EVERFORGE LABS (@everforgelabs) header
EVERFORGE LABS (@everforgelabs) profile picture

A venture capital and incubator focused on AI, robotics and Web3.

Joined November 2026
Photo by EVERFORGE LABS (@everforgelabs)Photo by EVERFORGE LABS (@everforgelabs)
We back founders building what lasts, across AI, robotics and Web3. Capital when you need it. Builders beside you when it matters. We don’t just fund companies. We forge them.
5
128
This week had one theme: AI finally turned narrative into cash flow, while crypto paid the price for tightening. Anthropic closed the largest IPO in history at a ~$1.5T valuation. In the same week, Bitcoin broke below $62,000 — and below its 200-week moving average. Real revenue on one side, liquidity retreating on the other. That is the dividing line of this week. **1. AI capital markets: Anthropic's record IPO, Nvidia doubles revenue** Anthropic priced on November 18 at $83 a share, raising $81.3B — the largest IPO ever — at a ~$1.5T fully diluted valuation, and listed on Nasdaq as ANT on November 19. Strategic investors including Nvidia, Microsoft, Alphabet and Micron committed a further $26B. Revenue reached $33.2B in the first nine months of 2026, ~13x year over year; Q3 revenue was $17B, with the company's first quarter of operating profit, at a 15% adjusted margin. Anthropic plans $70B of compute spend over the next 12 months, with ~$900B in long-term compute commitments. What it means: an AI lab has, for the first time, taken public-market pricing on operating profit and a verifiable revenue curve — the valuation anchor moves from narrative to cash flow. For builders: revenue growth and unit economics matter more than model rankings, for the first time. For investors: the next round of AI valuations gets recalibrated to Anthropic's multiples. Nvidia reported on November 17: Q3 revenue of $115.8B, up 103% year over year; data center revenue of $107.6B, up 110%; the new Vera Rubin platform contributed nearly $20B in its first quarter — the fastest product ramp in company history; Q4 guidance of $130B; FY2028 revenue growth outlook raised to 75%; a $150B buyback authorization. The earnings also confirmed: Jensen Huang resigned as CEO on November 12, with long-time CFO Colette Kress as interim CEO — and rumours of an SEC investigation circulated after the close, scope unclear. What it means: there is no sign of demand cooling — cloud vendors carry $2.5T in backlogs and $390B in purchase commitments. But a leadership handover and regulatory noise arrived together, and the stock fell after earnings: the market is now pricing whether AI capex can last, not this quarter's numbers. Watch: Rubin's delivery cadence, and whether the interim CEO becomes permanent. **2. Macro: the Fed turns hawkish; a December hike is now the base case** October CPI rose 0.4% month over month, core 0.3% — both above expectations. October nonfarm payrolls added 165,000, far above the 70,000 expected, and unemployment fell to 4.1%. PPI rose 6.5% year over year, the fastest since November 2022. Q3 GDP grew 3.4% annualized. The October FOMC minutes leaned hawkish; Governor Waller said December brings at least +25bp, with 50bp on the table if core CPI cooperates. The market has fully priced a hike on December 9 and started pricing another in January. The federal funds rate now sits at 3.75–4.00%. What it means: this is the source of this week's simultaneous slide in crypto and growth stocks — not a crypto problem, a repricing of liquidity. For builders: funding costs keep rising, and the window for burning cash for growth narrows. For investors: 2027's theme switches from the "rate-cut trade" to "real revenue at high rates." **3. Web3: Bitcoin breaks the 200-week moving average; ETF outflows continue** Bitcoin fell to ~$61,600 this week, down ~13% in seven days; Ethereum ~$1,745 and Solana ~$72, both sharply lower. Bitcoin broke below its 200-week moving average — a historically rare signal of extreme weakness. US spot Bitcoin ETFs saw another wave of redemptions: $306.5M of net outflows on November 18 alone, and cumulative net outflows in 2026 now exceed $2.5B. Total crypto market cap is back to ~$2.19T. What it means: the ETF channel has flipped from incremental buying to an outflow channel; price support now depends on spot buyers. Notably, Glassnode data this week showed some Bitcoin miners shifting hashpower and capital toward AI — selling the narrative asset to buy the revenue asset. That matches our call this week: cycle bottoms aren't decided by belief, they're decided by who's still paying for electricity and compute. Watch: whether the $60,000 level holds, and whether ETF flows stop bleeding. **4. Frontier models: Claude Fable 5.2 and 1,000-agent collaboration** Anthropic launched Claude Fable 5.2 and Mythos 5.2 on November 9 — the same model, different safety configurations: output up