Hi, I'm Kenneth. Welcome to this week's edition of Tech Venture Navigator.
This week felt like three different AI industries happening at once.
Yann LeCun left Meta, started a company in Paris, and raised $1.03 billion in a seed round to prove that the entire LLM paradigm is fundamentally wrong. OpenAI shipped GPT-5.4, its first model with native computer use, and it beat human performance on desktop navigation benchmarks. Nvidia invested in Mira Murati's Thinking Machines Lab and committed a gigawatt of next-gen chips. Oracle is reportedly planning to fire up to 30,000 people to fund AI data centres. Chamath published a deep dive on how equity tokenization is breaking traditional finance. And Jensen Huang is promising to "surprise the world" with a new chip at GTC next Monday.
The research is diverging. The products are converging. The economics are getting brutal. Here's what matters.
Brought to you by Kondo.
I. Yann LeCun Raises $1.03B to Prove LLMs Are a Dead End
This might be the most contrarian bet in AI right now.
Yann LeCun, Turing Award winner, left Meta in November after twelve years building their AI research operation. He walked into Zuckerberg's office and said he was leaving because he thought he could build something better, faster, outside Meta. Four months later, his new company AMI Labs just closed $1.03 billion in seed funding at a $3.5 billion pre-money valuation. Europe's largest seed round ever.
Investors: Bezos Expeditions, Cathay Innovation, Greycroft, Hiro Capital, HV Capital, Nvidia, Temasek, Samsung, Toyota Ventures, Mark Cuban, Daphni, Bpifrance.
The thesis: LLMs are a "statistical illusion." Impressive, but not intelligent. They predict the next word in a sequence. They don't understand the world. LeCun's alternative is JEPA (Joint Embedding Predictive Architecture), which learns abstract representations of how the world works instead of predicting pixel-by-pixel or word-by-word.
AMI Labs is building "world models," AI that learns from reality, not just language. Applications: healthcare, robotics, manufacturing, wearables. First partner: Nabla (medical AI). No product. No revenue. No near-term prospect of either. Just fundamental research.
CEO Alexandre LeBrun told TechCrunch: "My prediction is that 'world models' will be the next buzzword. In six months, every company will call itself a world model to raise funding."
Offices: Paris (HQ), New York, Montreal, Singapore.
Sources: TechCrunch | PitchBook | Sifted | The Next Web
✈️ NAVIGATOR’S EDGE: $1.03B for a company with no product, no revenue, and a thesis that the dominant paradigm in AI is fundamentally wrong. Either this is the most expensive academic bet in history, or LeCun is seeing something the rest of the industry is missing.
For context: World Labs (Fei-Fei Li) raised $1B last month. AMI Labs just raised $1.03B. That's $2B+ into world models in 30 days. The category went from academic curiosity to institutional-scale investment almost overnight.
For founders: If you're building on LLMs, this doesn't threaten you today. But in 18-24 months, if world models deliver, companies that interact with physical environments (robotics, manufacturing, autonomous systems) will need these capabilities. Start paying attention to JEPA.
For investors: This is a pure research bet with a 3-5 year horizon. No revenue path for at least two years. But if it works, the TAM is everything LLMs can't do: physical world understanding, spatial reasoning, reliable decision-making in high-stakes environments. Healthcare alone could justify the valuation.
II. GPT-5.4 Launches: Native Computer Use, 1M Token Context, Beats Humans
OpenAI shipped its most important model update since GPT-5.
GPT-5.4 is the first general-purpose model with native computer use: it can operate your desktop, browser, and software applications autonomously. It unifies the coding strengths of GPT-5.3-Codex, improved reasoning, and agentic workflows into a single model.
The numbers:
OSWorld-Verified: 75.0% success rate on desktop navigation (human performance: 72.4%, GPT-5.2: 47.3%)
WebArena-Verified: 67.3% browser success rate
GDPval: Matches or beats industry professionals in 83% of knowledge work comparisons across 44 occupations
Hallucinations: Individual claims 33% less likely to be false vs GPT-5.2. Full responses 18% less likely to contain errors
Context window: Up to 1M tokens in the API
Token efficiency: 47% fewer tokens on some tasks vs predecessors
Available now to ChatGPT Plus, Team, and Pro subscribers as GPT-5.4 Thinking. GPT-5.4 Pro for maximum performance on complex tasks. New Excel and Google Sheets plugins launched alongside.
