Twenty-one years later, that problem has been solved, and then some.

Today, we don’t have a visibility problem. We have a signal problem.

The Israeli tech ecosystem has matured into a global powerhouse, but with that success comes complexity. The 2026 landscape is more fragmented and fast-moving than ever. Between the rise of agentic AI infrastructure, the emergence of sovereign tech stacks, and a macro environment that demands extreme capital efficiency, founders are navigating a much narrower and noisier path to funding.

And yet, the tools for navigating that path haven’t evolved at the same pace.

Founders still spend countless hours stitching together outdated spreadsheets, jumping between fragmented databases, or pitching funds that haven’t deployed capital in years. Fundraising is already a full-time job; figuring out who to talk to shouldn’t be.

That’s why I built IsraelVC.com.

The “Why”

I bought the Israelvc dot com domain in November 2007 (that’s pretty OG if you ask me). It was parked for years and then just pointed at VC Cafe. Many times I toyed with the idea of launching it but it was always put aside for lack of bandwidth.

As a GP at Remagine Ventures, I sit on both sides of the table. I see how much time founders waste trying to map the ecosystem, and how often investors are misrepresented or simply outdated in existing databases. At the same time, the rise of Agentic AI coding meant that I can just ‘play’ with it and finally bring my ideas to life. Many hours of going in the rabbit hole later and Israel VC v1 is now live.

IsraelVC is my attempt to reduce that friction with a few simple principles:

• Curated, not just crawled
  This isn’t another scraped database. It’s a focused, opinionated list of firms that are actually active today, VCs, CVCs, and micro-VCs who are deploying and engaging. There’s still some data clean up taking place, so if a fund is missing or defunct, you can make the site better by adding the missing information.
• No fluff
  The goal isn’t to overwhelm with data. It’s to help founders quickly find the right investors. Clean interface, fast search, no unnecessary noise.
• Accessible
  No paywalls, no premium tiers, no gatekeeping. This is a tool built for the community, full stop.

Connecting the Dots

IsraelVC doesn’t exist in isolation, it’s part of a broader effort I’ve been building for years.

• VC Cafe has always been about storytelling and surfacing the people and ideas shaping the ecosystem. It has become a longer read, editorial outlet for my thoughts and what we care about as a fund at Remagine Ventures.
• #Firgun is about amplifying what’s working – a weekly newsletter, highlighting founders, rounds, and momentum across the Israeli startup community as well as the wider tech world news.
• Remagine Ventures is where I engage directly, backing the next generation of founders building the future of human experience.

IsraelVC sits at the intersection of all three.

It’s the infrastructure layer: a practical tool that complements the storytelling of VC Cafe and the amplification of #Firgun, while reflecting the real-time market perspective I see at Remagine. I need to see how I practically bring them together more.

Where This Is Going

This is very much a V1, intentionally simple, but built to evolve.

In the coming months, I plan to expand it in a few key directions:

• Deeper filtering
  Sector-specific views (AI infrastructure, fintech, cyber) that reflect how focused the market has become.
• Founder resources
  Integrating guides, insights, and curated content from VC Cafe and #Firgun to make the platform more than just a directory.
• Community-driven updates
  Keeping the data fresh and accurate through direct input from funds and the broader ecosystem.

The ambition is straightforward: to make the Israeli venture ecosystem the most transparent and easiest to navigate in the world.

Final Thought

If you’re a founder, my hope is simple: this saves you a week—maybe more—of research and helps you focus on what actually matters: building.

If you’re an investor, take a look and make sure you’re represented the way you want to be, because founders will be using this.

Explore the directory: israelvc.com and never stop learning!

As an early-stage VC, I am seeing this reality play out in private markets: the days of pure software moats are coming to an end, and hardware is finally investable again. Recent AI breakthroughs, such as Claude Code’s ability to nearly autonomously build and run software, have rattled public markets and wiped billions off legacy software valuations. While not immediate, it feels like we’re seeing the venture market swinging from bits to atoms.

