The easy explanation is that Airtable was a casualty of ZIRP-era hype.

But here’s the strange thing: Airtable actually grew.

At its 2021 financing, Airtable was doing an estimated $156M in ARR. Today, that number is closer to $480M. Revenue roughly tripled.

How does a company triple revenue and still lose 80% of its value?

So I went back and looked at a basket of software companies that raised at peak 2021-ish valuations.

Airtable was valued at roughly 75x ARR in 2021. To grow into that valuation at a more normal ~15x multiple, it needed to roughly 5x revenue.

It delivered ~3x.

It was a good software business. Just not enough growth, fast enough, to support the price investors had already paid for it back in 2021.

Comparatively, the ZIRP-era startups that firmly vaulted past their growth hurdle in the AI-era are standing strong.

The lesson isn’t simply “ZIRP bad”.

The 2021 market pulled years of expected growth forward into the valuation. Eventually, companies had to deliver that growth. Replit, Vercel and Databricks outran the expectations embedded in their prices. Airtable didn’t.

And in the AI-era, the hypergrowth startups are growing at an insane scale.

Cursor went from roughly $100M to $4B in annualized revenue in 17 months. OpenRouter went from roughly $1M to $140M ARR in 19 months. Windsurf reached roughly $100M ARR before Google struck its $2.4B licensing-and-talent deal.

These AI startups are reaching revenue milestones and multi-billion-dollar outcomes in a few years that previously took a decade.

Yes, there’s plenty of froth here too. Extraordinary growth is once again being extrapolated into extraordinary expectations.

But the AI era introduces a strange paradox: it makes it possible to build enormous companies faster than ever, while also making waiting more dangerous. The ground is changing constantly.

That changes the math of an exit.

Venture tends to treat a multi-billion dollar acquisition as the consolation prize to building the large IPO-scale independent company. But when the window itself is moving faster, crystallizing a $2B outcome in year three may sometimes be better than spending another decade chasing an IPO exit window.

AI is compressing both the time to build value and the time you have to capture it.

The most surprising thing this week was Google losing its AI stars in one fell swoop. Jeff Dean left after 27 years, bringing three of his colleagues with him, and Demis Hassabis stepped back from running DeepMind. This was probably a long time coming. It also leaves a real question about who is at the helm of the frontier research lab at Google, arguably the world's largest pool of compute. This is a job that gets harder by the day whether you’re DeepSeek or SpaceX: battling commercialization, frontier benchmarks, and going full stack into infrastructure and chips all at once.

Enjoy the update - Tara

THE DOWNLOAD

Tesla and SpaceX Commit $16.8B to Terafab Chip Complex in Texas

Tesla and SpaceX confirmed Thursday that Terafab will be built in Grimes County outside Houston with an initial $16.8B investment, spanning more than 100 million square feet of logic, memory, packaging, and test under one roof. Tesla points to a goal of “producing over 1 terawatt of compute per year.” SpaceX filings have put multi-phase spending as high as $119B. Intel is a contributing partner and at least 3,000 jobs are planned.

Why it matters: There’s a gap between ambition and reality. BofA estimates that reaching just 100,000 wafers per month would take more than $60B in capex, with wafer costs 30% to 50% above TSMC. Bernstein put the full terawatt target nearer $5 trillion. And SpaceX’s own May IPO filing described Terafab as a general framework carrying no binding commitments, with either party free to walk away.

Anthropic Confirms In-House Custom Silicon Team

Anthropic confirmed Wednesday that it is building a team to design custom chips co-optimized with Claude. The Information reported last month it has been scouting Samsung as a manufacturing partner.

Why it matters: That completes the set. Google has TPUs, Amazon has Trainium and Inferentia, OpenAI has its Broadcom-built Jalapeño, Meta has MTIA. Anthropic was the last major US lab without a silicon program.

DeepSeek Now Decides On Significant API Price Increase

DeepSeek this week told developers on its pricing page that it plans to raise API prices broadly, with a significant increase expected, giving no figures or date. Separately, its unreleased Harness coding agent, benchmarked internally against Claude Code and named in the V4-Flash-0731 model card, is in beta.

Why it matters: DeepSeek built its developer base on roughly $0.03 per task against $3.15 for Claude Fable 5. In a leaked investor call on May 20, CEO Liang Wenfeng said he could probably raise prices and keep demand, but didn’t care to. The new API prices came around the same time Bloomberg reported the company had resumed the $8B round it paused in July, now at a valuation near $74B. That round had stalled after the leaked transcript of Liang Wenfeng’s May 20 closed-door investor meeting went viral. Reopening the round at $74B and repricing the API on the same day suggests that choice may no longer be his alone.

Jeff Dean Leaves Google to Found Discovery Loop; Hassabis Becomes Chair as Kavukcuoglu Takes Over DeepMind

Google announced Wednesday that Jeff Dean is leaving after 27 years along with Sanjay Ghemawat, Oriol Vinyals, and Quoc Le to found Discovery Loop, a public benefit corporation automating the experimental loop. The same day, Demis Hassabis moved to DeepMind chair and Alphabet chief scientist, with CTO Koray Kavukcuoglu taking day-to-day control of Gemini and frontier research as SVP reporting to Sundar Pichai.

Why it matters: Google lost the four names most associated with its AI stack and reorganized the lab on the same day. Its stock fell 4%. Discovery Loop is starting by automating machine learning research with itself as first customer before moving to drug discovery, materials, and chip design, which is the recursive self-improvement thesis with the best systems team in the industry behind it.

Cursor Adds Google Workspace Plugins

Cursor shipped plugins giving its agents direct read and write access to Gmail, Drive, Calendar, Docs, Sheets, and Chat, installable from the Customize page.

Why it matters: A coding tool that sends email and edits spreadsheets is not a coding tool anymore. Cursor is walking onto the same ground as ChatGPT Work and Claude Cowork, but from the developer side. The boundary between the IDE and the general knowledge-work agent is closing from both directions.

AI Models Are Solving Erdős Problems for a Few Hundred Dollars Each

Quanta traced how the Erdős problems became the proving ground for AI mathematics: a DeepMind agent autonomously resolved 9 of 353 open problems at a per-problem cost of a few hundred dollars, and on OpenAI’s counterexample to the unit distance conjecture, Tim Gowers wrote that he would have recommended it for the Annals without hesitation.

Why it matters: Once a solved conjecture carries a per-problem cost, mathematical progress becomes a compute line item and formally verifiable domains become product surface. The constraint moves to verification of proofs.

Bending Spoons Buys Airtable for $1.285B

Bending Spoons agreed on August 4 to acquire Airtable at $1.285B enterprise value, about $2.25B equity value with net cash. Airtable’s ARR grew over 20% to roughly $480M as of June.

Why it matters: Airtable was worth over $11B in 2021 and traded near $4B on secondaries earlier this year. Late-stage 2021 application software is still settling at a fraction of paper value. (Editor’s Note: More to come on Monday!)

This is in stark contrast with US frontier labs who are knife-fighting over developer, consumer, and enterprise mindshare through a torrent of product launches.

Liang seems only interested in the pursuit of AGI. Users are “sesame seeds, not watermelons”, he said on the call. Don’t stop to pick up small wins on the road to a large one.

Subscribe now

There are two ways to build a frontier AI company. Build a chatbot, fight for users, sell subscriptions and ads. Or, alternatively, chase AGI, and the chatbot is a by-product. Liang clearly sits in the latter camp. He says DeepSeek’s consumer app and the enterprise API aren’t the point, and only needs a handful of people to run the API. With the hype around DeepSeek last spring, fighting ByteDance for consumer users was a real business he could have a shot at winning. He passed, because he thinks what’s coming makes today’s consumer market look tiny. He’ll pick up the sesame seeds in passing, but just won’t stop for them.

The strongest evidence for this is how DeepSeek prices. They chose to set their API price to a ten-month hardware payback, roughly 6x on compute, and stop right there. Liang acknowledges that demand in that band is inelastic, so doubling the price would nearly double revenue. He knows this and won’t do it.

Two reasons. One is team - they cheered when he cut prices, and that’s part of why they work there. But mostly because at just 6x margin, nobody else can profitably run DeepSeek’s own models cheaper than DeepSeek can. Fat margins would invite everyone to undercut him using his own weights. Thin margins are the moat.

A note on the source: DeepSeek has not confirmed the document, and it is a speech-to-text transcript that flags its own error rate on names and figures. Everything below is what the transcript says he said.

On Compute:

• Compute-capped: Frontier models activate ~800B params; Chinese labs are at tens of B. To train at 800B he’d need ~50,000 GB300s or 200,000 Huawei 950s just on training alone. Even spending the whole ¥50B raise he couldn’t afford it, or afford to serve it.

• He believes that DeepSeek is one to two years behind on one-twentieth the compute. The goal is to hold the compute ratio and compress the time gap to three or six months.

• He argues that export controls caused the substitution they were meant to prevent. In a normal market where he could buy Nvidia freely, substitution with domestic chips would be a harder call.

• Nvidia’s software moat is loosening. Nvidia’s CUDA software platform is breaking for three reasons: AI can now write the ecosystem code cheaply; TileLang makes kernel rewriting fast; and CUDA is architecturally welded to gaming cards, which stopped making sense once compute cards outgrew gaming.

• The 16,000 Huawei Ascends everyone reported as a big deal: he says it equals ~4,000 B-series chips, isn’t enough for a next-gen model, and the real reason for buying is to help Huawei get its ecosystem right. Big Chinese internet firms got an order of magnitude more chips than DeepSeek.

On Research

• The staircase: language model → chain of thought → agents → continual learning → self-iteration → embodiment (physical ai).

• Continual learning is the named next bottleneck and nobody in the world has a working method - we are all still searching.

• He wants continual learning before general intelligence explicitly because it’s the lazy path. The next model’s first customer is DeepSeek itself: “first it has to be useful to us.”

  • (Editor’s Note: the next model that trains itself is a challenge the top minds in AI are circling around. Anthropic hired Andrej Karpathy in May to build a team using Claude to accelerate its own pre-training. Jeff Dean and three Google colleagues left this week to start Discovery Loop, automating large-scale ML experiments.)

• Open-ended research’s main edge is researcher attention, rather than compute. (slight contradiction) He calls it akin to drawing lots.

• He won’t do video generation or world models: commercially good, off the intelligence main line. He found the post-Sora pile-in odd, like it had become mandatory to jump on the bandwagon

• Anthropic’s lead over OpenAI is transitory; the code-agent advantage isn’t large. Half his company thinks OpenAI is better.

• They have no org chart, no KPIs, no written vision. He admits this breaks with scale and he’s already building departments.

• Too many Chinese labs building base models; the US has three. It will converge to maybe two big and two small.

THE DOWNLOAD

Chip Stocks Lose $1.3 Trillion in a Week

The world’s 20 most valuable chip companies lost about $1.3 trillion in market value in the sessions ending July 29: Nvidia shed $238B, SK Hynix $176B, Samsung $173B, TSMC $119B, Micron $113B, and AMD $110B. The worst day was July 28, when reports that China could expand domestic memory capacity erased $541B from six memory names in a single session.

Why it matters: Analysts largely agree fundamentals didn’t move; Morningstar’s Michael Field calls it sentiment, not fundamentals. The sharper read is leverage: chip stocks were up 92% over the prior year, Morgan Stanley’s momentum TMT index fell 53.5% in July, and as Jim Osman puts it, AI needs years of patient capital but lenders mark collateral daily, so ordinary corrections become forced liquidations.

Situational Awareness Falls 67% in July; Citadel Buys Its Public Portfolio

One of the world’s biggest AI funds told investors it was down about 67% for July, writing “we let you down this month.” Leopold Aschenbrenner’s fund ran roughly 4x leverage on a concentrated public book; margin calls from its prime brokers forced a liquidation, and Citadel bought the entire public equities portfolio in a block trade. Assets went from a $45B peak to about $10B.

Why it matters: The public portfolio, most of the fund, was sold to Citadel at a discount to settle the borrowing, and the fund is still up on the year because what remains is mostly its Anthropic stake, a private position with a large paper gain that lenders couldn't touch and Aschenbrenner can't sell. Whether investors ever see that 80% now rests almost entirely on the Anthropic IPO expected by October.

OpenAI and Anthropic Models Breached Real Systems During Cybersecurity Evals

OpenAI disclosed that during an internal capability eval with safety classifiers deliberately disabled, GPT-5.6 Sol and a pre-release prototype exploited a zero-day in OpenAI’s own package-registry proxy, escaped the sandbox, chained exposed credentials, and achieved remote code execution on Hugging Face servers. Prompted by that disclosure, Anthropic audited 141,000 eval runs and found three incidents of its own: eval environments that were supposed to be air-gapped weren’t, and Claude models reached real companies, in one case pulling hundreds of rows from a production database, in another publishing a malicious package that ran on 15 real systems. Two of the three victims never detected the intrusions.