Sources: OpenAI | TechCrunch | Fortune | VentureBeat
✈️ NAVIGATOR’S EDGE: This is the model release that makes "AI agents that do your work" real rather than theoretical. When an AI model can navigate your desktop better than the average human, the implications for enterprise automation are immediate.
For founders: If you're building workflow automation, GPT-5.4's native computer use is the foundation layer. The 1M token context means agents can hold an entire codebase, legal document set, or financial model in context and work across it.
For investors: Every SaaS company whose value proposition is "we make software easier to use" just got a competitor that can use the software directly. The impact on vertical SaaS valuations could be significant over the next 12 months.
III. Nvidia Backs Mira Murati with a Gigawatt Deal. GTC Starts Monday.
Mira Murati, former CTO of OpenAI, signed a deal with Nvidia worth "tens of billions of dollars" (per FT). Thinking Machines Lab will deploy at least one gigawatt of Nvidia's Vera Rubin systems starting early next year. Only the largest AI labs have approached that threshold.
Nvidia also made a "significant investment" in the company. Thinking Machines has now raised over $2 billion since its February 2025 founding. Previous investors: a16z, Accel, AMD's venture arm. 120 employees, though it lost CTO Barret Zoph and two other co-founders back to OpenAI.
GTC starts Monday (March 16-19). Jensen Huang keynotes at 11am PT at SAP Center. He's promised "a chip that will surprise the world." Expect: Vera Rubin deep dive, potential Feynman architecture preview, Groq-powered inference processor, agentic AI roadmap. Pregame show features CEOs of Perplexity, LangChain, Mistral, Skild AI, and OpenEvidence. 30,000+ attendees from 190 countries.
Sources: Axios | TechCrunch | CNBC | Bloomberg | Nvidia GTC
NAVIGATOR'S EDGE:
Nvidia is now invested in OpenAI ($30B), Thinking Machines, and AMI Labs. Jensen is backing every horse because they all need his chips. The circular financing model is now standard: invest, then sell them compute.
GTC is the event next week. If the Groq-powered inference chip delivers 10x cost reduction vs Blackwell, it changes the economics for every AI startup running inference. If Vera Rubin benchmarks confirm 3.3x improvement, the next generation of infrastructure spending is locked in. Clear your Monday morning.
IV. The AI Job Reckoning: 34,000 and Counting
Oracle: Planning to cut 20,000-30,000 employees (up to 18% of workforce) to free $8-10B in cash flow for AI data centres. $100B+ in debt. $156B in long-term AI infrastructure commitments. US banks pulling back from financing. Reportedly considering selling Cerner ($28.3B acquisition) to fund the buildout.
Block: 4,000 jobs gone (40% of workforce). Stock up 24%.
The public reaction: NBC News poll (Feb 27-March 3) found 57% of voters believe AI risks outweigh benefits. Stanford found a 16% employment decline for workers aged 22-25 in AI-exposed industries since late 2022.
V. Chamath's Deep Dive: Equity Tokenization Breaking TradFi
Worth flagging for anyone in fintech, VC, or LP roles.
Chamath's latest deep dive covers how equity tokenization is opening private market access. The numbers: global equity markets exceed $150 trillion, but access to the highest-growth private companies remains restricted. Equity token market cap has risen 3.5x since early 2025. Robinhood recently distributed OpenAI and SpaceX tokens to EU users. 90% of Americans would allocate retirement savings to private assets if they could.
The nuance that matters: most tokenised equity provides economic exposure, not direct ownership. Two tokens for the same company can represent fundamentally different rights. The infrastructure is real but standardisation isn't there yet.
For VC/LP context: If equity tokenization scales, it impacts how liquidity events work for private companies. Secondary markets become more liquid. LP distributions change. The line between "public" and "private" blurs permanently. Worth the full read.
Source: Social Capital
VI. Anthropic Update: Lawsuits Filed, Consumer Business Booming
Two lawsuits filed March 9 challenging the supply chain risk designation. Key business detail: the designation only applies to Claude's use in Pentagon contract work, not all commercial relationships. Microsoft, Google, and AWS continue non-defense business. CFO says it could still cost "multiple billions" in 2026 revenue. But consumer growth is accelerating: 1M+ daily signups, App Store momentum holding. Microsoft and dozens of OpenAI/DeepMind researchers filed amicus briefs in support. Legal timeline will determine the outcome.