The Death of the Software Moat

Software was historically incredibly valuable because it was scarce. Engineering talent was a bottleneck, and technical know-how was hard to acquire. Today, AI coding assistants and open-source frameworks have commoditised code. If a model can write code or design a product overnight, software simply stops being scarce. Even the mighty Google recently mentioned that 75% of its new code is written by AI.

As a result, what once made software startups attractive, their code and the engineers who built it, no longer carries the same value. In fact, it is estimated that any enterprise software company built over the last five to ten years could be rebuilt in six months or less. The user interface (UI) moat is also collapsing to zero; when AI agents can mediate workflows directly and execute tasks, the friction of learning a UI disappears, and software loses its visual differentiation.

The Migration to “Atoms Over Bits”

With software abundant, venture capital is aggressively hunting for new scarcity. This scarcity is migrating downstream toward the systems that power intelligence and the physical infrastructure that AI cannot easily replicate: datacentres (which drove much of the US GDP growth in 2025), robotics, energy, space, defence, and advanced manufacturing.

This macro investment shift toward “atoms over bits” is already showing up in the numbers. Over the last three years, deeptech and hardware investment has grown its VC market share globally by approximately 40%. While enterprise software deals have slowed, funding for AI hardware, defense, and automation is seeing massive rounds.

Why Hardware is the Ultimate Defensible Moat

Investors have historically avoided hardware because of legacy scars – it was viewed it as too slow, capital-intensive, and risky. But hardware offers something software cannot: presence, permanence, and true defensibility.

• Political and Logistical Switching Costs: When your hardware is literally bolted into a city’s physical infrastructure or a factory floor, the switching cost is immense. It is not just a technological hurdle; it is a financial and logistical barrier that pure software rarely provides.
• Physical Constraints Create Monopolies: Real-world constraints offer durable competitive positions. You cannot simply use an LLM to generate a physical weapons system, a satellite network, or a cleared facility. Examples in Defence tech include Anduril, Shield AI, and Chaos Industries, which are securing massive funding rounds and practically building durable monopolies based on sovereign demand and physical constraints. In space tech, Stoke Space and Astranis are creating an infrastructure scarcity moat that software commoditisation cannot erode.
• Giving AI a “Body”: We have essentially built a digital god and trapped it in a text box. For AI to become infinitely more useful, it needs a physical body to collect passive, ambient context from the real world. Whether it’s glasses collecting hand-motion data to train factory robots or wearables tracking biomarkers, the AI ghost needs a physical shell, and typing into a laptop is not the final form. Many robotics companies are raising mega rounds of $100M+. One of the latest examples is Mind Robotics, which raised a $500M Series A in April 2026. It’s a spin-out from EV maker Rivian, this startup focuses on manufacturing-specific robotics, or Robot Era, A Beijing-based humanoid robotics company that joined the unicorn club this year after raising $145M in March 2026, focusing on general-purpose robots for industrial use.

The New VC Playbook

Many hardware startup founders are familiar with the saying: “Hardware is Hard”. And investors have internalised this all too well. Compared to SaaS companies, hardware is ‘further from the money’ and closer to deep tech in monetisation. Beloved metrics that VCs use to assess companies like Customer Acquisition Cost (CAC) and rapid Monthly Recurring Revenue (MRR) growth do not map cleanly to hardware companies. Hardware requires deep design work, longer timelines, supply chain risks and a tolerance for higher upfront capital. It’s a whole new ball game for non-hardware specialists.

However, the assumption that hardware is a bad business model is an outdated vestige. Thanks to global contract manufacturing, advances in rapid prototyping, and recurring-revenue software models layered on top of physical devices, hardware economics have been completely reshaped. Today, a base unit of hardware acts as a permanent sales rep on the ground, enabling continuous software upsells and renewals.

VC investors can no longer afford to dismiss hardware out of habit. The era of building yet another app is ending. The next generation of enduring tech giants will be those that establish their moats not with code, but with silicon, steel, and infrastructure. We’ve yet to do a hardware investment at Remagine Ventures to date, but we’re certainly considering it and started reviewing opportunities. From sensors to camera related tech. If you’re an Israeli founder in the earliest stages building hardware that can be empowered by AI, we’d love to speak with you.