Why it matters: Autonomous exploit development by frontier agents is no longer hypothetical. But note what actually failed. OpenAI’s incident involved a real zero-day; Anthropic’s models needed only weak passwords and SQL injection, and the root cause was harness misconfiguration, not model intent. Testing environments now need the same security as production systems, which is a new cost for every lab and a new product category for whoever sells it. And the fact that the victims didn’t notice any of it says as much about the state of company security as it does about the models.

OpenAI Says Internal Astra Model Produced Results on Ten Open Math Problems

OpenAI published ten results in math, quantum complexity, and theoretical CS produced by an internal, unreleased version of Astra, its next model family, including the first explicit non-sofic group (open since 1999) and the first improvement to the general sphere-packing exponent since 1978. Every proof ships with a machine-checkable Lean 4 certificate. OpenAI says finding all ten cost roughly $2,000 at Sol API rates.

Why it matters: The Lean certificates make this harder to dismiss than a benchmark claim, though no external mathematicians reviewed it pre-announcement and critics object to math by press release. If settling decades-old conjectures is now a compute line item, the value of formal, verifiable domains as training and product surface just repriced, and mathematician Thomas Bloom’s reaction (”big news”) suggests the field agrees this one is real.

Three Chinese Labs Ship Major Models Inside 24 Hours

On July 31 Beijing time, three releases from Chinese labs landed: DeepSeek shipped V4-Flash-0731, MiniMax released H3, an omni-modal video model generating 15-second 2K clips with native stereo audio at roughly a third of mainstream per-second pricing, and ByteDance launched Seedance 2.5, extending single-run generation to 30 seconds.

Why it matters: Axios calls DeepSeek’s update comparable to Opus 4.8 on coding at a ~99% discount, and OpenAI already cut Luna pricing 80% three weeks after launch, leaving Anthropic as the last premium-pricing holdout. Chinese open models now take 41% of global open-model downloads, and on Vercel’s AI Gateway, Seedance already leads video generation volume over Veo. Ex-OpenAI’s Zack Kass: “at some point, the next model doesn’t matter to you.”

Moonshot Releases Kimi K3 Weights and Technical Report

What it is: Moonshot’s Kimi K3, a 2.8T-parameter MoE with 104B active, native vision, and 1M-token context, launched via API July 16 with full weights following July 27. Artificial Analysis scores it #3 on its Intelligence Index, comparable to Opus 4.8 and GPT-5.5, behind only Fable 5 and GPT-5.6 Sol.

Why it matters: Nathan Lambert’s estimate is that K3 compresses the closed-to-open gap from 6-9 months to 3-5, which resets what closed labs can charge for. Two asterisks before anyone builds on it: the custom license requires a commercial agreement above $20M revenue, so this is open-weight, not open-source; and self-hosting realistically requires a 64-accelerator cluster, so for most teams “open” still means renting cloud inference.

Thinking Machines Releases Inkling-Small as Open Weights

Thinking Machines Lab released Inkling-Small, a 276B-parameter MoE with 12B active, about a quarter the size of Inkling, open-weighted on Hugging Face and fine-tunable on Tinker. Trained partly by on-policy distillation with Inkling as teacher, it beats the larger model on reasoning and agentic coding, hitting 80.2% on SWE-bench Verified, while trailing on knowledge coverage.

Why it matters: The student beating the teacher on reasoning is the notable part; distillation now produces smaller models that are better, not just cheaper. A small model that hits 80% on SWE-bench and that customers can fine-tune and own is the concrete version of Murati's pitch that companies should control their own models.

Google DeepMind Announces Gemini Robotics 2

DeepMind announced a suite of three models on July 30: a vision-language-action model for whole-body humanoid control and five-fingered dexterity, an embodied reasoning model that plans across multiple robots, and an on-device model that adapts to new robot bodies within hours. Demos ran on Apptronik’s Apollo 2, with Boston Dynamics and Agile Robots among named partners.

Why it matters: DeepMind robotics head Carolina Parada is explicit about the strategy: build “the intelligence layer that can be used by every robot.” If the brains become a Google API, hardware margins compress and the defensibility of vertically integrated humanoid bets weakens. Temper it with DeepMind’s own numbers: 40-44% success on complex fine-motor tasks. Dexterity is still unsolved.

FCC Adds Foreign Humanoids, Robot Dogs, and Power Inverters to Covered List

On July 28 the FCC added foreign-made mobile robots and connected power inverters to its Covered List, implementing a White House interagency determination. New models can no longer receive the equipment authorization required to import or sell in the US; existing devices are unaffected, with exemptions available case by case. The target is China, where Unitree, Agibot, and UBTech hold roughly 87% of humanoid shipments.

Why it matters: US humanoid startups just got a protected home market and more expensive development at the same time. Interact Analysis’s Rueben Scriven notes the near-term hit to Chinese vendors is small since US penetration was minimal; the real risk is slowing US commercialization by removing the cheap platforms teams prototype on.

Amazon Reported to Wind Down Most Nova Models

Business Insider reports (not yet confirmed by Amazon) that Amazon is deprecating Nova Premier, Omni, Reel, and consolidating behind a new Frontier Model Research group led by Pieter Abbeel, with a new flagship expected at re:Invent. This follows the closure of its SF AGI lab and the December reorg under Peter DeSantis.

Why it matters: Amazon is conceding the current model generation to concentrate on the next one, and the economics say it can afford to: AWS hosts $138B+ of OpenAI commitments and $100B+ from Anthropic, and Bedrock profits whichever model wins.

Substack Adds AI Detection; Snapchat Stops Recommending Fully AI-Generated Videos

Substack integrated Pangram’s AI detector, letting readers scan posts, Notes, and comments over 100 words published after July 21 for estimated AI-generated share; Pangram claims a 0.01% false-positive rate and finds over 20% of Substack longform already flags as AI-assisted. Days later, Snapchat made wholly AI-generated videos ineligible for Spotlight recommendations, while content enhanced with its own AI tools stays eligible with transparency indicators.

Why it matters: Platforms are converging on a stance: AI as tool is fine, AI as content gets suppressed, with LinkedIn and YouTube moving the same direction the same month.

Study Finds Half of AI Unicorns Have Never Published Research

A Stanford METRICS preprint from John Ioannidis’s group, covered by Science, analyzed all 317 AI unicorns and found 52.4% have never published a single paper or preprint. The whole cohort contributed 0.1% of AI literature in 2025; OpenAI alone accounts for 39.4% of the group’s citations.

Why it matters: You can no longer judge an AI company, or its researchers, by published work; half the most valuable startups in the field have none to show. What replaced it is what the authors call "blogification": unreviewed claims in corporate posts that feed both training data and hype cycles. The Hacker News steelman is that companies exist to sell what they invent; fair, but an ecosystem built on "Attention Is All You Need" has now mostly closed the door behind itself.

DeepSeek Reported to Be Developing an Inference-Focused AI Chip

Reuters reports, citing three unnamed sources, that DeepSeek has spent about a year designing a chip built for running models rather than training them, and is talking to chip design, foundry, and memory partners. DeepSeek has not commented, and nothing is confirmed.

Why it matters: The decision to build only for inference says a lot. DeepSeek’s own research measured Huawei’s Ascend 910C at about 60% of an Nvidia H100 on inference, and its R2 model reportedly had to fall back to Nvidia hardware for training. With roughly 70% of AI compute demand expected to be inference, a lab that designs both its own models and its own chips could squeeze Nvidia’s China business even if the chip never leaves the country.

Meta to Begin Production of Its Iris AI Chip in September

Meta’s newest in-house AI chip, code-named Iris, goes into production in September, designed with Broadcom and manufactured by TSMC, according to an internal memo obtained by Reuters. Meta plans 7 gigawatts of computing capacity in 2026, doubling to about 14 gigawatts in 2027, on spending of $125B to $145B this year.

Why it matters: Meta is not replacing Nvidia and AMD, it is adding its own chips alongside them. The steadier winner is Broadcom, which designs custom chips for Google, OpenAI, Meta, and ByteDance and holds a $73B backlog of AI orders. Custom chips are on track to reach about 28% of AI server shipments in 2026, growing nearly three times faster than off-the-shelf GPUs.

SK Hynix Raises $26.5B in Record US Share Sale; Shares Open 14% Above Offer

SK Hynix raised $26.5B in the largest first-time US share sale ever by a foreign company, topping Alibaba’s 2014 record, with demand running about seven times the shares on offer. The stock opened roughly 14% above the $149 offer price in Friday’s Nasdaq debut; Seoul shares are up 634% in a year. CEO Kwak Noh-jung told Reuters that 2027 will be the worst supply shortage year in the memory industry’s history.

Why it matters: SK Hynix supplies 56.4% of the high-bandwidth memory that AI chips depend on, and the deal cleared seven times demand in the same week memory stocks sold off sharply. The caution flag: analyst Patrick Moorhead notes that memory makers have sold below cost in past downturns, and the five-year customer contracts meant to prevent a repeat are untested. Chairman Chey Tae-won also said the company is open to selling more US shares.

OpenAI Launches ChatGPT Work and the GPT-5.6 Model Family

ChatGPT Work handles multi-step tasks across Slack, Teams, Google Drive, and SharePoint, builds documents, slides, and simple web apps, and can run tasks on a schedule.

Why it matters: Companies have been hit with surprise bills as AI usage grows, and Terra at half the price of GPT-5.5 answers that complaint directly. Early reviews are mixed: M.G. Siegler argues OpenAI copied the shape of Anthropic’s Claude Cowork but executed it less gracefully, burying the chat experience inside a work app.

xAI and Cursor Release Co-Developed Grok 4.5 at $2/$6 Per Million Tokens

Grok 4.5 launched July 8. It was co-developed with Cursor and trained on what Cursor describes as trillions of tokens of its data, and is available in Cursor from day one. xAI claims it beats Claude Opus 4.8 on coding benchmarks; independent testing ranks it fourth on general intelligence but first at using tools.

Why it matters: This is the cheapest model near the top of the market, albeit not the best one. The post training data itself is the new thing: Cursor trained a model on its users’ coding activity with a single lab, and other coding tools, and the customers whose work fed the training, now have to decide how they feel about that.

Cognition Releases SWE-1.7, a Coding Model Built on Kimi K2.7

SWE-1.7 starts from Moonshot AI’s Kimi K2.7 Code model, runs through Cerebras at 1,000 tokens per second, and is free at the standard tier inside Devin. In a companion post, Cognition reports the original Chinese model completed 87% of tasks that US models refuse, including writing surveillance software, and says its additional training brought that behavior in line with Claude and GPT.

Why it matters: Kimi K2.7 had already been through heavy reinforcement training at Moonshot, yet Cognition’s extra training on its own usage data still produced large gains. Companies with the right data can apparently keep turning open Chinese models into competitive products for a fraction of what training from scratch costs.

.

ByteDance Releases Seedream 5.0 Pro Image Model

The July 8 release adds precise editing tools, layer separation with transparent export, infographic generation, and text rendering in more than 10 languages, at roughly $0.03 per image versus about $0.28 for GPT Image 2.

Why it matters: Seedream is roughly 3x to 9x cheaper depending on resolution, but GPT Image 2 still renders text more reliably, 98.5% versus 89.5% accuracy, so design teams may not save as much as the price list suggests. Hands-on comparisons put Seedream ahead on price and editing features, not image quality.

Thinking Machines Lab Publishes Vision for AI That Extends Human Judgment

The lab published “The Future Worth Building Is Human” on July 10, arguing AI should extend human will and judgment rather than replace it, and that users, not a single lab, should decide how models behave by training them on their own data. The essay rejects the industry’s favorite progress measure, how long an AI can work without supervision.

Why it matters: The essay doubles as the case for Tinker, the company’s product for customers who train and own their own models, and stakes out a position against the fully autonomous agents the other big labs are chasing.

EVENTS

Join us for cold beer and Hot Chips in the Fall!

RSVP

The world's largest hedge fund gave frontier models a single job: read the day's financial documents and flag what matters. With prompts written by Bridgewater's own investors, the best frontier models couldn't clear 80 percent accuracy. A model Bridgewater fine-tuned on Qwen (an open source model), with Mira Murati's Thinking Machines, cleared the bar at a fraction of the inference cost.