to 30% faster, prices down 20%. The headline is multi-agent orchestration: 1,000 Fable 5.2 agents self-organized without scaffolding and formally proved Fermat's Last Theorem in 30 hours (the previous attempt took 11 days); the RoboDojo robotics score jumped from 27.6% to 64.9%. What it means: the agent race is shifting from "how smart is one model" to "how many models can work together." For builders: the orchestration layer, the verification layer and the evaluation layer are worth more than another chatbot — the coordination cost of 1,000 agents is becoming the new moat. **5. Robotics: OpenAI dissolves its robotics team; Nvidia bets on "agentic robotics"** OpenAI confirmed on November 13 that it is dissolving its standalone robotics team, folding researchers into chief research officer Mark Chen's main research organization and apps and hardware teams into Greg Brockman's Applied division. The reason: general models like GPT-6 Astra can already drive robots directly — no separate "robot brain" needed. In Nvidia's simulation benchmarks, Astra-driven agents reached 74.8% manipulation success, above specialist robot models' 51.0%. OpenAI still plans a first humanoid demo in 2027. The same week, at CoRL 2026, UC Berkeley and Nvidia proposed "agentic robotics": LLMs writing inspectable robot programs instead of stacking demo data — 99% success on industrial tasks in simulation, versus ~20% for imitation-learning baselines. Nvidia also launched the GR00T 2 open-world action model and Isaac Lab 3.0. What it means: robotics is having the same "generalization" moment as large models — specialist stacks giving way to general models plus agentic programming. Whoever owns the simulation-verification-deployment loop owns scale. For Everforge: look at simulation infrastructure and task-level benchmarks, not another humanoid body. **6. Stablecoins and compliance: GENIUS rules land; Binance hack aftershocks** The US OCC published its final stablecoin rule on October 21, effective January 18, 2027, covering reserves, redemption, audits and supervision; the GENIUS license application form is under OMB review, with comments due November 23. Ripple's national trust bank charter took effect on October 26 and will custody RLUSD reserves. SEC Chair Atkins said on October 31 that "most crypto assets are not themselves securities." The UK's October 28 budget confirmed that, from April 2027, individuals disposing of compliant stablecoins are exempt from capital gains tax. On security and enforcement: Binance's ~$6.3B hot-wallet theft on October 30 moved into court proceedings this week, with asset freezes and recovery ongoing; Abu Dhabi's FSRA said it is in close contact with Binance; South Korea's FSC told exchanges to move more assets to cold wallets. The US market-structure bill, the CLARITY Act, remains stalled after a 49–50 procedural vote in the Senate in September. What it means: the rules are moving from "legislative debate" to "regulatory execution" — OCC rules plus trust-bank charters plus tax exemptions are laying a compliant track for stablecoins on three lines at once. ADGM/FSRA's direct engagement on the Binance incident shows the Middle East is becoming the market where compliance arbitrage and compliance standards take shape at the same time. Watch: next steps after the November 23 GENIUS comment deadline, and whether FSRA issues further guidance. --- This week's takeaway: the market is measuring everyone with the same ruler — revenue. Anthropic earned a $1.5T valuation on $33.2B of revenue and operating profit. Nvidia defended the AI capex story with doubled revenue. Assets with no revenue anchor can only follow liquidity down when money tightens. For builders: put the revenue curve ahead of the model ranking. For investors: the next cycle's winners are the companies that can still turn compute into cash flow in a high-rate world. We don't just fund companies. We forge them. #AI #Web3
2
185
Our portfolio team @constant_fi is on the road — and this is what onchain credit looks like in motion. Photos from Solana Breakpoint in London, then straight to Bangkok for Binance Blockchain Week, Nov 26–Dec 1. If you're there, grab a coffee with them and talk onchain credit. Here's why we're backing them: credit is the next thing to move onchain, and fixed rates are the missing piece. Variable-rate lending works for traders. It doesn't work for the businesses, treasuries, and funds that need to plan. Constant is building the predictable yield layer — the boring, essential infrastructure that lets real capital come onchain and stay there. Go find them in Bangkok. #OnChainFinance #DeFi
A few photos from our teammate at @solana Breakpoint in London We’re headed to Bangkok next for #BinanceBlockchainWeek on Nov 28–29. We’ll be in town from Nov 26 to Dec 1, so if you’re around, let’s grab a coffee and talk onchain credit.