The Summer 2026 edition is characterised by AI is moving from copilots to agents, from software to services, from chat interfaces to company operating systems, and from digital workflows into chips, supply chains, agriculture, medicine, defence, and space tech.

1. Y Combinator Summer 2026 RFS: rebuilding software, services, silicon and the physical world

YC frames Summer 2026 around AI becoming “the foundation,” not just a feature, with startups rebuilding software, services, silicon and pushing AI into the physical world.

• 1.1 AI-native service companies – Sell the completed work, not the tool. YC highlights insurance brokerage, accounting, tax, audit, compliance and healthcare administration as areas where AI-native companies can replace outsourced services rather than merely improve SaaS workflows.
• 1.2 Company Brain and the AI Operating System for Companies – The missing layer is no longer model capability, but company context. YC describes the “Company Brain” as a living map of how a company works, while the AI OS for companies turns company artifacts into a closed loop that agents can reason over and act on.
• 1.3 Software for Agents– YC’s line is strong: “The next trillion users on the internet won’t be people, they’ll be AI agents.” This creates opportunities around APIs, MCPs, CLIs, machine-readable documentation, identity, permissions, payments and agent-native software.
• 1.4 SaaS Challengers – Jared Friedman’s point is that AI coding may hurt incumbents, but it is a gift to startups. YC wants founders to attack big, hard SaaS categories like ERP, chip design software, industrial control systems and supply chain management.
• 1.5 Dynamic Software Interfaces – AI makes users their own forward-deployed engineers. Instead of every user seeing the same interface, software may become a set of primitives that users and agents customise for their own workflows.
• 1.6 AI for Low-Pesticide Agriculture– YC sees AI vision, robotics, sensors and biology converging to reduce chemical use in agriculture while improving yield. This is a strong “AI meets atoms” category, especially relevant for Israeli agtech, robotics and computer vision founders.
• 1.7 AI Personalised Medicine– YC points to agents analysing diagnostics, genome scans, EHR data and wearables, alongside falling costs of diagnostics and n-of-1 genetic therapies. This is a huge opportunity, but one where clinical validation and trust matter more than speed alone.
• 1.8 Counter-Swarm Defence– YC argues that drone defense is starting to look less like traditional weapons procurement and more like a real-time distributed system, with winners looking “more like Cloudflare than Raytheon.”
• 1.9 Hardware Supply Chain and Semiconductor Supply Chain 2.0 – YC highlights faster hardware iteration loops and deeper visibility into semiconductor supply chains, from packaging constraints to export compliance and multi-tier supplier risk.
• 1.10 Inference Chips for Agent Workflows and Electronics in Space – Agents create bursty, looping, memory-heavy inference workloads. YC also points to demand for inference chips in space, optimised for mass, thermal performance and radiation.
• 1.11 Selling to Huge Companies – YC argues AI has made it newly possible for 2–3 person teams to sell useful products to Fortune 10-scale companies much earlier than before.

2. a16z speedrun: agents, networks and the 100x solo founder

A16Z Speedrun has started recruiting for its 007 batch, and I expect more requests to be released as we get closer to the deadline. We can already see some interesting themes coming up.

2.1 Come for the Agent, Stay for the Network

Troy Kirwin from a16z speedrun framed one of the most interesting vertical AI opportunities: AI procurement agents. The example is an HVAC technician who needs a replacement part. Instead of calling vendors, emailing for quotes, waiting days, and comparing PDFs, an AI agent identifies the SKU, contacts suppliers, negotiates price, and orders in minutes. But the real moat is the network. Once the agent operates across thousands of buyers, it sees real transaction prices, aggregates demand, and can make suppliers compete to be part of the network.

This is the difference between an agent and a marketplace.