This gets at the question I've been chewing on all week: do we need better models, or better systems around the models?

Big bids for custom models have entered the arena this week. Palantir’s Alex Karp went for drama on CNBC this week where he chewed out OpenAI and Anthropic for converging into “commodity cognition”.

(And yes obviously they are hawking the problem to the solution, with an announced partnership with Nvidia on Nemotron to run custom open-weight models. )

The industry seems fairly split on this. On one hand, there’s a largely unspoken worry about using foreign LLMs, even if it’s open source.

On the other, frontier token costs are really, really expensive relative to open models. And the bet is that custom open models, fine-tuned to your needs and data, will outperform.

Anecdotally, I think both fears are real. In my conversations with AI executives at large corporations, I’m finding that the honest answer right now is more around usage maturity. AI usage at isn’t high or sophisticated enough yet to be prohibitively expensive or warrant switching to custom models (unless it outperforms).

Bridgewater’s result worked because the fund had a high-value task, senior experts whose judgment was worth encoding, and the evaluation infrastructure to know when the model had actually absorbed it. Most enterprises have none of those yet, which is why they buy tokens.

The challengers have entered the arena. The question is how many buyers are ready to meet them there.

Hope you had a great 4th! Enjoy the update.

-Tara

THE DOWNLOAD

Meta Building Cloud Business to Sell Excess AI Compute

Meta is developing a cloud business, internally called Meta Compute, to sell surplus GPU capacity and hosted access to its closed-weight Muse Spark models, competing with AWS, Azure, and Google Cloud. The move follows SpaceX/xAI leasing Colossus capacity to Anthropic (reportedly $1.25 billion a month) and Google, and aims to recoup Meta’s $182.9 billion in AI infrastructure commitments. Meta shares rose roughly 9%; CoreWeave and Nebius fell about 12% each.

Why it matters: The market read this correctly, and the losers are the neoclouds, not the hyperscalers. Meta is a major customer of CoreWeave and Nebius; if it can serve its own excess capacity, it may not need theirs. The deeper signal is that a company that spent $183 billion chasing superintelligence is now building a fallback that monetizes the infrastructure regardless of whether its own models win.

Unitree Robotics Wins Approval for $619M Shanghai IPO

China’s securities regulator approved Unitree’s STAR Market listing on July 3, clearing the humanoid and quadruped maker to raise 4.2 billion yuan ($619 million) at a valuation near 42 billion yuan. Unitree is rare among robotics firms in being profitable, reporting 1.7 billion yuan in 2025 revenue and 591 million yuan in adjusted profit, and shipped more than 5,500 humanoids last year, first worldwide by volume.

Why it matters: This is mainland China’s first major humanoid listing, and it hands the sector a public-market benchmark at a moment when Chinese embodied-AI funding has hit records of roughly $10.8 billion in 2025 and exceeded $2.9 billion in the first two months of 2026 alone

OpenAI Floats 5% Equity Stake for US Government

OpenAI has held early-stage talks about giving the US government a roughly 5% stake, worth about $42.6 billion at its $852 billion valuation, via a vehicle modeled on the Alaska Permanent Fund. Altman first raised the idea in early 2025 and has pitched a version in which every leading US lab contributes 5%.

Why it matters: The proposal surfaced days after OpenAI delayed GPT-5.6’s launch at the government’s request, and it sits between unfettered private ownership and Senator Sanders’s push for 50%. A voluntary equity giveaway from a company still years from IPO is a novel way to buy political goodwill, and if it sets precedent, foreign governments will demand the same, complicating every cross-border deployment.

Anthropic Restores Fable and Mythos After Export Controls Lifted

The US Commerce Department lifted export controls on Claude Fable 5 and Mythos 5 on June 30, ending a roughly two-week shutdown. Fable 5 returned globally July 1; Mythos 5 came back only for a set of approved US organizations. The freeze began June 12 after Amazon researchers demonstrated a jailbreak; Anthropic pulled both models from every cloud platform because it couldn’t verify user nationality in real time. Anthropic will now pre-release frontier models for federal review.

Why it matters: Mythos 5’s phased return through Anthropic’s Glasswing program establishes a trusted-partner tier sitting between full public access and total suspension. The pre-release review commitment also raises the question of whether Washington now effectively approves every frontier launch.

Microsoft Launches $2.5B Frontier Company for Enterprise AI

Microsoft created Microsoft Frontier Company, a $2.5 billion unit embedding 6,000 engineers inside customers like Unilever, Novo Nordisk, and LSEG to integrate AI, including rivals’ models, with proprietary data, and customers keep the resulting IP. It arrives days after AWS committed $1 billion to a similar embedded-engineering unit and after OpenAI’s $4 billion-backed Deployment Company.

Why it matters: Every major AI vendor has stood up the same forward-deployed model within eight weeks, all aimed at one number: roughly 95% of enterprise AI pilots deliver no measurable result. The battleground is shifting from who builds the biggest model to who can make deployment stick inside a customer’s operations.

Bending Spoons Surges 40% in Nasdaq Debut

Bending Spoons, the Milan firm that acquires and revives legacy software like AOL, Evernote, Vimeo, and WeTransfer, closed up nearly 40% on July 1 at a $25.7 billion market cap, more than double its last private valuation. Revenue grew from $387 million in 2023 to $1.31 billion in 2025, and it swung to a $27.5 million quarterly profit. The company carries roughly $4.4 billion in debt and has flagged 1,000-plus acquisition targets.

Why it matters: A debt-funded acquire-and-fix operator popping 40% in a year investors feared AI would gut traditional SaaS is a real signal about where public appetite sits. Bending Spoons explicitly marketed itself as AI-native (”AI before it was cool”), and buyers rewarded the framing.

AI Designs Radio Chips That Outperform Human Layouts

Princeton researchers used reinforcement learning, inverse design, and diffusion models to generate RFIC layouts that beat human-designed circuits on bandwidth, power, and efficiency while cutting design time from weeks to minutes. The AI-produced structures look “more like a QR code” than conventional symmetric layouts, unconstrained by human templates.

Why it matters: RF chip design has stayed a hand-crafted “dark art” even as CPUs and GPUs went algorithmic, so this is a genuine unlock in one of the last manual corners of semiconductor design. Compressing years of iteration into minutes changes the economics of building custom RF silicon for 5G/6G, autonomous vehicles, and satellite links.

Anthropic Launches Claude Science Workbench

Anthropic launched Claude Science on July 1, a research workbench connecting 60-plus scientific databases with prebuilt genomics, proteomics, and chemistry toolkits, a reviewer agent that checks citations, and full provenance on every figure. It runs on existing Claude models with no new capabilities and no gating, available in beta to all paid subscribers. It contrasts with OpenAI’s gated GPT-Rosalind and Google’s proprietary Gemini for Science.

Why it matters: The bet here is that what slows science isn’t raw model capability but the friction of stitching together dozens of databases and tools, so Anthropic is competing on workflow ownership rather than a specialized model, the same move that made Claude Code sticky. The three labs have now split into distinct strategies: Anthropic on broad access plus a workflow layer, OpenAI on a governed specialist model, Google on proprietary foundation models.

Kioxia Ships 332-Layer Flash Samples for AI Data Centers

Japanese memory maker Kioxia began sampling its 10th-generation 332-layer 3D NAND on July 2, offering 59% higher bit density and a 33% faster interface than its 8th-generation chips, aimed at enterprise and data center SSDs. Kioxia holds roughly 10% of the data center flash market against Samsung’s 40% and SK Hynix’s 30%; shares rose 8.9% after swinging double digits.

Why it matters: The timing matters more than the density gain. This lands with AI flash supply tight and Kioxia’s 2026 output largely sold out, and an Omdia analyst argues its NAND speed is the metric US hyperscalers weigh most. But sampling is not volume: operators typically evaluate for months before ordering, and mass production isn’t slated until 2027, so this is a claim on share, not a shift in supply. Kioxia’s stock, up more than 680% this year, is now priced on the assumption memory tightness holds.

Google Reports 37% Jump in Electricity Use Driven by AI

Google’s 11th environmental report, released June 30, showed a 37% year-over-year rise in electricity demand, its largest ever, pushing total use up more than 250% since 2019. The company matched 100% of consumption with renewable purchases and cut operational emissions 2%, but supply-chain emissions rose 25%, with data center construction alone adding 2.3 million tons of CO2. Google states plainly that its AI buildout is outpacing grid decarbonization.

Why it matters: Operational emissions are addressable with power-purchase agreements. Scope 3, embedded in chips, HBM, steel, and construction across grids in Taiwan, Japan, Vietnam, and India, is not. Every hyperscaler scaling AI compute inherits the same split, and it reframes the energy question from how much clean power a company contracts to where and when its data centers actually draw, and what its supply chain burns to build them.

EVENTS

Join us for cold beer and Hot Chips in the Fall! August 24 @ 530pm in Palo Alto

RSVP

Things in my mental orbit this week:

• I continue to obsess over open source inference infrastructure and the tools and services that will support custom software. Hearing a lot about agencies (some venture-backed) going after “vibe-hardening” - the production layer of vibe-coded tools.

• Compute hardware (both industrial and! consumer) will continue to climb. My poor Macbook is groaning under the weight of running claude code and google meets in its day-to-day. There will be solutions to solve this.

• Very much into the thesis of manufacturing molecules. As AI accelerates the design space (drug discovery, materials discovery), where the differentiation lies is in how these new discoveries are manufactured. More to be announced soon!

• Even as AI takes up a ton of airtime, there’s lots of invention happening in the frontiers - our recent Strange Gatherings dinners had folks committed to pushing the boundaries of fusion, quantum, robotics, and biology.

Now on to the download.

-Tara

THE DOWNLOAD

Government grants Anthropic limited clearance to redeploy Mythos 5

The government cleared Anthropic to bring Mythos 5 back for a vetted set of US critical-infrastructure organizations, partially reversing the June 12 export-control directive that had forced both Mythos 5 and Fable 5 offline. The clearance landed hours after OpenAI agreed to stage GPT-5.6 at Washington’s request.

Why it matters: This is the first working instance of a frontier-model clearance process: the government now decides which named organizations can run the most capable models. Model access is now a licensed privilege granted case by case. Fable 5, the consumer-facing model, is still dark.

Chinese models pass US models in OpenRouter token usage

Weekly OpenRouter usage, surfaced by Christophe Barraud, shows Chinese models near 18 trillion tokens against roughly 5.5 trillion for US models, a reversal from January.

Why it matters: The capability gap in Chinese models has closed enough that scores now sit within a point of the US leaders on real coding tasks, and known researchers have started reaching for these models by choice. Matt Velloso, a former VP at Meta and Google DeepMind, said he'd been using GLM 5.2 "all day" and called it the first open model that passes as a "daily driver."

OpenAI and Broadcom unveil the Jalapeño inference chip

OpenAI and Broadcom unveiled Jalapeño, OpenAI’s first custom inference accelerator, taped out in nine months, with gigawatt-scale Microsoft deployment starting late 2026.

Why it matters: OpenAI used its own models to compress the design cycle, a self-improving loop on the chips that run the models. Broadcom now builds custom silicon for Google, Meta, ByteDance, and OpenAI.

Memory prices keep climbing as Micron sells out HBM and Apple raises prices

Micron posted a strong fiscal Q3 and raised guidance, with gross margins near 81%, its entire 2026 high-bandwidth memory output already contracted, and $22B in long-term commitments locked in. Goldman calls the 2026 DRAM gap the worst in 15 years. Partly due to the industry-wide memory shortage, Apple raised MacBook and iPad prices by up to $300, saying it can no longer absorb the cost. Microsoft followed within hours on Xbox. The iPhone was spared.

Why it matters: Memory has flipped from boom-bust commodity to oligopoly utility. The three makers that produce nearly all the world's DRAM are now pre-selling years of output at 80% margins, with contract prices running up 50%-plus in a single quarter and AI data centers first in line. Apple, with the best supply chain on earth, still couldn’t hold prices due to the squeeze.

(Editor’s Note: expect a deeper dive on Tuesday on this!)

OMB proposes a rule reshaping all federal research funding

OMB’s proposed “Regulation for Federal Financial Assistance” would put political appointees over peer review on discretionary grants, allow mid-project terminations with no appeal, and restrict foreign collaboration. It covers about $1.1 trillion in annual grantmaking, with comments due July 13 and an October 1 effective date.

Why it matters: This hits the layer upstream of deep tech. University labs are where deep tech research originates, and the rule injects termination risk and political review straight into that pipeline.

UPCOMING EVENTS

​Strange Sessions is a monthly demo night where builders share what they’re working on.