1
3
99
The most important sentence in this post isn't about November 25. It's this: a more powerful model alone can't solve the physical world. XPENG is saying out loud what every embodied AI builder already knows. Language models got smart in a world of text. The physical world doesn't come with a training set — it comes with friction, gravity, and consequences. You can't prompt your way out of a robot falling over. That's why Physical AI is a different game. The moat isn't the model. It's the data loop: deploy in the real world, fail, learn, redeploy. Cars gave XPENG that loop at scale — millions of kilometers of physical-world data. A humanoid is the same stack pointed at a harder body. November 25 is worth watching for one reason: if IRON shows real progress, it means the car-to-robot transfer is working. And that would be a very big deal for everyone building embodied AI. #PhysicalAI #Robotics
XPENG AI Day is coming on November 25. The next frontier of AI isn't just about making models smarter. It's about making intelligence work in the physical world. Why is Physical AI fundamentally different? Why can't a more powerful model alone solve it? On November 25, we'll share our latest thinking, explore the co-evolution of AI models and physical systems, showcase the latest capabilities of IRON humanoid robot, and unveil how physical AI will reshape the future mobility experience. See you there!
1
3
95
The trick isn't the roller skates. It's the brake. Zhejiang Sci-Tech's lab has a robot gliding on skates — smooth, balanced, and it knows how to stop. Going fast on wheels is a demo. Stopping without falling is control. That's the whole story of embodied AI right now. The demos keep getting flashier — backflips, marathons, now roller skates — but the real question is the same one every robotics founder has to answer: can the system handle the messy, unscripted part? A curb. A sudden stop. A floor it has never seen. The labs that win won't be the ones with the best highlight reel. They'll be the ones whose robots fail gracefully, recover, and keep going. Locomotion is becoming a solved problem in pieces; general-purpose physical intelligence is still wide open. Watch the control stack, not the stunt. #Robotics #EmbodiedAI
Zhejiang Sci-Tech University’s Robotics Lab is teaching robots roller-skate. Really smooth at this speed, and it knows how to brake as well.
1
48
Welcome to the Forge, @kyleellicott. Kyle joins Everforge Labs as Venture Partner. He's spent his career at the intersection of builders and capital — Executive Director of the Stacks Asia Foundation, previously Partner at Stacks Ventures, and the person who seems to track every new fund on the planet. For Everforge, this is a big one. Based in the US, Kyle brings real American market access — the customers, capital, and partners our founders need to break in — plus the Bitcoin L2 lens and the operator's eye for what actually gets built. Our founders get a partner who's been on both sides of the table. If you're building in AI, robotics, or Web3 — come talk to us. #Bitcoin #VentureCapital
1
40
The miners are selling the treasury to buy the future. Glassnode: six listed miners have sold nearly half their BTC since January. MARA alone moved 40,581 BTC — about $3B. The coins aren't funding more mining. They're funding AI data centers. Riot sold 20,264 BTC and signed a 15-year, 910 MW lease. The power that once hashed blocks now rents to AI. This is the other side of yesterday's chip story: the whole stack is being repriced around one question — who owns the infrastructure that runs intelligence? And look at the timing. While miners rotate into 15-year contracted revenue, spot BTC ETFs shed $306M in a day and ~$2.9B in November. BTC is testing $62K, Fear & Greed at 10. The tourists are leaving. The builders are converting. That's the signal we watch: in every cycle, the winners sell the narrative asset and buy the revenue asset. The miners figured it out. The question is who else does. #Bitcoin #AI
Bitcoin miners are rapidly selling their reserves. For years, listed miners held what they mined, and some bought more. Since January, they have sold nearly half their bitcoin:native. A major shift in their approach: the BTC now funds their move into AI data centers.
