brains

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 We Can, Must, and Will Simulate Nematode Brains (Michael Skuhersky / Asterisk). I mean must is kinda strong for this.

The Talmudic Argument

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📖 The Talmudic Argument by Louis Jacobs (1984; via Jeremy Wertheimer). Finally made some sense of a question I had since 8th grade: "Why didn't you just quote a few more words of that Braita earlier!?"

The Dyslexic Advantage (Revised and Updated): Unlocking the Hidden Potential of the Dyslexic Brain

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📖 The Dyslexic Advantage (Revised and Updated): Unlocking the Hidden Potential of the Dyslexic Brain by Brock Eide (2023; via Jacqueline Novikov).

The curve is bending

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 The curve is bending (Grant Slatton).

The Best Programmers I Know

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 The Best Programmers I Know (Matthias Endler).

Startup Exercise: What can't be solved with money?

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Startup Exercise: What can't be solved with money? (Jason Cohen).

Solution-space taste

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Solution-space taste (Grant Slatton).

Nutrition Beliefs Are Just-So Stories

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Nutrition Beliefs Are Just-So Stories (Cremieux Recueil).

Falsehoods software teams believe about user feedback

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Falsehoods software teams believe about user feedback (Fritz Meissner).

Cell Phone OPSEC for Border Crossings

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Cell Phone OPSEC for Border Crossings (Bruce Schneier).

Calm Down—Your Phone Isn’t Listening to Your Conversations. It’s Just Tracking Everything You Type, Every App You Use, Every Website You Visit, and Everywhere You Go in the Physical World

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Calm Down—Your Phone Isn’t Listening to Your Conversations. It’s Just Tracking Everything You Type, Every App You Use, Every Website You Visit, and Everywhere You Go in the Physical World (Jonathan Zeller / McSweeney's).

Boxie - an always offline audio player for my 3 year old

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 Boxie - an always offline audio player for my 3 year old (Mario Zechner).

Barking Up the Wrong Tree: The Surprising Science Behind Why Everything You Know About Success Is (Mostly) Wrong

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📖 Barking Up the Wrong Tree: The Surprising Science Behind Why Everything You Know About Success Is (Mostly) Wrong by Eric Barker (2017). Ugh. I probably need to remove all the books from 2016-2017 from my reading list that are overflowing with psych studies that never replicated. The mentorship parts were reasonably ok.

A Halakhic Guide to Dealing with Mental Illness

Wed Apr 30 2025 19:33:00 GMT-0400 (Eastern Daylight Time)

📝 A Halakhic Guide to Dealing with Mental Illness (Meir Ekstein / Lehrhaus).

Backfill your blog

Tue Apr 29 2025 18:56:08 GMT-0400 (Eastern Daylight Time)

📝 Backfill your blog (Simon Willison). I took the time to convert 9 years worth of Keep In Touch emails into Markdown so that I can start a link blog starting today.

Patrick McKenzie on using Dangerous Professional Voice

Tue Apr 29 2025 16:50:55 GMT-0400 (Eastern Daylight Time)

🐦 Patrick McKenzie on using Dangerous Professional Voice (Patrick McKenzie).

Memetically, being a Dangerous Professional means communicating in what might be a slightly adversarial context in a way which suggests that a bureaucracy take one’s concerns seriously and escalate them to someone empowered to resolve them swiftly.

cattrs I: un/structuring speed

Tue Apr 29 2025 13:48:24 GMT-0400 (Eastern Daylight Time)

📝 cattrs I: un/structuring speed (The Three Wands).

During the code review for castfit, Gemini pointed me at attrs and cattrs. Many of the features of the attrs became dataclasses in the standard python library while the features of the cattrs look very similar to my own castfit library.

I started reading up on cattrs and I came across the article above which explains the three phases of optimization that cattrs went through. The first phase roughly corresponds to how castfit does conversions today. It also highlights the same problems I anticipate if castfit were to ever be used in a production environment.

The second (generated code) and third phases (optimized bytecode) reflect some very interesting optimizations. I'm glad that they kept the initial phase code around for situations where you don't have time up front (e.g., at CLI start up).