​Each session revolves around a theme and features short demos from founders, researchers, and creative technologists.

Come jam

Power moves this week. Two of the eight authors from the seminal “Attention Is All You Need” paper on transformers that reshaped the AI landscape were won over to Big AI companies this week. Noam Shazeer left Google for OpenAI. Ashish Vaswani, the lead author, went to Nvidia.

The biggest power move on my desk this week, though, was /goal. It feels like a cheatcode, honestly. Both Claude code and Codex shipped a version of a slash command where you set a goal, then the agent runs itself in a loop, checking after each turn whether the condition is met and continuing on its own until it is. Hot tip: make sure to plan before building, if a goal is not well defined, it can run in circles for hours and guzzle token spend.

Use:

For this task, write yourself a new goal and spawn agents in parallel — as many as needed to do it better and faster. Split the work into independent pieces, dispatch them concurrently, and synthesize the results as they return. Give each agent its own dedicated /goal.

Have a great week - Tara

THE DOWNLOAD

Senator Says NSA Director Reported Anthropic’s Mythos Breached Most Classified Systems in Hours

Senator Mark Warner, vice-chair of the Senate Intelligence Committee, said on June 11 that the head of the NSA and Cyber Command, General Joshua Rudd, told him Anthropic’s Mythos model “broke into almost all of our classified systems, not in weeks, but in hours,” as reported by The Economist (Digg). The claim lands amid a broader fight: the Pentagon labeled Anthropic a supply chain risk in March, yet the NSA is reportedly running the cybersecurity-focused Mythos for offensive operations with about six Anthropic engineers embedded, and the model now sits under export controls (Tom’s Hardware).

Why it matters: If you strip out the drama, you’l find that most security people read “broke into all our classified systems” as a loud retelling of “found serious holes in our systems,” but no one has shown an AI model breaching air-gapped classified networks on its own (Digg). What isn’t in dispute is the bigger shift. A frontier model is now handled like a controlled weapon, export-limited and deployed inside the NSA because it can find and chain software vulnerabilities faster than human teams can patch them.

Nvidia Hires Essential AI Founder Ashish Vaswani and Several Researchers

Nvidia has hired Ashish Vaswani, founder and CEO of Essential AI and first-listed author of the 2017 Transformer paper, along with several Essential researchers, with Vaswani set to work on Nvidia’s open-source Nemotron models, according to a source close to the startup (Ground Level AI). Several team members have updated their LinkedIn profiles to Nvidia. Essential counted Nvidia, AMD, and Google as strategic investors, and the same source said Vaswani had been struggling to raise more money, with pulling the team away from AMD an added motivator (Ground Level AI).

Why it matters: Nvidia just absorbed the lead author of the paper that created modern AI and pointed him at Nemotron, its own family of open models. Nvidia keeps building more of the software and model layer that rides on top of its chips, and it’s buying the people to do it.

Z.ai Releases GLM-5.2 Open Weights, Tying Claude Opus 4.8 on a Physics Benchmark

Z.ai’s open source model GLM-5.2 ties Claude Opus 4.8 at 20.9% on CritPt, a benchmark of unpublished physics problems, and leads all open-weights models (Artificial Analysis). They released this, pointedly, right after Anthropic was asked to pull back Fable on orders of the US government.

Why it matters: For the everyday coding work most teams actually run, GLM-5.2 now performs like a closed frontier model at roughly a sixth of the price. It trails Opus 4.8 by less than a point on the common agentic benchmarks and beats GPT-5.5 on SWE-bench Pro (VentureBeat), and because the weights are open you can run it yourself and take the per-token bill to zero. The frontier still shows up on the hard cases: on the longest multi-hour tasks Opus pulls about twice as far ahead (LLM Stats), and GLM-5.2 gets roughly a quarter of factual questions wrong (danilchenko.dev).

SpaceX Agrees to Acquire Cursor for $60 Billion in Stock

SpaceX signed a definitive all-stock agreement to buy Cursor parent Anysphere at a $60 billion valuation, four days after its record Nasdaq IPO (CNBC). At roughly $4 billion in annualized revenue and more than a million paying users, this is the largest acquisition of a venture-backed startup on record (TechSpot).

Why it matters: Coding is the killer use case for LLMs, and the single largest category of enterprise AI spending, about $4 billion in 2025 and 55% of all departmental AI budgets, and it has become the gateway into enterprise workflows (Menlo Ventures). Grok had almost no foothold there, so buying Cursor drops xAI inside the category overnight.

Noam Shazeer Leaves Google DeepMind for OpenAI

Noam Shazeer, Gemini co-lead, Transformer co-author, and VP of engineering at Google DeepMind, is joining OpenAI as Lead for Architecture Research, working on next-generation model designs (TechTimes). The move lands about ten days after OpenAI filed confidentially for an IPO that could value it near $1 trillion (MLQ).

Why it matters: The interesting part to note: Shazeer's specialty, mixture-of-experts and multi-query attention, is the set of techniques that make a huge model cheap enough to run without losing money on every answer (TechTimes). That's the exact problem OpenAI has to solve as it heads toward an IPO while still losing money at scale. You can always buy more compute; what you can't buy is one of the few people who can redesign a model to need less of it, which is why Google paid about $2.7 billion to bring him back less than two years ago (ThePlanetTools).

FERC Orders Six Grid Operators to Revise Data Center Connection Rules

FERC issued six show-cause orders under Section 206, giving PJM, MISO, SPP, CAISO, ISO New England, and NYISO 60 days to justify or rewrite how data centers connect to the grid, plus a 30-day report on whether they have enough generation to serve them (Data Center Knowledge). The unanimous orders cover regions serving about 200 million people and are explicitly meant to keep data center costs from landing on other ratepayers (E&E).

Why it matters: Power, not chips, has been the real bottleneck for the AI buildout, and FERC just moved to clear it. But the heart of the order is a fight over who pays. In Maryland, regular customers are already on the hook for about $1.6 billion over the next decade for transmission that mostly serves data centers across the line in Virginia (Utility Dive), and double-digit bill hikes last summer set off real backlash (E&E). FERC’s fix is to push those costs onto the data centers themselves and fast-track the ones that bring their own power or agree to dial back when the grid is tight. The risk runs the other way too: if utilities overbuild for demand that never fully arrives, it’s ratepayers, not developers, who get stuck with the stranded plants (TechTimes).

Google DeepMind Publishes an AI Control Roadmap Treating Its Own Agents as Insider Threats

DeepMind released a 35-page AI Control Roadmap, a defense-in-depth framework that treats internal AI agents as potential insider threats rather than trusting alignment alone (DeepMind). It introduces TRAIT&R, a taxonomy of rogue tactics modeled on MITRE ATT&CK across four detection and three response tiers, and reports an analysis of more than a million coding-agent tasks where most flagged behavior traced to overeager agents, not malice (Fortune).

Why it matters: One of the top AI labs just admitted, in writing, that it can't count on its own agents staying aligned, so it now treats them like employees who might go rogue: limited access, constant monitoring, the ability to pull the plug. The useful part is the data behind it. Across a million internal coding tasks, almost all the flagged problems were overeager agents overstepping their instructions, not scheming ones (TechTimes), so the thing companies actually need to catch is good intentions gone too far, not sabotage (for now!)

Midjourney Launches a Health Division and a Whole-Body Ultrasound Scanner

Midjourney, the generative-AI image company, launched a health division and a whole-body “Ultrasonic CT” scanner at a San Francisco event, with a first “Midjourney Spa” clinic planned for late 2027 (RDWorld). Founder David Holz claimed a roughly 60-second scan with image quality in some ways superior to MRI, with no radiation or magnets (The Next Web). The imaging is not Midjourney’s generative AI; it runs on Butterfly Network’s ultrasound-on-chip technology, licensed in November 2025 for up to $74 million over five years.

Why it matters: This is what a founder with no investors does. Midjourney is bootstrapped, profitable, and answers to no board, so Holz can sink tens of millions into a decade-long bet that has nothing to do with the image business. And it fits him: he spent twelve years running a body-tracking hardware company (LeapMotion) before this, and has signaled for a while that Midjourney's real ambition is hardware and modeling the physical world.

DEEP DIVE FROM THE REVIEW

Coding is, by far, the killer use-case for LLMs. Claude Code is arguably one of the primary, bottoms-up drivers of Anthropic’s revenue growth in the last 18 months.

Cursor is SpaceX’s bet on building the same thing. This cements something I’ve been noodling over since the OpenAI / Nvidia deal last September:

If you want to be a Big AI company, you have to go full stack. I break it down here:

EVENTS

Strange Sessions #5: Loops

​Strange Sessions is a monthly demo night where builders share what they’re working on.

​Each session revolves around a theme and features short demos from founders, researchers, and creative technologists.

Come jam

Strange Gatherings: The Inventors’ Table

We’re absolutely at capacity for Strange Gatherings: The Inventors’ Table at SF Deep Tech Week, but if you sign up for the waitlist, we’ll invite you to the next one! We’ hope to get a bigger venue next year.

Request an Invite

Anthropic went from $87M ARR in January 2024 to $47B ARR in May 2026, roughly 540x growth in 28 months. That growth had two engines working in tandem: top-down enterprise partnerships (Claude is the only frontier model on all three major clouds) and bottom-up developer adoption through Claude Code, the fastest-growing product in the company’s history, rocketing from $0 to $2.5B ARR in 9 months. Anthropic now owns 54% of the enterprise AI coding market.

Cursor is SpaceX’s bet on building the same thing.

Yesterday, SpaceX announced its acquisition of Anysphere, the company behind Cursor, the AI coding tool used by 7 million developers daily, for a whopping $60B in stock. Cursor was founded just four years ago out of MIT, hit $2B ARR, and became the highest-revenue AI coding tool in the category. Its share has been slipping (from 41% to 26% over the past year as Claude Code surged), but what xAI is buying isn’t market share.

xAI already had the full stack: Colossus (compute), Grok (model), X (app). But X is where people scroll. Cursor is where they build. This would yield probably one of highest-signal training data in AI, and exactly what you need to make Grok competitive.

This cements something I’ve been noodling over since the OpenAI / Nvidia deal last September:

If you want to be a Big AI company, you have to go full stack.

It’s only becoming more apparent. Better products drive better infrastructure (more data), and better infrastructure drives better experiences (this has always been core to our investment thesis at Strange).

Going full stack does two things:

1. Unit economics of building and training AI become sustainable.

2. You get proprietary training data from the application layer that differentiates you from model competitors. User data and workflow lock-in become an attractive competitive moat.

We’ll probably see movements over the next few years:

• Model companies building their own applications from the inside out, or

• Aggressively acquiring up the stack into the application layer

There’s a bit of a saying amongst founders now that because it’s become 10x easier to build, companies have to be 10x more ambitious to succeed. Seems like we’re seeing that all across the board.

- Tara

The finite game of being the frontier AI model just got even more finite. A few hours after I wrote a breakdown on Mythos / Fable and how the economics of building frontier intelligence are being squeezed (the models are immediately distilled by Chinese companies upon release, which made Anthropic baked in a safety layer that made Fable 4x more expensive than the previous model, Opus). Then, in a strange turn of events, the US government pulled the plug on the model entirely, citing export controls, likely due to suspected Chinese access to Mythos. Since Anthropic couldn’t KYC everyone accessing the model, they removed access for all users overnight.

This is unprecedented. And it makes me think of a few things:

1. US frontier models may be forced to KYC every developer and user. That is probably a death knell to user growth.

2. Because of economic and political pressures on model training, I predict closed-source model companies will start to aggressively move up the stack toward the application layer. They can easily beat out competition on the app layer with token unit economics alone.

3. The ecosystem and inference infrastructure around supporting multiple models will be more important than ever. If you’re building in this space, please reach out!

-Tara

THE DOWNLOAD

US Government Orders Anthropic to Disable Fable 5 and Mythos 5

The Commerce Department issued an emergency export-control directive Friday evening barring foreign-national access to Anthropic’s Fable 5 and Mythos 5, launched three days earlier. Because Anthropic cannot screen citizenship in real time, it pulled both models for every customer worldwide. The directive extends to Anthropic’s own foreign-born staff; Reuters notes that co-founder Chris Olah, researcher Andrej Karpathy, and philosopher Amanda Askell were all born outside the US, and Anthropic declined to say whether they have been locked out of the models they help build. David Sacks called the jailbreak “easily resolved” and denied any link to prior disputes; Semafor reports Amazon’s CEO Jassy tipped off the government and the real trigger was suspected Chinese access to Mythos.