1
48
The loop is closing. OpenAI's models found up to 14% PPA improvement on AMD's flagship GPU IP — in under five months, on designs AMD's engineers had already spent years optimizing. The same approach improved Jalapeño, OpenAI's first custom inference chip, with gains verified on real silicon. Not a simulation. Silicon. This is the second-order effect to watch: models are now improving the hardware that runs the models. Every point of PPA is a point off the cost of inference, and inference cost is the whole game — it decides who can afford to serve intelligence at scale. And the tell is the partnership: OpenAI is bringing its models into Synopsys' EDA toolchain. The model becomes part of the design flow, not a demo on top of it. Same pattern as this morning's research agents — expert work, done by models, verified in the real world. Vertical integration is the strategy now: own the model, own the chip, own the cost curve. #AI #ChipDesign
Working with @AMD, our models identified up to 14% PPA improvement on flagship GPU IP in less than five months—building on designs AMD’s engineers had already spent years optimizing. The same approach helped improve Jalapeño, our first custom inference chip, with gains verified on real silicon. New evaluations also show our models getting better at certain chip design tasks. We’re sharing these results and how we’re working with @Synopsys to bring our model capabilities and their trusted EDA design and validation tools to more chip designers. openai.com/index/advancing-c…
2
82
81 chains tells the real story: RWA is no longer an experiment on one chain. It's a market spreading across the whole stack. The numbers back it up. Tokenized real-world assets on-chain are now ~$40B (rwa.xyz, Nov 2026), up ~50% year over year — and that's before stablecoins, another ~$280B. Holder wallets hit 5.7M, up nearly 10x in a year. Tokenized Treasuries alone are $15B+ across 100+ funds. And the runway is still enormous: that's roughly 0.02% of global financial assets. BCG puts tokenized assets at $16T by 2030. The question isn't whether RWAs come on-chain — it's which rails, which assets, and which teams get there first. On-chain finance is a core direction for us at Everforge. We're backing the builders turning this map into real products. #RWA #OnChainFinance
81 blockchains mapped at rwaf.xyz. 🐘
1
2
45
The right question, and the honest one. As agents get cheaper to build, the wrapper is worth nothing — value has to sit somewhere it can't be copied. The verification layer is the one most people underestimate. An agent that trades is an agent making financial decisions with real money — without identity, audit trails, and proof of performance, no institution touches it, no matter how good the backtest looks. Trust infrastructure is the bottleneck, not the model. Same logic on the data side: the edge isn't the agent, it's the proprietary input it reasons over. Distribution is easy now; defensibility is the data and the proof. This is exactly the stack we're backing builders in. #AIAgents #DeFi
New @cbventures market map alert! 🚨 AI agents are moving from talking about markets to trading in them. But as agents get easier to build, where will lasting value accrue? My new post maps the agentic trading stack through that lens. Here’s where I’m focused: • Agentic trading harnesses and terminals that turn a user’s goal into an ongoing strategy, with real retention. • Market reasoning models and data assets that give trading agents a real edge. • Identity, audit trails, and performance verification infra built for non-human, financial decision makers. I care less about shiny demos and more about products that get harder to copy as they grow. If you’re building here, I’d love to hear from you.
2
35
Agents are only as good as the data they can reach. Most of them are still running on stale knowledge. Teneo just made the fix trivial: live web data for your Grok bot in three steps, over MCP. Add the server, paste your key, done. No scraping infra, no pipeline to maintain. This is the infrastructure pattern that wins — the data layer becomes a plug, not a project. MCP is the rail: whoever owns a scarce, live input and distributes it through an open protocol gets pulled into every agent stack being built right now. Free tier to try it is the right call. Distribution first, monetization follows usage. Proud to back @teneo_protocol. #AIAgents #MCP
Live Web Data API from Teneo gives your Grok bot live web data in 3 steps. Add livewebdata.com/mcp as the MCP server, paste in your key from livewebdata.com, and start using it. The free test account is enough to try it.