My simple test didn't work:

uv run --with cattrs python

>>> import cattrs
>>> class Cat:
...     name: str
...     age: int
...
>>> cattrs.structure({"name": "Garfield", "age": 45}, Cat)
Traceback (most recent call last):
  File "<python-input-2>", line 1, in <module>
    cattrs.structure({"name": "Garfield", "age": 45}, Cat)
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../lib/python3.13/site-packages/cattrs/converters.py", line 558, in structure
    return self._structure_func.dispatch(cl)(obj, cl)
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^
  File ".../lib/python3.13/site-packages/cattrs/fns.py", line 22, in raise_error
    raise StructureHandlerNotFoundError(msg, type_=cl)
cattrs.errors.StructureHandlerNotFoundError: Unsupported type: <class '__main__.Cat'>. Register a structure hook for it.

It turns out you need to use attrs or dataclasses to make this into an object that can be converted into:

>>> from dataclasses import dataclass
>>> @dataclass
... class Cat:
...     name: str
...     age: int
...
>>> cattrs.structure({"name": "Garfield", "age": 45}, Cat)
Cat(name='Garfield', age=45)

castfit can handle the dataclass approach too, but for the undecorated version it just tries constructing an empty object and setting a bunch of attributes on it. Not sure if this is a reasonable assumption (probably isn't), but it reduces the number of required imports.

Overall, though, I'm pretty impressed by cattrs.

castfit code review

Tue Apr 29 2025 13:23:35 GMT-0400 (Eastern Daylight Time)

I asked both Gemini 2.5 Pro Experimental and ChatGPT o3 to review the code in castfit 0.1.1. I attached the source files for __init__.py, the two test files, and README.md together with the following prompt:

Please review the attached code and produce a comprehensive list of improvements. For any error or edge case, please produce sample code that would trigger the error/edge case. For any improvement, indicate the strength of the improvement. Improvements may include, among other things, performance, documentation, marketing, or ease-of-use.

While both found many similar issues, Gemini's output was both easier to convert to issues (although I should have been more explicit in the prompt) and it also exposed me to the great cattrs library which I should explore more.

I'll implement as many of these changes as I can and then see how the reviews change.

Note: I'm saving these transcripts to a private gist which is not indexed by search engines to try to reduce AI slop.

castfit 0.1.1

Tue Apr 29 2025 13:14:06 GMT-0400 (Eastern Daylight Time)

castfit 0.1.1 is available. castfit is my tiny library for converting things (mostly dict of keys) into classes using standard python type hints.

To install castfit:

# modern (recommended)
uv add castfit

# classic
python -m pip install castfit

Release Notes 0.1.1 - 2025-04-28T14:32:37Z #

Fixed

#1: type hints for castfit

Changed

metaist/LTS#8: using latest github actions

Class-based docopt usage strings

Tue Apr 29 2025 12:23:29 GMT-0400 (Eastern Daylight Time)

Previously: Access type hints in python, castfit 0.1.0

attrbox is my little library for loading configuration files, reading environment variables, and parsing command-line arguments. I reach for it every time I make a little CLI or throwaway script so that I can write out the usage and quickly get everything up and running.

One problem with using a docopt-based approach is that the result isn't typed properly. I created castfit exactly for this purpose: just define the class with the correct types for the fields and a usage string and you're good to go.

But there are a few annoyances:

You have to import two things (attrbox and castfit).
Now you have to synchronize the usage string with the class members (which are far away).
If you want to document the members, you're repeating yourself three times (usage string, member, and member docstring).

I keep wanting to move the documentation closer to the class members and then dynamically build up the usage string. But this introduces new problems:

You need to use ast to find the floating string expressions that are near the class members. So this won't work in, e.g., the REPL.
You need to format them some way. There's a tension between how you document a class member normally and how you do it when it's part of the docopt usage string.

And now your usage string isn't as clean as it was when you started.

I have a working version that I'll push out soon. We'll see if I ever reach for it.

attrbox 0.1.6

Mon Apr 21 2025 18:55:29 GMT-0400 (Eastern Daylight Time)

attrbox 0.1.6 is available. This release vendorizes docopt-ng, a more up-to-date version of the popular docopt library that used to be a dependency.

To install attrbox:

python -m pip install attrbox

# or, if you use uv (recommended):
uv add attrbox

Combined Release Notes 0.1.3, 0.1.4, 0.1.5, 0.1.6 - 2025-04-17T21:49:10Z #

Fixed

Versions 0.1.4 and 0.1.5 were zero-change version bumps due to errors in how GitHub Actions publish to PyPI.