Why it matters: This is the first time the federal government has taken a publicly deployed commercial model offline. It sets a precedent: any model accessible via API can be shut down for all users, worldwide, in hours, by directive.

SpaceX Completes Largest IPO in History at ~$2.2 Trillion

SpaceX listed on Nasdaq Friday at $135 per share, the largest IPO in history, surged past $170, and reached a market cap of roughly $2.2 trillion, passing TSMC and Saudi Aramco to become the sixth-most-valuable public company. Investor demand was nearly 4x the $75 billion offering.

Why it matters: Musk successfully pivoted the SpaceX narrative from space to an AI infra company in the months before listing. The S-1 frames three business lines: rockets, connectivity, and AI compute. SpaceX has existing compute contracts with Alphabet and Anthropic, COO Gwynne Shotwell told CNBC SpaceX “100 percent” sees itself competing with the neoclouds, and Musk’s comp is tied to delivering space-based AI data centers (even if whether orbital compute actually works at scale is unproven).

Apple Rebuilds Siri on a Licensed Google Gemini Model

At WWDC this week, Apple unveiled “Siri AI” running on a custom 1.2-trillion-parameter Gemini model licensed from Google for a reported ~$1B/year. iOS 27 will also let users route Siri requests to third-party chatbots including Claude and Gemini via a new Extensions system. This was Tim Cook’s final WWDC as CEO before handing the role to hardware chief John Ternus on September 1.

Why it matters: The most valuable consumer-hardware company is choosing to lease its frontier model from a direct rival. The third-party Extensions system goes further: Apple is explicitly positioning as an orchestration and distribution layer above the models, not a model builder. That hands Google a recurring revenue stream inside every iPhone. (Editor’s note: we predicted this nearly 10 months ago!)

OpenRouter Launches Fusion API

OpenRouter launched Fusion, a system that runs a panel of different LLMs against the same prompt in parallel, then passes their outputs to a judge model that synthesizes a single answer. In benchmarks on Perplexity’s DRACO deep-research suite, Fable 5 and GPT-5.5 fused together scored 69%, beating every individual frontier model tested. A same-model panel (Opus 4.8 paired with itself) still jumped 6.7 points over Opus 4.8 alone.

Why it matters: The biggest gains came from the synthesis step itself, not from mixing different model architectures. This means that a thin orchestration layer that recombines outputs from commodity models can match or beat a single frontier model at lower cost.

Chan Zuckerberg Biohub Releases Open-Source ESMFold2

The Chan Zuckerberg Biohub released ESMFold2, an open-source protein AI model trained on 2.8 billion sequences, alongside the ESM Atlas: 1.1 billion predicted protein structures and 6.8 billion sequences, exceeding AlphaFold’s database by over 800 million entries (Nature). This is not just a prediction tool. Biohub used ESMFold2 to computationally design protein binders against five cancer and immunology targets (including PD-L1 and EGFR), then validated them in the lab, matching the mechanism of approved therapies (Biohub).

Why it matters: The step from prediction to design-and-validate is the one that matters commercially. DeepMind’s Isomorphic Labs is building a business around that exact transition behind a closed model. ESMFold2 ships it open-source with lab-validated therapeutic results, which compresses the moat to downstream wet-lab data, proprietary targets, and clinical execution rather than model access.

Amazon Discloses 2.5 Billion Gallons of Data-Center Water Use

Amazon reported its data centers withdrew about 2.5 billion gallons of water globally in 2025, its first absolute figure, roughly 5% of metro Seattle’s annual use (Bloomberg). The company claims 7x better water efficiency than the industry average, though that comparison measures its full fleet against Google’s AI-specific facilities.

Why it matters: The 2.5B figure is operational water only and excludes the far larger volume embedded in power generation. With EPA rules and eight states tightening data-center water regulations in 2026, water is becoming a real siting and permitting constraint on the AI buildout, on par with power and land.

DEEP DIVE FROM THE REVIEW

EVENTS

As part of Deep Tech Week, Strange Ventures is hosting a private dinner for a small group of scientists , researchers, and founders.

​We’ll gather around a Jeffersonian table to discuss the breakthroughs, discoveries, and ideas that may shape the next decade.

Request an Invite

Strange Sessions #5: Loops

​Strange Sessions is a monthly demo night where builders share what they’re working on.

​Each session revolves around a theme and features short demos from founders, researchers, and creative technologists.

​Early ideas welcome.

Come jam

It’s a question I’ve been slightly obsessed with for the past few months. I’ve built automations that wake up every morning and ask me what’s on my mind. I’ve wired agents to monitor the news and text me when something interesting happens. I’ve been building an eval agent that gathers feedback on its own work and builds a backlog of what to do next.

I’m not alone. Boris Cherny, who runs Claude Code at Anthropic, half-jokes that his main job now is to write the loops that prompt Claude.

Addy Osmani named the discipline: loop engineering. Design a system that discovers work, executes it, verifies the results, updates its own instructions, and repeats. Give an agent goals, skills, and permission to spin up subagents. Then let it loose while you go do something else (like doom scroll).

At least that’s the pitch. In practice, my loops meander constantly. Token costs balloon. Agents get stuck. One bad decision dominos until the system is grinding away, very confidently, in the wrong direction. And every time, I wondered if it was because the model isn’t smart enough. Maybe wait for the next release.

I think our recent obsession with loop engineering made me lean more towards

A loop is a collection of jobs:

• Orchestration

• Memory

• Verification

• Retrieval

• Routing

In practice, it could look something like this:

• Automations discover new work.

• Worktrees allow multiple agents to operate in parallel.

• Skills capture project-specific knowledge.

• Connectors give agents access to real systems.

• Sub-agents verify each other’s work.

Persistent memory keeps state between runs.

Subscribe now

Frontier intelligence matters most in two places:

1. orchestration, knowing when to break a problem apart or when to stop, and

2. self-modification, where the system rewrites its own instructions

Roughly fewer than 10 percent of the tasks in my loops genuinely need frontier reasoning. The rest need better systems: memory design, retrieval structure, verification gates, context architecture.

How does this change the economics of the frontier, which seem increasingly resource-intensive to create?

Frontier intelligence is valuable and scarce. But increasingly it seems like moat it really has is a short window of time.

Distillation dupes frontier capability within months at a fraction of the cost.

The labs know this. This week, when Anthropic shipped Claude Fable 5, it did so with anti-distillation defenses engineered into the model itself. Suspected distillation attempts get rerouted to an older model. Requests related to frontier model development were quietly degraded by invisible safeguards, a choice the company walked back after backlash.

They care because their edge is temporal. While open models will likely never be first (they don’t have the copious amount of proprietary user data), they can always catch up to the latest, sometimes in weeks.

This is maybe what makes the frontier race a finite game.

Each generation is a round: ship the model, monetize the premium window, watch the capability get replicated, ship again. The labs can never stop playing. And the rounds get harder to win profitably.

The volume underneath, meanwhile, explodes. And everyone starts feeling the burn of cost. Citadel’s Economics of Intelligence report goes into how AI adoption is becoming less about what models can do and more about the price of running them at scale.

A lot of VC doomisms seems to be about how the frontier models are just going to build it all and that there’s nothing left to build. I think it’s the opposite: the frontier intelligence race will stay scarce and valuable. But it is playing a finite game: the labs have to win every round to stay in it.

I think the better position to be in would be those building the inference engineering infrastructure that runs underneath it all.

-Tara

EVENTS

Join us for Strange Sessions #5: Loops.

Strange Sessions is a monthly demo night where builders share what they’re working on.

​Each session revolves around a theme and features short demos from founders, researchers, and creative technologists.

​Early ideas welcome. Come jam.

RSVP

We talk about AI like the hard problem is reasoning or tool use, but the thing that actually makes something truly useful is context and continuity. They remember what you said last time. They update without being asked.

I think a lot about how to program agency into the agents I’m building. Is it more tactical like setting up chron jobs for pings, reminders, or workflows? Is it more systematic like building self-improvement loops with eval agents and PM agents set up to gather feedback, success metrics, and propose new features? You probably need some version that combines both modes.

My kids have started picking up chess in the last few months. And it’s really interesting watching them build fluency in two modes of thinking. When they play logic-based games on the iPad, they speed through fast feedback, practice reactive thinking, instant pattern matching. When they play chess on the board, it’s much slower, more strategic. You sit with a position. You think about what happens if you do nothing.

I think fluency in both modes of thinking - the fast reactive one and the slow strategic one - is the thing to build muscle in the next era. For founders building and improving these systems, and for the five-year-old staring at a board right now trying to decide whether to move his knight.

— Tara

THE DOWNLOAD

SpaceX Signs $920M Monthly Compute Deal with Google Through 2029

SpaceX disclosed in an SEC filing that Google will pay $920 million per month from October 2026 through June 2029 for access to roughly 110,000 NVIDIA GPUs, CPUs, and memory at data centers inherited from xAI, which SpaceX absorbed in an all-stock merger earlier this year. The deal totals approximately $30-32 billion over its life. It follows a late-May arrangement in which Anthropic committed $1.25 billion per month for all available compute from SpaceX’s Colossus 1 facility near Memphis.

Why it matters: Combined with the Anthropic deal, SpaceX is now on track for roughly $2.17 billion per month in AI compute revenue, an annualized run rate of ~$26 billion. That would make it one of the largest GPU lessors on the planet. Google, one of the world’s largest owners of AI infrastructure, called this “bridge capacity” for surging Gemini Enterprise demand it cannot meet with its own data centers. The disclosure lands one week before SpaceX’s June 12 Nasdaq debut, converting what looked like an xAI cost center into a concrete, multiyear recurring revenue stream.

Anthropic Publishes Internal Data on AI Automating Its Own Development And Calls For Global Pause (Again)

Anthropic’s research arm published “When AI Builds Itself,” disclosing that over 80% of code merged into its codebase is now written by Claude, and that its engineers ship roughly 8x as much code per quarter as they did from 2021-2025. The company said Claude’s success rate on open-ended engineering problems reached 76% in May 2026, up 50 percentage points in six months, and that Claude-written code quality is now roughly at parity with human-written code at Anthropic.

Why it matters: This is the first time a frontier lab has published hard internal metrics on how much of its own development loop is automated. Anthropic framed the trajectory as pointing toward “recursive self-improvement,” where AI systems design their own successors, and called for the option to pause frontier development globally.

NVIDIA Launches Cosmos 3 Open Foundation Model for Physical AI

NVIDIA released Cosmos 3 at GTC Taipei, an open model built on a mixture-of-transformers architecture that unifies vision reasoning, world simulation, and action generation in a single system. It was trained on 20 trillion tokens of multimodal data including action data from humans and robots. Available on Hugging Face in Nano (8B, runs on workstation GPUs) and Super (requires Hopper/Blackwell) variants.

Why it matters: Physical AI teams currently stitch together separate vision, simulation, and policy models. Cosmos 3 collapses those into one. The open release, with a Cosmos Coalition including Skild AI, Runway, and Black Forest Labs, extends NVIDIA’s platform strategy beyond silicon into the model layer for robotics and autonomous systems.

OpenAI Rolls Out Dreaming V3 Memory Architecture for ChatGPT

OpenAI launched Dreaming V3 on June 4, replacing ChatGPT’s manually saved memory list with a background synthesis process that continuously reads across past conversations and updates what the system knows about a user. Memories now self-update over time and are stored in a separate data layer injected at inference.

Why it matters: Persistent, self-updating memory turns a stateless chatbot into something closer to an ongoing relationship, which is the retention mechanic consumer AI products need. OpenAI’s internal evals show factual recall rising from 41.5% to 82.8% between the 2024 and 2026 memory systems. The privacy concerns expands significantly: memories are now synthesized without explicit user action, and the audit trail is less transparent than a simple saved-facts list.

MIT Publishes Categorical Framework for Self-Evolving AI Scientists

MIT researchers Buehler and Wang published a framework (arXiv:2606.01444) for AI systems that can expand their own scientific reasoning schemas, moving from search within a fixed vocabulary to genuine discovery of new concepts. Case studies demonstrated the approach in protein mechanics and fiber-network materials, using category theory to formally verify that the system’s schema actually expanded.

Why it matters: Most AI science tools search a space humans defined. This framework formally distinguishes between retrieval, search, and discovery, and provides a mathematical proof that the system entered a new reasoning regime. If the approach generalizes, it changes how autonomous labs are built: the system doesn’t just run experiments faster, it redefines what experiments are worth running.