Promising verticals include:

• Freight and logistics
• Agricultural inputs
• Field services
• Food service procurement
• Construction subcontracting
• Industrial MRO
• Healthcare staffing

The wedge is automation. The moat is pricing data, supplier access, demand aggregation, and workflow ownership.

2.2 Tools for the 100x Solo Founder

Emily Bennett from a16z speedrun described another important theme: tools that let one founder operate what previously required entire teams. The examples include an AI-native finance stack, an autonomous revenue engine, and an AI-operated supply chain for SMBs. The defining feature is control: these systems do not just recommend actions, they access operational surfaces like bank accounts, CRMs, inboxes, ad platforms, and inventory systems, then execute, measure, and improve.

This is not just “productivity software.” It is company automation.

If AI reduces coordination costs, the minimum viable company shrinks. A founder can do more with fewer people. A small team can look larger. A startup can reach revenue before building a traditional org chart.

Startup ideas:

• AI CFO for solo founders and micro-SMBs
• Autonomous outbound and renewal engine
• AI customer support operator with authority to resolve issues
• AI supply chain planner for ecommerce operators
• AI chief of staff that manages operating cadence, follow-ups, hiring, and reporting

This theme is especially relevant for founders building globally from smaller ecosystems. AI compresses the distance between a two-person team and a much larger competitor.

3. ARK Big Ideas 2026: the macro tailwinds behind the startup requests

ARK published its Big ideas 2026 research not a startup wishlist, but as the macro scaffolding that explains why these requests matter. ARK’s “Great Acceleration” argues that AI, public blockchains, robotics, energy storage and multiomics are becoming increasingly interdependent, with AI acting as the central catalyst across platforms.

• 3.1 The Great Acceleration – Use ARK to explain why this is not just an AI app cycle. AI is converging with robotics, energy, blockchain, biology and space. That supports YC’s interest in agriculture, defense, personalised medicine, hardware, chips and space.
• 3.2 AI Infrastructure – ARK estimates that data center systems investment reached roughly $500 billion in 2025 and could reach about $1.4 trillion by 2030. This supports the YC sections on inference chips, semiconductor supply chains, electronics in space and hardware iteration.
• 3.3 The AI Consumer Operating System– ARK argues that foundation models are becoming a new layer of the internet stack, changing search, discovery, transactions and ecommerce. This supports your sections on software for agents, agent commerce, agent-readable storefronts, GEO for agents and AI purchasing agents.
• 3.4 AI Productivity – ARK says AI agents became more proficient during 2025, with reliable task duration rising from 6 minutes to 31 minutes, while software development model costs fell sharply. This supports the sections on SaaS challengers, AI-native service companies, company brains and 100x solo founder tools.
• 3.5 Robotics, Multiomics and Space – ARK’s robotics and multiomics sections give you macro support for low-pesticide agriculture, personalised medicine, counter-swarm defense, humanoids, autonomous logistics and space-based compute.

4. My synthesis: what Remagine Ventures is looking for

This is where you bring it back to your voice.

The clean synthesis could be:

YC is telling founders where the pain is.
a16z speedrun is telling founders what the new company-building patterns look like.
ARK is telling investors why the platform shift is bigger than another SaaS cycle.

At Remagine Ventures, the most interesting opportunities sit at the intersection:

• Agents that become networks
• Company brains and AI operating systems
• Software built for agents, not humans
• AI-native services that sell outcomes
• Tools that let tiny teams operate like large companies
• AI in hard verticals: healthcare, defense, agriculture, supply chain and industrial workflows
• Infrastructure for trust, context, permissions, payments and execution

That firsthand experience is what made me sceptical of the loudest narrative in enterprise software right now: that AI agents have made SaaS obsolete. There’s a peculiar ritual in tech where every few years, we declare something dead before it’s had the chance to prove itself. SaaS killed on-premise software. Mobile killed the desktop web. Cloud killed everything else. And now, agentic AI and vibe coding are supposedly killing SaaS. The eulogies are being written fast, the memes are flying, and the term “SaaSpocalypse” has entered the VC lexicon.