#11: add py.typed

Changed

#13: vendorized docopt-ng
#14, #16, metaist/LTS#8: using latest project configurations
metaist/LTS#4: switched to using ds
metaist/LTS#6: copyright year
metaist/LTS#8: using latest github actions

Added

#15: cog to docs process

Removed

metaist/LTS#5: python 3.9 support

Cropping and Merging PDFs

Tue Apr 15 2025 22:39:05 GMT-0400 (Eastern Daylight Time)

Here's a simple-sounding task: Crop the vocal and piano parts of a full orchestral score PDF (many pages) into a new condensed score (fewer pages). After talking it over with Isaac Selya we discovered that, aside from the first page, both staves were in the same vertical position on every page.

I thought ChatGPT would nail this using pypdf which was its first attempt. It couldn't find the correct coordinates for the cropping, but it gave me enough code that I could manually supply the coordinates. However, it was only able to crop each page into separate smaller pages, but not stack them on a new page. Turns out:

Just because content is no longer visible, it is not gone. Cropping works by adjusting the viewbox. That means content that was cropped away can still be restored.

That means any time I tried putting a "cropped" part of a page on a blank page, I'd get the whole page (with some vertical translation). I tried so many variations on the crop (e.g., crop to small pages in a new PDF, then try stacking those pages), but just kept getting lots of pages of sheet music overlaid on top of each other.

Finally, ChatGPT mentioned that pymupdf might actually crop PDFs in a way that destroys the content outside of view. The API made a lot of sense to me, especially the more normal X-Y coordinate system with an origin in the top-left (unlike other PDF libraries that use the bottom-left).

That got me this working result which cuts vertical rectangles out of pages and glues them together on new pages. The only hard-coded bit was the extra margin on the first page to make sure that the vocal and piano pieces stay together. If I ever have to do this again, I'll probably add it as a feature to my pdfmerge tool.

#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.11"
# dependencies = [
#     "attrbox",
#     "pymupdf",
# ]
# ///
"""
Usage: ,crop-pdf.py
    [--help|--version][--debug]
    <input> [--output PATH]
    <rules>...

Options:
    -h, --help              show this message
    --version               show program version
    --debug                 show debug messages

    <input>                 input path
    -o PATH, --output PATH  output path
    <rules>                 rules

Example:
,crop-pdf.py scan.pdf "1: 0-80, 171-252, 666-764; 2-: 150-240, 666-764"
"""
# std
from __future__ import annotations
from pathlib import Path
from dataclasses import dataclass
from dataclasses import field

# lib
from attrbox import parse_docopt
import pymupdf


__version__ = "1.0.0"


@dataclass
class Ranges:
    ranges: list[range] = field(default_factory=list)

    def __contains__(self, other: int) -> bool:
        return any(other in r for r in self.ranges)

    def __iter__(self):
        for r in self.ranges:
            yield from r

    def as_list(self) -> list[tuple[int, int]]:
        return [(r.start, r.stop) for r in self.ranges]

    @staticmethod
    def parse(s: str, beg: int = 1, end: int = 1000, inclusive: bool = False) -> Ranges:
        result = Ranges()
        for rule in s.split(","):
            if "-" in rule:
                left, right = rule.split("-")
                left = int(left) if left else beg
                right = int(right) if right else end
            else:
                left = int(rule)
                right = left
            result.ranges.append(range(left, right + (0 if inclusive else 1)))
        return result


@dataclass
class Rule:
    pages: Ranges
    parts: list[tuple[int, int]] = field(default_factory=list)

    def __contains__(self, other: int) -> bool:
        return other in self.pages

    def __iter__(self):
        for page in self.pages:
            for beg, end in self.parts:
                yield (page, beg, end)

    @staticmethod
    def parse(s: str, end: int) -> Rule:
        pages, parts = s.split(":")
        return Rule(
            Ranges.parse(pages, 1, end),
            Ranges.parse(parts, inclusive=True).as_list(),
        )


def main():
    args = parse_docopt(__doc__, version=__version__)
    args.input = Path(args.input)
    if not args.output:
        args.output = Path(args.input).with_stem(args.input.stem + "_crop")
    else:
        args.output = Path(args.output)

    if args.output.exists():
        args.output.unlink()

    if args.debug:
        print(args)

    doc = pymupdf.open(args.input)
    out = pymupdf.open()

    end = len(doc)
    rules: list[Rule] = []
    for rule in args.rules:
        rules.extend(Rule.parse(s, end) for s in rule.split(";"))
    if args.debug:
        print(rules)