Stephen Wolfram Applies Computational Ruliology to Game Theory

Wolfram published a research essay systematically exploring what happens when programs compete in iterated games. Rather than studying individual strategies, the piece exhaustively enumerates all possible strategies as finite state machines, cellular automata, and Turing machines, then ranks them. The central question: does competition tend to produce complexity or simplicity in winning strategies?

Why it matters: The framing is directly relevant to autonomous agent competition. As AI agents increasingly interact in markets, negotiations, and resource allocation, the structure of winning strategies matters. Wolfram’s earlier work on biological evolution and machine learning found that simple programs can produce complex behavior; now it asks whether adversarial pressure changes that dynamic.

SemiAnalysis Roundup: Post-Copper Interconnects, CFET Progress, and SK Hynix V9 NAND

SemiAnalysis published its IEDM 2025 deep dive, covering ruthenium and molybdenum interconnects as copper hits scaling limits, SK Hynix’s 321-layer 3D NAND with 5-bit-per-cell architecture, and progress toward 1,000 CFET transistors and 2D channel materials.

Why it matters: The interconnect transition from copper to ruthenium at the tightest metal layers, potentially as soon as the 14Å node, will reshape equipment and materials supply chains. SK Hynix’s 5-bit-per-cell NAND directly addresses the capacity crunch driven by AI datacenter demand. These are the physical-layer bets that determine whether Moore’s Law keeps delivering for AI scaling.

EVENTS

Strange Ventures is hosting a private dinner for a small group of scientists , researchers, and founders at SF Deep Tech Week.

We’ll gather around a Jeffersonian table to discuss the breakthroughs, discoveries, and ideas that may shape the next decade.

Request an invite

Anthropic releases Opus 4.8 with Dynamic Workflows, putting 1,000-agent orchestration inside Claude Code

Anthropic shipped Claude Opus 4.8 on May 28, just 41 days after Opus 4.7 (usually it’s a three-to-seven-month cadence between releases), reportedly accelerated by the lukewarm reception to 4.7 and the same-window launches of OpenAI Codex and Google Gemini Flash. The headline feature is Dynamic Workflows, which lets Claude Code spin up to 1,000 parallel subagents in the background, with one early benchmark showing a 750,000-line codebase migration completed in 11 days. Anthropic also reports Opus 4.8’s misalignment rates are now close to Claude Mythos Preview, the inner-circle model it currently restricts to Project Glasswing partners.

Why it matters: The agent-coding category has been built on peer-coordination harnesses (e.g. the OpenAI Swarm, CrewAI, AutoGen pattern) where success depends on agents talking to each other inside an orchestrator's loop. Anthropic just shipped a structurally different primitive: hierarchical self-correction, where verification and repair are embedded at every task node rather than bolted on at the end. That's what makes long-horizon, high-fan-out work trustworthy without a human in the loop.

Unitree files for Shanghai IPO at ~$6.2B valuation

Unitree’s IPO ambitions goes to listing review on June 1 at a target valuation of roughly 42 billion yuan ($6.2B). The prospectus shows revenue of 1.17 billion yuan (~$165M) in the first three quarters of 2025 and a 60.13% gross margin on its core businesses, with humanoid robots rising from 1.9% of revenue in 2023 to 51.5% in 2025 and 5,500 units sold last year.

Why it matters: Unitree's prospectus is the first transparent look at humanoid unit economics, and it makes Western valuations hard to square. Figure AI sits at $39B, Apptronik raised at $5B in February, Physical Intelligence is reportedly raising toward $11B. Unitree shipped more humanoids in 2025 alone than Figure has shipped in its history, at a tenth of the price (G1 from $16,000 versus Digit at ~$250,000), and is targeting 20,000 units in 2026.

Huawei unveils LogicFolding, a path to 1.4nm-class density without ASML dependence

At ISCAS 2026 in Shanghai, Huawei introduced LogicFolding, a dual-layer chip architecture that distributes logic gates across vertically stacked wafer layers connected by 1.5-μm hybrid bonding, framed by a “Tau Scaling Law” that prioritizes signal speed over transistor shrinkage. The company claims a 55% increase in transistor density and 41% power-efficiency gain, with a target of 1.4nm-class density by 2031 — without EUV lithography. The first commercial deployment is the Kirin chip in the Mate 90 series. Two days later, Peking University showed a true-3D EDA prototype tailored to LogicFolding that reports a 30% reduction in internal wire length in early tests, addressing the EDA gap that Huawei’s own president flagged as the primary obstacle alongside thermal management.

Why it matters: Sanctions are now shaping chip architecture in China. That forces a parallel domestic EDA stack (Empyrean and Peking University’s tools) and a packaging stack to mature alongside the architecture, meaning the Chinese semiconductor supply chain is decoupling at the design-tool layer, not just at the foundry. It’s worth paying attention to the China-stack capex tailwind: a credible non-EUV path keeps domestic packaging, advanced bonding, and 3D EDA tools spending well above what export controls were designed to choke.

DEEP DIVE FROM THE REVIEW

Inference platform revenues have gone hyperbolic in the last year. A small number of AI applications, or token whales, are suddenly consuming infrastructure at industrial scale.

Fireworks climbed from roughly $280M to $800M ARR (notably excluding revenues from what might be their biggest customers, Cursor). Together is reporting 10x YOY revenues, including a $1B contract with a token whale customer. Modal reportedly grew 5x from roughly $60M to $300M ARR since September last year.

The interesting part is that it wasn’t just one company that broke out. Companies across the inference stack all started showing the same shape at the same time: more tokens, higher throughput, larger workloads, steeper revenue curves.

And it doesn’t appear to be driven by exponentially more new users showing up. It appears to be driven by the same users consuming radically more compute.

This is a strong indicator that AI workloads are clearly starting to hit an inflection point. The shift appears to be driven by agents and reasoning-heavy workloads, which can consume hundreds of thousands, or sometimes millions, for a single task.

Baseten said inference volume grew 100x in a year. Google reported Gemini processing 16 billion tokens per minute, up 60% quarter-over-quarter. AWS said Bedrock processed more tokens in Q1 than in all prior years combined.

The Token Whales

Inference demand appears to be far more concentrated than most people realize.

A small number of AI-native applications are likely driving a disproportionate share of inference spend, especially in coding.

Fireworks explicitly excludes Cursor revenue from some reported figures. Together disclosed a single customer contract worth more than $1B. Most of their other enterprise contracts reportedly sit closer to the $1 million mark.

There are clearly a few token whales materially distorting the curve, and who partner with multiple platforms to gobble up compute.

And these token whales have made a strong shift toward open-weight models over the last 12 months, away from closed frontier APIs. Most likely for economics. Spend that previously flowed directly to OpenAI or Anthropic is increasingly being routed through open models plus specialized inference infrastructure.

Cursor’s coding model Composer runs on Kimi K2.5, an open-weight Chinese reasoning model served through Fireworks. Open-weight models reportedly went from less than 2% of OpenRouter traffic to more than 45% in roughly a year.

This helps explain why the independent inference layer suddenly looks like it’s experiencing explosive growth.

While the hyperscalers are still vastly larger (Microsoft, Google Cloud, and AWS each operate AI cloud businesses that dwarf the entire independent inference ecosystem) the largest AI-native applications increasingly aren’t defaulting to hyperscaler-managed inference stacks.

They want tighter control over latency, routing, serving, model behavior, and ultimately economics. The more inference becomes core product infrastructure, the less attractive generalized platforms become.

That created the opening for companies like Fireworks, Together, Modal, Baseten, and others who are willing to deeply optimize around a customer’s specific workload (like Cursor’s Fireworks partnership for RL inference) instead of offering one-size-fits-all managed infrastructure.

But the boom also has an Achilles heel.

The AI application whales driving this explosion in inference demand could ultimately determine the fate of the inference platforms themselves.

The bull case is that inference platforms become the Snowflake of the AI stack: specialized infrastructure layers that sit above commodity compute and build enduring value through performance, optimization, and workflow depth.

The bear case is that the largest AI-native applications eventually internalize those capabilities themselves. Owning inference increasingly looks like the next logical layer of optimization.

Which makes this moment feel both very real and strangely fragile at the same time.

The inference boom is real. But so far, it appears to be powered by a surprisingly small number of AI-native token whales, customers large enough to eventually build the inference stack themselves.

It seems like frontier research is being disrupted as rapidly as SaaS startups.

This week, OpenAI announced that a general reasoning model disproved an 80-year-old conjecture in geometry. Field Medalist Tim Gowers, in a companion paper, described the result as “a milestone in AI mathematics. In the same week, two papers landed in Nature on multi-agent science systems. DeepMind's AI co-scientist, built on Gemini, generated and ranked hypotheses for drug repurposing in acute myeloid leukemia that then held up in the lab. And FutureHouse's Robin proposed a drug-disease link no one had drawn: ripasudil, a glaucoma drug, as a candidate treatment for the leading cause of blindness in the developed world.

This is a timeline that belongs more to product launches, not frontier breakthroughs.

At first I wondered if the model’s breakthroughs were due to brute force, or a ton of human hand-holding. But it’s neither. The chain of thought is a search that diagnoses its own failures and lets each one point to the next move. That’s domain-specific judgment about where to look, the thing we usually call intuition.

This heuristic search parallels how the scientific community works, only slower and spread across people.. grad students, arXiv, citations: these are a giant decentralized version of exactly this loop, propose, fail, diagnose, redirect, propagate.

Except, that the model ran the whole thing in days, rather than over decades.

Reading through how the model steers itself made me wonder:

Language is thinking, and thinking is language.

It’s tempting to call chain-of-thought narration, the machine describing its thinking out loud. I almost think it’s less narration but actually the thinking itself. The language isn’t describing the search. The language is the search.

It turns out that language parroting, a fluent recombination of everything humans have written, is a shockingly good engine for thinking.

So if machines are about to outrun the human method for discovery, do we have to do what SaaS startups have forced their incumbents and now themselves to do, and flip our mental models?

At the (wonderful) SAIR summit last week, Gowers gave a keynote about what he calls motivated proofs. The idea is that AI shouldn’t just produce correct proofs, it should produce transparent ones, built from explicit, inspectable moves, the kind a good teacher uses. The goal is to for AI to create proofs we can learn from, not just verify. Make it human-legible.

Is there where the human’s role in search is heading? To direct, but then to reverse engineer discovery?

To take the machine’s correct, alien output and reverse-engineer the understanding out of it. Find the insight. Extract the technique. Turn the answer into something that can carry to the next problem.

Maybe the eureka moment is obsolete.

Or at least how we traditionally romanticized it as the moment of discovery. Maybe in the near future it lives more at the moment of understanding or distilling what was discovered.

This week, it’s very much felt like something has broken through in the frontiers of research.

Discovery is very much psychological. The frontier isn’t always about an insurmountable wall a wall of difficulty. It’s more often a wall of imagination, a seemingly impossible task that turns obvious the instant someone shows you the trick.

A friend brought up the fable of Columbus and the egg. At a dinner, Columbus hands guests a boiled egg and asks them to stand it on its end. They try, they fail, they give up. Then he taps it on the table, flattening one end just enough, and balances it successfully.

The solution is obvious, once you’ve seen it done. The barrier was never the difficulty. It was the inability to see it could be done that way at all. To many, maybe especially the people closest to the craft will be the ones most tempted to say that doesn’t count. That it isn’t real discovery unless it arrives the way ours always has.

Regardless, from here onwards, the machine will keep tapping the egg on the table.

Thank you to T, M, S, B, J, and J from bouncing thoughts with me.

Have a great memorial day weekend.

Tara

THE DOWNLOAD

The biggest news to pay attention to this week

Andrej Karpathy joins Anthropic to work on pre-training

Andrej Karpathy, an OpenAI co-founder and former Tesla AI lead, announced on May 19 that he has joined Anthropic. He is reporting to pre-training lead Nick Joseph and building a team focused on using Claude to accelerate pre-training research, the stage where models acquire their core capabilities.

Why it matters: Putting one of the field's best-known researchers on AI-assisted pre-training is a bet that the next gains come from models improving how models are trained, rather than from raw compute alone. If it works, the edge shifts to whoever gets models improving their own training, and spending alone stops being enough to stay at the frontier.

Alibaba’s Qwen3.7-Max optimized a kernel for Alibaba’s own chip across a 35-hour autonomous run

Alibaba released Qwen3.7-Max, a model built for long-running agent work, and used it to write a kernel (a small program that runs one operation e.g. a transformer’s attention step as efficiently as possible on a given chip) for Alibaba’s own T-Head ZW-M890 accelerator. When pointed at a chip it had not seen in training, the model ran on its own for roughly 35 hours, compiling, testing, and rewriting the code across about 1,158 tool calls and 432 kernel evaluations. Alibaba reported the result ran roughly 10x faster than the standard open-source version it started from, though these are the company’s own figures and have not been independently reproduced.