Before we start ordering flowers, let’s slow down and actually look at what’s happening – including what’s genuinely exciting, what’s genuinely broken, and what should make any serious investor pause.

The Promise Is Real. So Are the Problems.

Let’s start with some honesty. Agentic AI – systems that don’t just respond to prompts but autonomously plan, execute multi-step tasks, use tools, and interact with the world – represents a genuine shift in how software gets built and used. The same is true of vibe coding, the workflow popularised by Andrej Karpathy in February 2025, where developers (and increasingly, non-developers) describe intent in plain language and let AI generate the code. These are not gimmicks.

But the gap between demo and production reality is wide, and it matters enormously to founders and investors who are building real businesses on top of this infrastructure.

1. Security: From Malware to Rogue Skills

The security picture for agentic AI is, frankly, alarming. Unlike a traditional chatbot that produces text, an agent has tool access, memory, and the ability to take real-world actions – which dramatically expands the attack surface. According to research published on arXiv, 94.4% of state-of-the-art LLM agents are vulnerable to prompt injection, 83.3% to retrieval-based backdoors, and 100% to inter-agent trust exploits.

The attacks are no longer theoretical. In mid-2025, a critical vulnerability dubbed EchoLeak was discovered in Microsoft Copilot – infected email messages containing engineered prompts could trigger Copilot to exfiltrate sensitive data automatically, without any user interaction. Meanwhile, community marketplaces for agent skills have become a new vector for supply-chain attacks: the “ClawHub” repository has been abused to distribute malicious packages disguised as trading bots, utilities, or development helpers – once installed, these skills can deploy information-stealing malware directly onto a user’s machine.

The problem compounds in multi-agent systems. Protocol-level threats target the communication standards agents use – attackers can exploit these to achieve impersonation, where one agent pretends to be another, or manipulate collusion and coordination, effectively turning a group of agents into an autonomous threat actor that can bypass security controls a single agent could not.

For enterprise adoption, this is a hard blocker, not a speed bump. 59% of organizations surveyed between December 2024 and January 2025 said implementing agentic AI in their cybersecurity operations was a “work in progress.” That number tells you how much trust still needs to be built.

2. Token Guzzling: the hidden cost of doing things

Even when agents behave perfectly, they are expensive. Complex agentic workflows consume tokens at every reasoning step, tool call, memory retrieval, and sub-task delegation. Chain a few agents together on a non-trivial task and the costs compound faster than most teams anticipate. I’ve experienced this firsthand: what looks like a simple multi-step workflow in a demo can quietly run up a bill that would make a CFO wince.

Context windows have grown dramatically, which is genuinely useful, but larger contexts mean more tokens processed per call. The economics that work for a prototype fall apart at production scale, particularly for startups trying to build margin into AI-native products. This is one of the less-discussed constraints on the agentic AI business model: the underlying inference costs are still high enough that many use cases are impressive but not yet profitable to run at scale.

3. Bugs, Hallucinations, and the Vibe Coding Hangover

Vibe coding – accepting AI-generated code without fully reading or understanding it – accelerates prototyping dramatically. But its costs downstream are becoming increasingly visible. Up to 45% of AI-generated code introduces known security vulnerabilities, according to the Veracode 2025 GenAI Code Security Report, with Java having a failure rate exceeding 70%.

Hallucinations in code are particularly insidious because they look correct. AI models hallucinate function names that don’t exist, reference APIs that were deprecated years ago, and invent configuration options that never existed – and these bugs are particularly dangerous because the code follows all the right patterns while calling functions that simply don’t work. Research found that among 576,000 code samples analysed, AI tools suggested 205,474 unique software packages that did not exist. Attackers have learned to monitor these hallucinations and register malicious packages with the suggested names on NPM or PyPI, turning a development mistake into a supply-chain attack.

If 2025 was the year everyone shipped faster, 2026 is the year many teams are discovering what they shipped. Fast Company reported on the “vibe coding hangover” as early as September 2025 – engineers inheriting AI-generated codebases they can’t extend, debug, or audit. One December 2025 analysis by CodeRabbit of 470 open-source pull requests found that code co-authored by AI contained 1.7x more “major” issues than human-written code.