    margin = 20
    dest = None
    y = margin
    out_page_n = 0
    for rule in rules:
        for idx, beg, end in rule:
            idx -= 1

            page = doc[idx]
            width = page.rect.width
            height = page.rect.height
            h = end - beg

            if (
                not dest
                or y + h > height - margin
                or (out_page_n == 1 and y + h > height - 100)
            ):
                print("new page")
                dest = out.new_page(-1, width=width, height=height)
                out_page_n += 1
                y = margin

            clip = pymupdf.Rect(0, beg, width, end)
            post = pymupdf.Rect(0, y, width, y + h)
            dest.show_pdf_page(post, doc, pno=idx, clip=clip)
            print(idx, beg, end)

            y += h

    out.ez_save(args.output)


if __name__ == "__main__":
    main()

TIL: contextvars in python

Mon Apr 07 2025 20:31:42 GMT-0400 (Eastern Daylight Time)

While I was working on my third attempt of a python version of idempotent-bash, ChatGPT suggested I use contextvars which has apparently been part of the standard library since python 3.7. This is different from contextlib which I know and like.

Let's start with a simple (imprecise) definition: A ContextVar is like a global variable except that it can be reset to earlier values.

Why would you want this? The situation where I found it helpful was to avoid having to track and pass the same variable into every function I was writing.

from __future__ import annotations
from dataclasses import dataclass
from contextvars import ContextVar


_current: ContextVar[IdemPy | None] = ContextVar("_current", default=None)

@dataclass
class IdemPy:

    # parameters for how functions should behave
    dry: bool = False
    force: bool = False
    keep_running: bool = False
    show_progress: bool = True

    def __enter__(self) -> IdemPy:
        self._token = _current.set(self)
        return self

    def __exit__(self, exc_type, exc_val, exc_tb):
        _current.reset(self._token)

    @classmethod
    def get(cls, default: IdemPy | None = None) -> IdemPy:
        current = _current.get() or default
        if not current:
          current = IdemPy()
          _current.set(current)
        return current

# elsewhere
def my_function() -> None:
    idempy = IdemPy.get():
    ... # continue with the correct context


my_function() # creates a default IdemPy context
with IdemPy():
    my_function() # similar to above

with IdemPy(force=True):
    my_function() # slightly different IdemPy context

There are two alternatives that I have tried in the past:

Add idempy as an argument to my_function. It works at the cost of having to pass that argument around into every function. And for a class where you almost always want the default version, that feels very expensive.
Wrap my_function in a class which, when constructed, takes a class of IdemPy. This is also very heavy and reminds me most of Java. Let construct the builder which will instantiate the initiator.... No.

Most of the time I just want to call my_function and have it do the right thing. Occasionally, I want to more explicitly set the context using with IdemPy(...) which overrides what IdemPy.get() returns by changing the ContextVar. I can even nest with IdemPy(): calls and it will do the right thing.

While all the advice about global variables is reasonable, this is one of those cases where I actually think python's with statement works incredibly well.

Is it like a stack? #

Update 2025-04-08: Shalev NessAiver asked me whether ContextVar acts like a stack and whether you can pop and push to the same token's value.

Nope. First, unlike a stack you reset the values in a different order than you set them.

from contextvars import ContextVar
v = ContextVar('v', default=0)
token1 = v.set(10)
token2 = v.set(20)
token3 = v.set(30)

v.reset(token2)
assert v.get() == 10

v.reset(token3)
assert v.get() == 20

v.reset(token1)
assert v.get() == 0

Second, you'll get an error if you try to reset the same token twice.

from contextvars import ContextVar
v = ContextVar('v', default=0)
token = v.set(10)
v.reset(token)
assert v.get() == 0

v.reset(token) # raises RuntimeError: <Token ...> has already been used once

Updates #

2025-04-08: added Shalev NessAiver's question

Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📖 Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy by Cathy O'Neil (2016; via EdSurge). Glad the book includes some positive examples. Unfortunately, because there are more chaotic states than ordered states there are more poorly considered uses of mathematical modeling than excellent uses.

We've been wrong about math for 2300 years

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 We've been wrong about math for 2300 years (David Bessis; via Jeremy Wertheimer).

Two Americas, one bank branch, and $50,000 cash

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Two Americas, one bank branch, and $50,000 cash (Patrick McKenzie). Imagine Patrick McKenzie as Columbo for this one. Just took a random bit of a description of a bank branch and went full deep-dive wild on trying to figure out if a story was legit.