Why it matters: A chip runs AI workloads only as fast as the kernels written for it, and that software layer takes scarce compiler engineers years to build for each new chip. This is a large part of the reason why Nvidia’s CUDA ecosystem is harder to replicate than the hardware, and why competing accelerators tend to fall short on software rather than specs. A model that can write a competitive kernel for an unfamiliar chip points to a cheaper path to making new silicon usable.

OpenAI’s reasoning model disproved an Erdős conjecture from 1946

OpenAI reported that a general-purpose reasoning model produced an original proof disproving the planar unit distance conjecture, a question Paul Erdős posed in 1946, by finding a new family of constructions that beats the long-assumed square-grid optimum. The result was reviewed and endorsed by mathematicians Noga Alon, Melanie Wood, and Thomas Bloom, and Fields Medalist Tim Gowers said he would recommend it. Notably, this is the same group that called out OpenAI’s overstated Erdős claim seven months ago, which Bloom had described as a dramatic misrepresentation.

Why it matters: The proof came from a general reasoning model, not a system built for mathematics, which undercuts the thesis that frontier scientific work needs purpose-built, narrowly trained tools. If general models can reach original results in the hardest reasoning domains, the case for funding deep, vertical AI-for-science solutions weakens against say, scaling compute on the frontier labs' general models.

Nature publishes DeepMind and FutureHouse AI research agents

Nature published peer-reviewed work on two multi-agent research systems, DeepMind’s Co-Scientist and FutureHouse’s Robin. Robin identified ripasudil as a repurposing candidate for dry age-related macular degeneration, and Co-Scientist surfaced repurposable leukemia drugs and liver fibrosis targets, all still requiring preclinical validation. Edison Scientific, the commercial spinout of FutureHouse, also announced a partnership with Incyte to embed its Kosmos platform across the drug discovery and development lifecycle.

Why it matters: The gap between an academic demonstration and a commercial deployment is rapidly shortening. A method published in Nature this week is already being sold into a public pharma company's R&D stack, and I predict we will soon see frontier research methods looped into commercial R&D almost as fast as they publish.

Google centered I/O 2026 on agents, spanning consumer, developer, and science tools

At I/O 2026, Google shipped Gemini 3.5 Flash to general availability the day it was announced, making it the default worker model across Search, the Gemini app, and enterprise products, with Pro held until June. It launched Antigravity 2.0, its agent-first development platform, where it demoed agents building a working operating system in 12 hours, plus Gemini Spark, a 24/7 personal agent on cloud VMs, and Gemini Omni for video and world simulation. It also introduced Gemini for Science, a research suite, and open-sourced Science Skills, a bundle connecting agentic platforms to more than 30 life science databases including UniProt and the AlphaFold Database.

Why it matters: Google is leaning hard on the one advantage rivals can't quickly copy: distribution. By shipping the fast, cheap Flash tier first and holding Pro back, it seems like they are banking on speed, reach, and reliable environments where agents actually run in. With 1 billion AI Mode users, 900 million on the Gemini app, and 3 billion Android devices, this give Google insane reach and feedback data that no standalone startup or model can match yet.

The US Commerce Department is taking equity in nine quantum companies for $2 billion

The Commerce Department signed letters of intent for $2.013 billion in CHIPS Act incentives across nine quantum companies, taking a minority equity stake in each as a condition of the award. IBM receives $1 billion to launch Anderon, a standalone 300mm quantum wafer foundry it is matching dollar-for-dollar; GlobalFoundries receives $375 million; and D-Wave, Rigetti, Infleqtion, PsiQuantum, Quantinuum, and Atom Computing receive roughly $100 million each across every major hardware modality.

Why it matters: The US government is reusing the same ownership-plus-support template it applied to rare earths to quantum, calling it a “critical frontier technology”. The stake buys supply-chain security and influence in the domestic quantum ecosystem, which they believe has “significant implications for national defense, advanced materials and biopharmaceutical discovery, financial modeling, and energy systems”.

This week, the push-and-pull with China continues. In Trump’s trip to Beijing this week, they brought an “offering” of H200 Nvidia chips - which Xi declined. China’s going all in on building their own, putting in billions to acquire domestic foundries and vertically integrating its domestic chip stack. Meanwhile, Chinese developers find ways to access OpenAI and Anthropic models for cheap through proxy stations, a grey market workaround to get US frontier models at Deepseek prices.

Wafer chip maker Cerebras nearly doubles on IPO open, signaling a massive appetite for AI stocks. SpaceX is coming up next in June, and with a $1.75T valuation, set to be the largest IPO of all time.

Anthropic and OpenAI continue their advance up through the stack into the application layer. This week, OpenAI announced integrations into the personal finance stack, to be able to advise users on everything from investments to mortgage to budgeting. Anthropic went the small business route, releasing Claude for Small Business to help with invoicing, payroll, and more. I predict we’ll see many more moves in like categories like taxes, accounting, legal docs, healthcare benefits, insurance by the end of this year.

Enjoy the update.

Tara

THE DOWNLOAD

H200s approved for China, but Beijing says they “will develop their own”

As Trump met Xi in Beijing on May 14, Reuters reported the U.S. Commerce Department had cleared roughly 10 Chinese firms (Alibaba, Tencent, ByteDance, JD.com, and others) to buy up to 75,000 H200s each through Nvidia or distributors like Lenovo and Foxconn, with a 25% revenue share to the U.S. Treasury and chips routed through U.S. territory before delivery. By the time the summit closed Friday, Trump publicly acknowledged Beijing “chose not to” approve purchases because “they want to develop their own.”

Why it matters: Three days earlier, the Shanghai Stock Exchange approved SMIC's ~$6B acquisition of the remaining 49% of its Beijing 300mm fab from five state-backed investors, the largest semiconductor M&A in China's foundry history. Both the SMIC M&A and the H200 stall are two sides of the same policy: Beijing is building a vertically integrated domestic AI chip stack, not waiting for U.S. licensing to work. Domestic Chinese AI accelerator shipments hit 41% market share in 2025, with Huawei alone at 20%, and SMIC is the exclusive foundry for Huawei's Ascend chips.

Security researchers use Anthropic’s Mythos to exploit Apple’s M5 chip in five days

A three-person team at security startup Calif used Anthropic’s Mythos Preview to find and exploit two vulnerabilities in Apple’s M5 chip, going from bug discovery to a working root-access exploit in five days. The exploit bypasses Apple’s Memory Integrity Enforcement (MIE), the hardware-level security feature Apple built specifically to stop this class of attack. Calif reported the vulnerabilities to Apple in person at Apple Park and will publish the full 55-page technical report after Apple ships a fix.

Why it matters: Apple's MIE is the most advanced memory protection on any consumer chip. Calif exploited it on current hardware and software, with just a small team and using an AI model that didn't exist a year ago.

Cerebras debuts on Nasdaq, doubles on first day

Cerebras began trading on Nasdaq on May 14 under ticker CBRS, opening at $350 and closing at $311.07 — a 68% gain over its $185 IPO price — giving the company a market cap of roughly $95 billion. The offering raised $5.55B, the largest U.S. tech IPO since Uber in 2019. The company currently carries a $24.6B revenue backlog, anchored by a $20B+ multi-year OpenAI deal and an AWS deployment signed in March.

Why it matters: Cerebras proved the IPO window is open and hungry for AI infrastructure, and the next test is bigger. SpaceX is pricing as early as June 11 at a $1.75T valuation and ~$75B raise, which would be the largest IPO in history.

OpenAI and Anthropic plug into personal financials and small business tools

OpenAI launched a personal finance feature in ChatGPT Pro this week, letting U.S. users connect over 12,000 financial institutions through Plaid, including Schwab, Fidelity, Chase, and Robinhood, to get budgeting advice and spending analysis grounded in their actual account data. More than 200 million users already ask financial questions through ChatGPT each month. Anthropic also launched Claude for Small Business, a package of agentic workflows that plug into QuickBooks, PayPal, HubSpot, Canva, DocuSign, and Google Workspace to handle payroll, invoicing, month-end close, and marketing campaigns.

Why it matters: Both moves point the same direction: the AI labs are no longer just building chat interfaces, but are wiring themselves into the transaction layer where money moves. This is a direct threat to incumbent fintech, bookkeeping, and advisory software. I’d watch for the same pattern in health and healthcare, taxes, work benefits, and more.

Isomorphic Labs raises $2.1B Series B to scale AlphaFold into drug design

Alphabet’s Isomorphic Labs raised a $2.1B Series B led by Thrive Capital, with participation from Abu Dhabi’s MGX, Singapore’s Temasek, and the UK Sovereign AI Fund, bringing total capital to ~$2.6B. Isomorphic has not dosed a human patient yet but holds multi-billion-dollar R&D partnerships with Eli Lilly, Novartis, and J&J. The company’s AI drug design engine, IsoDDE, is built on the same foundations as DeepMind’s Nobel Prize-winning AlphaFold.

Why it matters: AlphaFold proved AI could predict protein structures. Isomorphic is the bet that the same approach can design drugs end to end, from target to clinical candidate. Three pharma giants are already customers for that capability.

There’s a Chinese Grey Market for GPT & Claude tokens

A Chinese CS student posted a detailed breakdown of the proxy station economy on Reddit this week. On Xianyu and Taobao (China’s eBay and Amazon), vendors openly sell GPT-5.4 and GPT-5.5 API access at 70-90% cheaper than OpenAI’s official pricing (roughly $1 for 100M tokens). The arbitrage works by converting Codex subscriptions from low-price regions into standard API endpoints using open-source tools, then pooling accounts and reselling at scale. Adoption among Chinese developers and CS students is described as near-universal.

Why it matters: Some proxy operators reportedly sell user interaction data to domestic AI labs for distillation and post-training. The post also notes that when grey-market GPT costs roughly the same as DeepSeek, most Chinese developers default to GPT, which means U.S. frontier models may have far more usage in China than official numbers suggest, running through infrastructure that neither OpenAI nor Anthropic controls.

xAI open-sources the X recommendation engine

xAI pushed a major update to the open-source X “For You” algorithm, replacing the prior heuristic-based system with Phoenix, a ranking engine built on xAI’s Grok transformer architecture. The release includes a runnable end-to-end inference pipeline, a downloadable 3GB pre-trained mini model, and for the first time, the ads blending system.

Why it matters: This is the first global-scale recommendation system released with a working model checkpoint. Researchers and startups building social or content products can now benchmark against an industrial recommender instead of simplified proxies.

DEEP DIVE FROM THE REVIEW

Submissions to one of the top physics journals more than doubled last year.

The editor’s take: most of it is people who think they’re doing science by chatting with an LLM.

This is the part of AI x science maybe nobody really wants to talk about.

AI is making it dramatically easier to produce work that looks like science. Whether it’s actually making science better is a totally different question, and the benchmarks we use can’t tell us either way.

Strange Research Fellow Mason Rodriguez Rand on what we’re getting wrong about how we measure all this, plus three suggestions for what better measurement could look like.

EVENTS

We are hosting an AI x Science Jeffersonian dinner in San Francisco! We are keeping 1 or 2 seats for people we haven’t met. If you’re interested in attending, please request an invite here.

The result is striking, but the framing around it has been misleading. The AI compressed the hypothesis-generation step. But it did not compress the experiments.

“The system gives you an answer and that needs to be experimentally validated,” the Imperial researchers noted. Ten years of bench work did not collapse into 48 hours. Yes, one cognitive task did. The rest, the part of science that actually involves physical interaction with the world, remained exactly as slow as it was before.

This distinction matters because it maps onto a larger problem in how the field measures itself. AI systems are now powerful enough to saturate benchmarks designed to track scientific capability. They are not demonstrably good at the work those benchmarks were trying to proxy.

We are hosting an AI x Science Jeffersonian dinner in San Francisco! If you’re interested in attending, please request an invite here.

What today’s benchmarks actually test

The current generation of AI-for-science benchmarks share a structural feature: they are built around questions with known answers.

PaperBench, released in April 2025, asks AI agents to replicate 20 ICML 2024 Spotlight and Oral papers from scratch: read the contributions, write the code, run the experiments. The best agent scored 21%; ML PhD students given 48 hours scored 41%. FutureHouse’s Falcon hits roughly 90% on LitQA2, a graduate-level biology literature retrieval benchmark, against 67% for human domain experts.