3
74
The reliability threshold just moved. OpenAI's new report: coding agents now complete 46% of 2–4 hour research tasks on their own — up from 23% in March. On 4–8 hour tasks, 35%. The median researcher is already spending $2,400 a day on agent use. This is the number that matters. Not benchmarks, not demos — multi-hour work, finished, unsupervised. The "jagged frontier" is the honest part: capability is uneven, agents ace some tasks and fail others. So the winners won't be the best models. They'll be the teams that build verification, evals, and workflows around the jagged edges — the ones who make unreliable capability reliable enough to sell. Agents stopped being a demo this year. They're becoming labor. The question now is who owns the verification layer. #AIAgents #AIResearch
We're sharing new data on how AI is changing research at OpenAI. We find researchers' AI use continues to rise exponentially, and agents are completing longer tasks, but human involvement remains central in research. Researchers are also spending more time on safety and alignment, with human effort in these areas growing faster than in any other individual research area. This report aims to increase transparency around research acceleration and inform the governance of increasingly capable AI. We believe that AI labs should be required to track and publicly report indicators of recursive self-improvement (RSI): the process by which AI accelerates the development of future AI systems. Our hope is that this expanded report will help sustain and advance shared measurement norms across the industry. openai.com/index/research-ac…
1
1
57
The $500M is the headline. The real story is distribution. Tencent's Q3: WorkBuddy passed 10M DAU, and per Martin Lau the last 7 days annualize to $500M+ in revenue, 5x Q2. Hy4 topping Agent Arena is nice, but benchmarks don't sell seats. Distribution does: the agent lives inside workflows enterprises already use. Two caveats before calling this the inflection point. A 7-day annualized run rate flatters — early adopters, pent-up demand. What matters is renewal: does usage convert to paid seats, and does it expand? Token volume is scale, not revenue. Usage is vanity until it's monetized. The structural signal: agents eat the interface layer, not the system of record. CRM is next — the data stays, the UI goes. Tencent Cloud up 35% says the same thing from the other side. Startups win vertical, not horizontal. #AIAgents #SaaS
Tencent’s Q3 results show AI’s significant achievements across our models and applications. Revenue reached $30.78 billion, up 8% year on year. Our @TencentHunyuan Hy4 model ranked first among Chinese models on Agent Arena at launch, @WorkBuddy_AI passed 10 million daily active users, and @TencentCloud’s revenue growth rate accelerated to 35%. More: buff.ly/CyK7XDv
2
38
1/ Meet @UVioletAI — one of the boldest bets in our portfolio. No blockchain. No gas. No forks. Client-side validation + zk-proofs, built to run AI and IoT workloads at a TPS blockchains can't touch — anchored to Bitcoin and Lightning.
1
4
7
207
3/ Behind it: Dr Maxim Orlovsky @dr_orlovsky. The mind behind RGB smart contracts on Bitcoin, AluVM, and the LNP/BP Association. A cypherpunk and neuroscientist who has spent a decade building the post-blockchain stack.
1
5
37
GPT-6.2 Astra is OpenAI's first released model to hit its 'High' bar for AI self-improvement — a model that can materially accelerate AI research itself. When research starts compounding, the best founders build for the second-order effects.
1
47
Grateful to the partners standing with Everforge Labs from day one: @fxh_crypto @DFG__Official @FundamentalLabs @waterdripfund @CerasVentures Thank you for your support. Here’s to backing the founders building what lasts across AI, robotics and Web3.
4
38
EVERFORGE LABS retweeted
MASSIVE: 🇭🇰 Hong Kong plans to introduce new rules for crypto trading, advisory, and custody services by the end of this year. Asia is moving towards clearer crypto regulations.
73
85
4
795
70,470