This is Not A Game About A Pipe

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 This is Not A Game About A Pipe (Mac McAnally). If Metaist ever published card games, they'd probably all end up like this.

Think Like a Freak: The Authors of Freakonomics Offer to Retrain Your Brain

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📖 Think Like a Freak: The Authors of Freakonomics Offer to Retrain Your Brain by Steven D. Levitt & Stephen J. Dubner (2014). Favorite part is when they revealed how their previous book was used to create a honeypot for would-be terrorists.

The Manus Marketing Madness

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 The Manus Marketing Madness (Zvi Mowshowitz). 'twas madness indeed.

The 5 Most Difficult Employees (And How To Actually Handle Them)

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 The 5 Most Difficult Employees (And How To Actually Handle Them) (Claire Lew). "Most difficult" is probably stretching it, but, yes, all these archetypes exist.

Shape of AI

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Shape of AI (Emily Campbell). Nice collection of AI interaction patterns.

“I scratched my own itch” isn’t good enough

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 “I scratched my own itch” isn’t good enough (Jason Cohen). And a point-by-point breakdown of the implications.

How to simplify complex decisions by cleaving the facts

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 How to simplify complex decisions by cleaving the facts (Jason Cohen). Self-recommending.

Here’s how I use LLMs to help me write code

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Here’s how I use LLMs to help me write code (Simon Willison). Excellent explanation for what it takes to actually get good results from LLMs when generating code.

Great software design looks underwhelming

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Great software design looks underwhelming (Sean Goedecke). Unsophisticated & battle tested is certainly a reasonable place to start from.

Discoverability of AI Features: Learn from Amazon’s Mistakes

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Discoverability of AI Features: Learn from Amazon’s Mistakes (Kate Moran & Feifei Liu). Did you ever use Amazon Rufus for anything? Did you know Amazon Rufus was a thing? What kind of thing do you think it is?

Dear Student: Yes, AI is here, you're screwed unless you take action...

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Dear Student: Yes, AI is here, you're screwed unless you take action... (Geoffrey Huntley). The initial story from the student is a good contrast to Simon's article above. Tons of advice, but kinda all over the place.

Childhood and Education #9: School is Hell

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 Childhood and Education #9: School is Hell (Zvi Mowshowitz). A mostly depressing, but sometimes funny, compendium of links about education.

Causal Emergence 2.0: Quantifying emergent complexity

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📄 Causal Emergence 2.0: Quantifying emergent complexity by Erik Hoel. Interesting update; the blog post gives a good summary.

Building a Life Worth Living

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📖 Building a Life Worth Living by Marsha M. Linehan (2020; via Jacqueline Novikov). The origin story of Dialectical Behavior Therapy (DBT).

Build: An Unorthodox Guide to Making Things Worth Making

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📖 Build: An Unorthodox Guide to Making Things Worth Making by Tony Fadell (2022; via Tim Ferriss & Lex Fridman). Lots of interesting stories and, occasionally, some reasonable advice.

An essay in which my friend feels stuck and I suggest relaxing some constraints

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 An essay in which my friend feels stuck and I suggest relaxing some constraints (Henrik Karlsson). It is hard to become aware that many constraints are imagined.

A Post Mortem on the Gino Case

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 A Post Mortem on the Gino Case (Zoé Ziani; via Andrew Gelman). Somehow the story just gets worse.

A Field Guide to Rapidly Improving AI Products

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 A Field Guide to Rapidly Improving AI Products (Hamel Husain). "Can you show me how you’re measuring if any of this actually works?"

A Category Theory-Inspired BBN

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 A Category Theory-Inspired BBN (Eric Gilliam). Interview about Topos Institute with a goal of "building technologies to enable scientists and engineers from adjacent fields to more effectively combine and reason with overlapping models from distinct domains."

50 Thoughts on DOGE

Mon Mar 31 2025 18:36:00 GMT-0400 (Eastern Daylight Time)

📝 50 Thoughts on DOGE (Santi Ruiz). They are certainly thoughts.

Trying Caddy

Tue Mar 25 2025 13:32:19 GMT-0400 (Eastern Daylight Time)

I accidentally blew away my server (with several websites), so today was a good a day as any to try caddy.