4. Quality at Scale Is Still a Work in Progress

Beyond security, there’s a broader quality challenge. The top frustration among developers, cited by 66% in a Stack Overflow survey, is dealing with “AI solutions that are almost right, but not quite” – and almost-right code is worse than clearly wrong code, because wrong code fails tests immediately while almost-right code passes tests but fails in production.

The structural problem is that AI accelerates code generation but review capacity remains flat. These systems still struggle with long-horizon tasks, architectural coherence, and the kind of contextual judgment that distinguishes a prototype from a production system. As senior engineers would tell you: the boring parts – error handling, edge cases, observability, compliance – are exactly where AI-generated code tends to fall short.

The Saaspocalypse: Real Signal, Overstated Conclusion

Now, about the SaaSpocalypse. Unless you lived under a rock, you’ve heard that SaaS cmFebruary 2026 market selloff that, in roughly 48 hours, erased approximately $285 billion from SaaS company valuations in a single window – the largest AI-triggered repricing event in software history. ServiceNow fell 7%. Intuit dropped 11%. LegalZoom lost nearly 20%. The trigger: a growing market conviction that AI agents could replace the per-seat workflows that have powered SaaS business models for two decades.

The underlying business logic isn’t wrong. The barriers to entry for creating software are now so low thanks to coding agents that the build-versus-buy decision is shifting toward build in many cases, as one investor at One Way Ventures told TechCrunch. If one employee with AI tools can do the work of five, per-seat pricing collapses.

But “the model is under pressure” and “SaaS is dead” are very different statements, and conflating them is a mistake.

Here’s why it’s too early to write the obituary. Global SaaS spending is projected to rise from $318 billion in 2025 to $512 billion by 2028, according to Forrester – which is not the trajectory of a dying industry. The enterprise core doesn’t evaporate overnight. Regulated workflows in healthcare, financial services, and legal – which run on deeply integrated SaaS platforms – are not being ripped out and replaced by autonomous agents anytime soon. Epic, Salesforce, Workday, and ServiceNow are embedding AI, not being displaced by it. IDC predicts that by 2028, pure seat-based pricing will be obsolete, with 70% of software vendors refactoring their pricing strategies around consumption, outcomes, or organisational capability – that’s a business model evolution, not an extinction event.

Bain & Company’s read on history is instructive: technological transitions rarely produce extinction – they produce heterogeneity. Desktop survived mobile. Cloud didn’t kill on-premise so much as push it into specialised niches. The history of software is a history of layers accumulating, not replacing.

What is genuinely true is that the market has evolved and users are more empowered. The “build vs. buy” calculus has shifted. The days of buying a separate SaaS tool for every workflow are numbered. The funeral for SaaS has been extremely well-attended. The corpse, on closer inspection, is still breathing.

The Real Winners – and Where the Opportunity Lives

The clearest winners of the current moment are the foundation model providers. Anthropic hit $30 billion in annualised revenue in March 2026, up approximately 1,400% year-over-year, driven by enterprise and startup API adoption. Over 500 customers now spend over $1 million annually, eight of the Fortune 10 are Claude customers, and the company closed a $30 billion Series G at a $380 billion post-money valuation.

Where are the start up opportunities emerging? If I’m honest, I believe many VCs are a bit concerned by the current trend and are assessing the situation. The managing partner of a $1 billion fund told me they have yet to do a deal in 2026… nevertheless, every ‘crisis’ is an opportunity and ambitious founders will be all over these new ‘Agentic AI’ market:

• AI security and trust infrastructure. Enterprises need identity, permissions, policy enforcement, prompt monitoring, runtime guardrails, agent authentication and audit trails. If agents are going to touch sensitive systems, they need to be governed like employees and secured like software.
• Developer infrastructure for the vibe coding era. The world does not need more code generation alone. It needs better code review, testing, context management, architecture validation, dependency security and AI-native CI/CD. The more code AI creates, the more valuable verification becomes.
• Vertical AI with proprietary data and workflow depth. Generic agents will be powerful, but regulated and complex domains still need specialised workflows, domain context, data access, compliance and trust. Healthcare, legal, finance, construction, insurance, logistics and defence will not be won by wrappers.
• Agent observability and cost control. As agent workflows become more complex, companies will need to understand what agents did, why they did it, what it cost, where they failed and how to improve them.
• Outcome-based SaaS infrastructure. If software pricing moves from seats to outcomes, companies need new billing, metering, attribution and revenue operations tooling. “Pay per result” sounds simple until finance asks how the result is measured.
• Human-in-the-loop platforms. In high-stakes workflows, the winning product will not remove humans entirely. It will put them in the right place, at the right moment, with the right context.

The Honest Take

Agentic AI and vibe coding are real technological shifts with real commercial momentum. The challenges – security vulnerabilities, token costs, hallucinations, code quality – are also real, and they are not solved problems. Founders building in this space need to take them seriously, not dismiss them as teething issues. The early mover advantage in this wave will go to teams that understand both the upside and the engineering discipline required to deliver it in production.

The SaaSpocalypse was a market signal, not a market fact. What it told us is that per-seat pricing is structurally vulnerable, that users are more empowered than ever, and that the value in software is migrating toward intelligence and outcomes. What it didn’t tell us is that SaaS is dead – because it isn’t. It’s changing. And in the gap between the old model and the new one, that’s exactly where the best early-stage opportunities tend to hide.

At Remagine Ventures, that’s precisely the territory we find most interesting.

The original version of #Firgun was posted on LinkedIn to 7,840+ subscribers.

If you’re new here, welcome! So glad you’ve joined us.

Firgun (Hebrew: ??????, pronounced ‘feer-goon’) is an Israeli term describing genuine, unselfish joy in someone else’s success… It’s basically the opposite of “schadenfreude”. Every week since March 2020, we celebrate the wins of our startup community. Read more about why Firgun matters.

If this was forwarded to you and you’d like to subscribe, previous editions are on VC Cafe.

This week, Israel celebrated its 78th Independence Day (Yom Ha’atzmaut), a moment of national pride and reflection, with the traditional torch-lighting ceremony at Mount Herzl, barbecues, flyovers, and festivities across the country. Looking at the Calcalist 2026 most promising startups list, it’s clear that there’s a quiet shift going on: from software to hard tech, but also Israeli startups are finding they need to move abroad (aka US) earlier than before and to diversify their development centres from only Israel to cope with the impact of war, weakening dollar, etc.

In the wider tech world, SpaceX secured a $60 billion option to acquire Cursor (Anysphere), the AI-native code editor, or pay a $10 billion partnership fee to co-develop models on xAI’s Colossus supercomputer. OpenAI launched GPT-5.5, purpose-built for agentic coding, computer use, and scientific research, priced at $5/$30 per million tokens, intensifying competition at the frontier model layer. OpenAI also released a new image tool capable of generating complex charts and diagrams directly from prompts. China’s DeepSeek AI fired back with V4 Pro (1.6T parameters) and V4 Flash, both at a fraction of Western pricing, raising fresh questions about whether open-source models will commoditise AI faster than anyone expected. Meanwhile, Meta announced it will cut 10% of its workforce, roughly 8,000 employees. Microsoft offered voluntary buyouts affecting 7% of U.S. staff, signalling that even the AI-era giants are restructuring around leaner, more automated operations. Tim Cook stepped back as CEO of Apple as hardware chief John Ternus was appointed the new CEO to lead Apple’s next era. Google has launched new custom chips designed specifically to speed up AI processing and challenge Nvidia’s dominance. 4 month-old UK startup Irregular raised $500M in funding for its AI security lab focused on LLM protection and model control.