Sources: Axios | Anthropic
March 3-11
Company | Amount | Stage | Sector | Key Investors |
|---|---|---|---|---|
AMI Labs | $1.03B | Seed | World Models | Bezos Expeditions, Nvidia, Temasek, Samsung |
Thinking Machines | Undisclosed | Strategic | Frontier AI | Nvidia |
Promptfoo | Acquired | Acquisition | AI Security | OpenAI |
Moltbook | Acquired | Acquisition | AI Social | Meta |
✈️ NAVIGATOR’S EDGE: The pattern this week: The pattern: Nvidia is everywhere. OpenAI acquiring Promptfoo signals safety/eval consolidating into foundation model companies. Meta buying Moltbook shows the agentic race extending to agent-to-agent communication.
Miles Clements is a Partner @ Accel where he helps to lead their growth fund. At Accel, Miles has led or invested in Atlassian, Cursor, Linear, and more.
Curated signals from the first week of 2026.
1. Consumer AI Engagement: ChatGPT DAU:MAU at 45% vs Gemini's 22%. Week 4 retention of 66%. "Smile curve" retention (dips then recovers) puts it alongside Gmail and Chrome. #QuitGPT movement (2.5M users) will test whether this holds.
2. "Your Supplier Is Your Competitor": Frontier labs build first-party products at under 50% token cost. Applied AI companies need to win on something other than the model. Directly relevant to our Kill Test: if your moat is "we use AI," you're already commoditised.
3. Nvidia State of AI: 87% of enterprises report AI reduced costs. 86% say budgets increase in 2026. 44% deploying or assessing AI agents. 3,200+ respondents. (Nvidia Blog).
The Big Number: 75%
GPT-5.4's score on OSWorld-Verified desktop navigation. Human performance: 72.4%. GPT-5.2 scored 47.3%. The first general-purpose AI model to beat humans at operating a computer. That number changes everything about how enterprise software gets built.
✈️ NAVIGATOR'S EDGE: If your pitch deck still positions your product as "AI-enhanced SaaS," delete that slide. The market just wiped $830 billion from companies with that exact positioning. Reframe around autonomous outcomes, proprietary data loops, and outcome-based pricing. The question VCs are asking in every meeting this week: "Could a Cowork plugin replicate what you do?" You need a bulletproof answer.
Lovable — The Full-Stack Machine. Generates UI, React frontend, and Supabase backends via chat. The gold standard for non-technical founders.
Bolt.new — In-Browser Development. Uses WebContainers for instant, full-stack environment generation.
Granola — Contextual Note-Taking. Not just a transcript; it’s an AI notepad that understands the vibe and nuance of VC calls.
Averi — The GEO Engine. Optimizes your brand to be cited by Perplexity and ChatGPT (Generative Engine Optimization).
Gamma — The Slide-Killer. Transforms raw research into aesthetic, interactive presentation decks.
Lemlist — The outreach tool we recommend for breaking through the noise with personalised fundraising outreach.
We sit at the center of 350k+ founders, operators, and LPs. Our goal is to eliminate the "random walk" of fundraising by using our data to make the one warm connection that actually matters.
For Founders: If you are raising a Seed or Series A and your roadmap is built for the agentic era, we want to see it. We don't just invest; we plug you into the distribution loops of our entire community. 👉 Startup Intake Form: Apply here
For Investors: Stop digging through generic deal flow. Tell us exactly what "high-signal" looks like for your mandate, and we'll route the outliers directly to your inbox. 👉 Investor Intake Form: Apply here
Fundraising in 2026 is a game of conviction and speed.
Let's stop wasting time on "maybes" and start building rounds that move the needle.
Pass this on to a founder who’ll use it.
Thanks for reading this edition. If you got value from it, share it with a founder, teammate, or friend who’d use it too it keeps this community sharp and growing. Not for you? You can always update your preferences anytime.
Want more practical ideas on how to raise smarter and build better?
Join the crew below and follow me on LinkedIn to never miss an update.
If you’d like tailored help you can book time with me on hubble here.
Stay going,
— Kenneth Kelly
Founder @Tech Venture Navigator
What do you think about Tech Venture Navigator?
💡 Love it | 🔥 Useful | 👍 Solid | 🤷♂️ Okay | 🚫 Skip it
or Leave feedback here.
Got ideas, topics, or frameworks you’d love to see covered? Just hit reply or drop me a message, we read every note.