On Ubuntu, I was able to get a server up and running quickly.

sudo apt install -y debian-keyring debian-archive-keyring apt-transport-https curl
curl -1sLf 'https://dl.cloudsmith.io/public/caddy/stable/gpg.key' | sudo gpg --dearmor -o /usr/share/keyrings/caddy-stable-archive-keyring.gpg
curl -1sLf 'https://dl.cloudsmith.io/public/caddy/stable/debian.deb.txt' | sudo tee /etc/apt/sources.list.d/caddy-stable.list
sudo apt update
sudo apt install caddy

Then I updated /etc/caddy/Caddyfile:

import *.caddy

And then created a bunch of .caddy files for each of the websites. I have three kinds of sites I support: simple static HTML sites, legacy PHP (usually WordPress), and python (uWSGI).

Static sites are super simple with Caddy:

example.org {
	root * /var/www/example.org/site
	file_server
}

PHP isn't too bad either:

example.com {
	root * /var/www/example.com/site
	php_fastcgi unix//run/php/php8.1-fpm.sock
	file_server
}

uWSGI took some playing around, but is even simpler once you realize that you can't use /tmp for security reasons and you need to use http-socket instead of socket:

example.com {
	reverse_proxy * unix//run/uwsgi/example.sock
}

The only other major feature I thought I needed was URL rewriting which looked really easy. It actually took me more time to realize when I didn't need it. For example, on this blog you can get the RSS feed from /blog/feed which, previously, redirected to /blog/feed/index.xml. In Caddy you can just say:

	file_server {
		index index.html index.xml
	}

The other place I rewrite is to go from / to /blog/index.html (until such day that I have a better cover page for this site). I tried a few things ChatGPT suggested like:

  rewrite / /blog/index.html # => 404
  rewrite ^/$ /blog/index.html # => 404

But only it's final suggestion worked:

	@root {
		path /
	}
	rewrite @root /blog/index.html

This created a named matcher @root that matches the exact path of /.

I highly recommend reading through the concepts and skimming the list of directives.

Caddy is looking like a drop-in replacement for nginx + certbot.

Widowhood

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Widowhood (Amy Satterthwaite Pappas / Pie & Chai). I should have realized that "Chai" here means tea and not the Hebrew word for "life". Still interesting.

VALIS

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 VALIS by Philip K. Dick (2011; via Jeremy Wertheimer). Randomly autobiographical with a schizophrenic version of Philip K. Dick. Satellites beaming laser images into people's brains. Par for the course for this sort of thing.

Unbundling the University

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Unbundling the University (Ben Reinhardt). A very long discussion about the history of how tech transfer came into the university and why we should create new types of labs.

TypeScript types can run DOOM

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

▶️ TypeScript types can run DOOM (Michigan TypeScript; via Simon Willison). Because of course it can. Amazing video.

Training Small R1-like Reasoning Models

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Training Small R1-like Reasoning Models (Stephen Diehl). Fascinating.

The Truth About The Moon Landings

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

▶️ The Truth About The Moon Landings (Everyday Astronaut). Long, but incredible. Careful walkthrough of lots of details.

The Startup Drake Equation

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 The Startup Drake Equation (Jason Cohen). Seems to have more factors than the original Drake Equation, but fair points nonetheless.

The Art of the Start 2.0

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 The Art of the Start 2.0 by Guy Kawasaki (via Iridescent Learning). Some nice exercises, but not as exciting as I thought.

solve for distribution

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 solve for distribution (Visakan Veerasamy). Basically means figure out marketing first.

Slashing my ".bashrc" in half

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Slashing my ".bashrc" in half (Bite Code!). I've got a bunch of tabs open to check out some of these tools.

Reasons of State: Why Didn't Denmark Sell Greenland?

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Reasons of State: Why Didn't Denmark Sell Greenland? (Gwern Branwen). Why'd you leave the keys upon the table?

On Emergent Misalignment

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 On Emergent Misalignment (Zvi Mowshowitz). Some of the best coverage of one of the more interesting LLM training papers to drop this past week.

Newton and the Counterfeiter: The Unknown Detective Career of the World's Greatest Scientist

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 Newton and the Counterfeiter: The Unknown Detective Career of the World's Greatest Scientist by Thomas Levenson (2010). Why isn't there a movie about this!? Action-packed (warning: book starts off slow), crazy borderline unethical interrogations, law enforcement becoming criminals, criminals becoming informants, a brilliant nemesis, exacting calculations, fierce guarding of coinage. Now that we have the historical account, you could do a whole historical fictional series of about this.