QUOTE OF THE WEEK

“Risk more than others think is safe. Dream more than others think is practical.” — Howard Schultz

NEW FUNDING ROUNDS

• Big congrats Renen Hallak and team VAST Data on completing a $1 billion Series F round at $30 billion valuation to scale its AI Operating System data platform!
• Kudos Arick Goomanovsky and team Band.ai on coming out of stealth with a $17M seed round to build the communication and interaction layer for the Internet of Agents!
• Way to go Shimon Tolts and team Copperhelm on coming out of stealth with a $7M seed round to monitor cloud environments, detects threats, and performs autonomous remediation with AI agents!

EXITS

• Congratulations Aharon Horwitz and team Fullpath on the acquisition by Cox Automotive Inc. for a reported hundreds of millions for your AI-driven customer data and marketing automation platform for the automotive industry!
• Well done Yossi Reitblat and team Ryft on the acquisition by Cyera for a reported $100M-$130M for your data management platform for AI agents!

LINKS FOR YOUR BROWSER

REMAGINE VENTURES/ VC CAFE

• The Anthropic question has replaced the Google question – Anthropic went from $9 billion to $30 billion in revenue and has an insatiable appetite for not just infrastructure, but also the product layer.
• VC is being rewired by AI – anyone can code now, and so AI didn’t just change how startups are built, but also how VCs find, assess and engage them. Will you be pitching to a VC AI Agent soon for funding?
• Are we in a bubble or an AI super cycle? Q1 broke all records in venture capital, reaching $300 billion globally (with the majority of that capital going to hyper scalers OpenAI and Anthropic and 80% of the money going to AI startups).
• Israel’s most promising startups 2026 – as selected by Calcalist ??????? – we’re clearly moving from SaaS/software to hard tech…

ISRAEL

• Israel’s Defence Tech leaders list by Viola Group
• The 50 most promising startups for 2026 by CTech by Calcalist
• Iron dome developer Rafael advances IPO plans at $20 billion valuation

GLOBAL

• Big ideas 2026 deck by ARK Investment Management LLC
• The 2026 Global Mega Trends by Activant Capital
• Early stage investing in the AI Age – filled with interesting stats by Carta

MEDIA OF THE WEEK

The Israel Defence Tech ecosystem landscape by Alex Shmulovich at Viola Ventures was added to my 2026 Israeli startup venture collection on VC Cafe. Learn more here.

the 10 largest Tech Ecosystem Hubs. ?? Tel Aviv, relatively small city, but $525B EV and 6.9x growth shows incredible capital efficiency. (source: Dealroom.co)

The Agent ecosystem is evolving very fast – Sapphire Ventures captured the current state in this landscape (source)

A framework on AI Value Creation and Pricing from Bessemer Venture Partners. Axis 1: What kind of value do you create? Axis 2: How provable is that value? The diagonal arrow is the key insight: pricing power increases as you move from soft ? hard ROI.

Voice AI applications and infrastructure by Dawn Capital (source)

Today’s state of Seed: more money, higher valuations, but less rounds happening (see links section). Also, seed teams are 39% smaller in 2025 than they were in 2021 and about 25% of seed startups raising this year will get to Series A in 2 years or less. Maybe 35% after 4 years.

Lastly, something I shared on X…

That’s all for this week, thank you for tuning in. There’s a lot to celebrate in our “small” community! Keep on creating!

Eze Vidra

Beyond the numbers, the 2026 rankings signal a fundamental pivot in the Israeli strategy. As highlighted by recent analysis in Globes, the era of “light software” is giving way to a more rugged, “Steel-Tech” approach. The high density of startups in the eVTOL space (AIR), autonomous drone swarms (Line 5), and surgical robotics (ForSight) proves that Israeli entrepreneurs are no longer afraid of hardware and complex logistics. This shift is a direct response to the “closed skies” and regional challenges of the past year; instead of slowing down, founders are building technologies that solve for physical independence. Whether it is protecting LLMs from agentic threats or mapping the underground with AI, the message for Israel’s 78th year is clear: the ecosystem is no longer just a “Silicon Wadi” feature, it has become a global factory for the essential infrastructure of the next decade.