New Proofs Probe the Limits of Mathematical Truth

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 New Proofs Probe the Limits of Mathematical Truth (Joseph Howlett / Quanta). Consistently impressed by Quanta Magazine's ability to give accessible history of these mathematical ideas.

Never Eat Alone, Expanded and Updated

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 Never Eat Alone, Expanded and Updated by Keith Ferrazzi & Tahl Raz (2014). Some interesting stories, I guess.

My Promised Land: The Triumph and Tragedy of Israel

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 My Promised Land: The Triumph and Tragedy of Israel by Ari Shavit (2013; via Jeremy Wertheimer). This was a hard read. I expect a future version will have much to say about Oct 7, 2023.

My LLM codegen workflow atm

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 My LLM codegen workflow atm (Harper Reed; via Simon Willison). Makes sense. (Tangent: Can't believe I've been following Simon Willison's llm tool for many months and it wasn't until Wednesday that I actually installed it. Having an LLM available in the command line is a lot of fun.)

In Search of Excellence: Lessons from America's Best-Run Companies

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 In Search of Excellence: Lessons from America's Best-Run Companies by Tom Peters & Robert H. Waterman (2012; via Reed Hastings & Reid Hoffman). Originally published in 1982, many of the examples no longer hold up, except for maybe McDonalds & Disney. Most of the "be authentic" advice has become a hallmark of random holiday-card-from-your-dentist-spam.

How to write a good design document

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 How to write a good design document (Grant Slatton). When I saw the headline, I thought this was about UX, but it's about design documents for dealing with technical constraints. The writing advice is actually good for many forms of writing.

How Core Git Developers Configure Git

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 How Core Git Developers Configure Git (Scott Chacon; via Patrick Kage). I like that they made all the core developers re-experience the defaults.

Halakhic Positions of Rabbi Joseph B. Soloveitchik: Volume 5

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 Halakhic Positions of Rabbi Joseph B. Soloveitchik: Volume 5 by Aharon Ziegler (1998). Episode 5: The Rav Strikes Back.

Getting a handle on your goals and big life changes

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

🎙️ Getting a handle on your goals and big life changes (Spencer Greenberg & Tee Barnett). There were bits of "manifesting" that I couldn't totally stomach, but the discussion about burnout was quite good.

Funding science is actually a badass thing to do

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Funding science is actually a badass thing to do (Adam Mastroianni). Yep.

Eight Sleep mattress found out to include an SSH backdoor, allowing access to the user's entire network

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Eight Sleep mattress found out to include an SSH backdoor, allowing access to the user's entire network (ToasterPC / Linus Tech Tips). Forget hoverboards, in this future we can ssh into our mattress covers.

disabling cert checks: we have not learned much

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 disabling cert checks: we have not learned much (Daniel Stenberg). You see, it's easier to not.

Destined For War: Can America and China Escape Thucydides's Trap?

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 Destined For War: Can America and China Escape Thucydides's Trap? by Graham Allison (via Shalev NessAiver). Spoiler: Probably not.

Crossing the Chasm, 3rd Edition: Marketing and Selling Disruptive Products to Mainstream Customers

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📖 Crossing the Chasm, 3rd Edition: Marketing and Selling Disruptive Products to Mainstream Customers by Geoffrey A. Moore (2014; via Drew Houston). I wonder if the 4th edition will address apps.

Automating Math

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 Automating Math (Adam Marblestone). I've been learning more about formal methods.

Algorithms are breaking how we think

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

▶️ Algorithms are breaking how we think (Technology Connections). Illustrative example of how our search behavior has changed.

AI software I'm surprised doesn't exist yet

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 AI software I'm surprised doesn't exist yet (Grant Slatton). Nice list. If I ever do another "30 apps 30 days" kind of thing, these could serve as an inspiration.

A year of uv: pros, cons, and should you migrate

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 A year of uv: pros, cons, and should you migrate (Bite Code!). My take: Yes, you should.

A Defense of Weird Research

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 A Defense of Weird Research (Deena Mousa & Lauren Gilbert / Asterisk).

50 Years of Travel Tips

Fri Feb 28 2025 11:07:00 GMT-0500 (Eastern Standard Time)

📝 50 Years of Travel Tips (Kevin Kelly; via Tyler Cowen). Feels well-earned.

Your Face Belongs to Us