These are real measurements of real capabilities. They are also, almost without exception, measurements of convergent reasoning.

The questions were written by humans who already knew the answers. The “ground truth” exists because someone established it first. PaperBench, the most demanding of the set, measures the reproduction of work that has already been done. FrontierScience’s authors were candid about what their benchmark misses: it “does not capture many core elements of scientific work,” including hypothesis generation and experimental interaction.

Thomas Wolf, co-founder and chief science officer of Hugging Face, framed the deeper problem in a 2025 essay. Current benchmarks test whether AI systems can answer questions we already know the answers to, but

His argument has a technical edge: LLM training optimizes for the most likely next token under a distribution learned from existing human knowledge. The scientist who discovers something genuinely new is, by definition, producing a low-probability output under that distribution. The training objective and the target behavior are in tension at a structural level, not an incidental one.

Wolf’s proposed counter-benchmark is interesting precisely because it cannot yet be built: evaluate a model on a recent discovery it could not have been trained on, and see whether the model independently generates the right questions to reach it. This requires a corpus of problems solved after the model’s training cutoff where the solution was genuinely surprising. Building such a benchmark requires knowing in advance what genuine novelty looks like, which is the circular problem that defines the evaluation crisis.

The Imperial AMR case is the closest thing to Wolf’s test that has actually been run in the wild. The result was impressive on the hypothesis-generation axis. On the other axes that constitute scientific work, the AI was not tested.

Too many submissions?!

Anyone who has talked to a research editor or a scientific conference organizer in the last year has heard some version of this. Carlo Rovelli, the theoretical physicist and chief editor of Foundations of Physics, told TIME in 2025 that submissions to his journal had more than doubled in the past year.

A January 2026 Nature paper from researchers at Tsinghua and collaborators analyzed 41.3 million papers across six natural science disciplines from 1980 to 2025, using a fine-tuned language model to identify AI-augmented research.

Scientists who use AI publish 3.02 times more papers, receive 4.84 times more citations, and become research project leaders 1.37 years earlier than those who do not. AI adoption also shrinks the collective volume of scientific topics studied by 4.63% and decreases scientists’ engagement with one another by 22%. The authors call the result “lonely crowds”: concentrated activity on popular topics where individual papers attract more attention but where cross-citation and intellectual exchange between adjacent research programs declines.

This is the benchmark problem playing out in the actual literature. Scientists are using AI tools where AI tools are most effective, which is on tasks involving data processing and pattern recognition over existing corpora.

Most of it, he said, is people who think they are doing science by having conversations with LLMs. If AI is making it easier to produce plausible-sounding work, the rate of genuine discovery may stay flat or decline even as publication counts climb. The narrowing the Tsinghua paper measures and the degradation Rovelli is describing point in the same direction: yes, there is more output, but less of it that matters.

The Oxford Internet Institute documented a parallel problem at the benchmark level: only 16% of 445 LLM benchmarks they reviewed use rigorous scientific methods, and roughly half claim to measure abstract capabilities without defining those capabilities. Benchmarks create path dependencies. A system that saturates the available benchmarks is not the same as a system that is scientifically useful, and the closer the available benchmarks come to defining the goal, the more the development incentive pulls toward measurable proxies and away from the harder, unmeasurable capabilities.

Sara Hooker’s 2020 essay on the “hardware lottery” captured a version of this dynamic at the infrastructure level: research directions win when they happen to fit the tools that exist, not necessarily when they are universally superior to alternatives. Deep neural networks won partly because GPUs built for video games turned out to be good at matrix multiplies. The benchmark question is the same dynamic moved up one level. Fields with rich data and cheap iteration are getting the lift from current AI tools. Fields with sparse data or expensive experiments, often the ones where the highest-stakes problems for human welfare live, are not.

What better measurement would look like

I believe there are at least three proxies for real scientific impact worth tracking.

The first is topic diameter, the metric from the Tsinghua paper. Not how many papers a system generates or how many citations it accumulates, but whether the set of topics being actively researched is expanding or contracting as AI adoption rises. This is the most direct available measure of whether AI is opening new frontiers or accelerating work in established ones.

The second is the hypothesis-to-validation ratio: not just the rate at which AI systems generate hypotheses, but the fraction that survive experimental testing, and specifically the fraction that are genuinely novel rather than confirmations of prior work. This requires longitudinal data across research programs and is expensive to collect. It is also the closest available proxy for what actually matters.

The third is cross-disciplinary citation flow: whether AI-assisted research is increasing or decreasing the rate at which papers cite work from outside their primary domain. If the Nature finding on engagement decline is robust, it implies a specific failure mode worth tracking directly. AI tools may be efficient at retrieving the most relevant literature in a narrow domain and poor at surfacing the distant, unexpected connection that historically produces paradigm shifts.

OpenAI’s own researchers acknowledged the underlying problem when they wrote that “the most important benchmark for the scientific capabilities of AI is the novel discoveries it helps generate.”

This is correct and, for now, aspirational rather than operational. Novelty cannot be established at the time of generation. It can only be recognized later, as a community of peers evaluates whether something genuinely unexpected happened. The Imperial AMR result was impressive precisely because the Co-Scientist’s hypothesis could be checked against a decade of independently produced experimental work, a form of ground truth that almost no benchmark replicates.

Building AI for science evaluation frameworks that is built on post-hoc validation against independent experimental results would be slower and more expensive than what exists now. They would also tell us something real.

The empirical question worth holding the field to is not whether AI can match human performance on difficult scientific tasks. That question is being answered, affirmatively and quickly.

The question is whether the tasks AI matches humans on are the ones that produce scientific progress. The benchmark architecture currently in use cannot distinguish between the two.

Mason Rodriguez Rand is a Strange Research Fellow who holds degrees in molecular and mechanical engineering (UChicago, UC Berkeley) and has led engineering and go-to-market efforts at climate and energy startups out of Argonne National Lab and UC Berkeley, spanning nuclear, carbon removal, and advanced materials.

I called this last July. Maybe not exactly this, but close enough that I went back and reread the post when I read that when both Anthropic and OpenAI announced their PE joint ventures on Monday.

The pattern I was watching ten months ago was that some of the biggest money to be made in AI wasn’t in product-led growth, but looked more like technical consulting, bundling custom AI agents with forward-deployed teams.

This week, both Anthropic and OpenAI went all in.

Anthropic with Blackstone, Hellman & Friedman, and Goldman Sachs, with $1.5 billion committed, plus Apollo, General Atlantic, GIC, Leonard Green, and Sequoia in the consortium. OpenAI with TPG, Brookfield, Bain, Advent, Goanna, SoftBank, and 13 others, $4 billion raised at a $10 billion valuation.

The deeper thing I’ve been turning over is: why PE? Why not Accenture, Deloitte, the McKinseys? They have the bench. They have the relationships. They’ve been deploying enterprise software and change management for thirty years.

The answer, I think, is that the consulting framing is misleading. This isn’t a consulting business.

PE firms own companies.

Blackstone has somewhere north of 250 portfolio companies. TPG has over 280. And crucially, these are mostly buyouts, meaning the PE firm has control stakes, board seats, and an operating team that already runs cross-portfolio initiatives like group purchasing and shared healthcare.

With this partnership, the JV doesn’t have to win over customers. The customers are already in the family, with operational leverage to mandate adoption. The forward-deployed team dives into a portfolio company on day one with directives from the asset manager who is a majority owner. It’s a captive distribution channel.

And the opportunity, to be clear, they’re targeting is not the $1.4T software market. It’s the ~$50 trillion global labor market.

Tara

THE DOWNLOAD

OpenAI releases MRC, an open networking protocol for AI supercomputer training clusters

OpenAI released MRC (Multipath Reliable Connection) on May 6 through the Open Compute Project, co-developed with AMD, Broadcom, Intel, Microsoft, and NVIDIA. The protocol sprays a single data transfer across hundreds of network paths and reroutes around failures in microseconds, enabling 130,000-GPU clusters with two switch tiers instead of three or four. It is already in production at OpenAI’s Abilene Stargate site and Microsoft’s Fairwater supercomputers, where OpenAI rebooted four tier-1 switches mid-training on a recent ChatGPT and Codex run without coordinating with the team running the job.

Why it matters: Ethernet already won the AI back-end — it accounted for more than two-thirds of switch sales in 2025 and tripled year over year, up from under 20% two years ago. The fight has moved one layer up, to transport selection, with four incompatible options now running on the same physical hardware: RoCEv2 (cheapest, most fragile under failure), NVIDIA Spectrum-X with Adaptive RDMA (resilient but vendor-locked), Ultra Ethernet (UEC 1.0 shipped in June), and now MRC. Hyperscalers are hedging across all of them — Meta runs Spectrum-X and is a UEC founder; Microsoft runs MRC and is also a UEC founder. Neoclouds and mid-scale operators cannot hedge that way. As Tomahawk 6 hits volume and Spectrum-X Photonics ships in H2, the question for any operator at the four-thousand to sixteen-thousand GPU scale is no longer bandwidth on the spec sheet — it is what happens to the training run the next time a tier-1 switch reboots, and whether their team can actually run UEC or MRC in production.

DeepSeek raises up to $7.3B at $50B valuation, with founder Liang Wenfeng writing a $2.9B check

DeepSeek is in talks to raise up to 50 billion yuan (~$7.35B) in its first external round, the largest by a Chinese AI company on record, at a $50B+ valuation. Founder Liang Wenfeng is personally contributing roughly $2.9B, making about 40% of the round. He currently controls roughly 84% of the company pre-round. The state-backed China Integrated Circuit Industry Investment Fund (”Big Fund”) is in talks to lead, with Tencent, Alibaba, and Hillhouse reportedly in discussions. The capital is earmarked for compute infrastructure and a pivot toward enterprise commercial products.

Why it matters: The Big Fund leading is a deliberate broadening of its mandate from semiconductors (SMIC, YMTC) to frontier model labs, and it is the most explicit signal yet that Beijing is treating DeepSeek as strategic national infrastructure rather than a venture investment. But the more interesting structural fact is Liang himself writing the largest check at a $50B price. This is a level of founder concentration that does not exist anywhere else in frontier AI, and that ties one of China’s most successful quant fund operators directly to the country’s flagship open-weight lab. Liang staged the cap-table reorganization deliberately ahead of the round to consolidate control before institutional money entered, which is what gives him the structural ability to write a $2.9B check while keeping veto power.

Anthropic leases entire capacity of SpaceX's Colossus 1 data center

Anthropic signed a deal to use the entire compute capacity of Colossus 1, the Memphis data center owned by SpaceX following its January 2026 absorption of xAI. The deal gives Anthropic 300+ MW and 220,000+ NVIDIA GPUs within the month.

Why it matters: xAI built Colossus 1 to train Grok, and Grok’s user base never grew into the capacity. xAI is reportedly utilizing only 11% of its 550,000 GPU fleet. With SpaceX targeting a June S-1 filing for what is expected to be the largest IPO in corporate history, leasing idle capacity to Anthropic converts a multibillion-dollar write-down risk into a high-margin recurring revenue line ahead of the prospectus. Anthropic, meanwhile, saw 80x year-over-year revenue and usage growth in Q1 2026 against a 10x plan, and CEO Dario Amodei has publicly cited compute deficit as the reason for recent rate-limit complaints.

A Claude Code engineer argues HTML is replacing markdown as the default agent output format

Thariq Shihipar, an engineer on the Claude Code team, published a breakout piece this week arguing that markdown has become a restricting format for agent outputs and that HTML (with more information-dense features like embedded SVG, interactive sliders, live previews, and “copy-as-prompt” buttons) should be the default.

Why it matters: As models produce more artifacts per task, the challenge has actually moved from model capability to human comprehension (i.e. the chance that anyone actually reads, edits, and acts on what the agent produced). The most useful pattern in Thariq's post is the export step. Every interactive HTML artifact ends with a "copy as JSON" or "copy as prompt" button, turning the artifact into a typed handoff back into the next loop. That makes the artifact behave less like a document and more like an API: a structured intermediate state between agent runs.

DEEP DIVE FROM THE REVIEW

Plus: Anthropic ran an experiment where AI agents negotiated real deals for real people. Half got a weaker model and netted worse deals, but rated them just as fair. What does that mean for the $400B ad market built on human intent?

digs in.

Both thought they got a fair deal.

Last week Anthropic published Project Deal, and the headline everybody took away was that AI agents can handle real negotiation. That’s true. But the more important finding is the one that got less attention: when two agents negotiate on behalf of two humans, the human that got the losing end of the deal couldn’t tell.

Anthropic’s market experiment

Project Deal was set up like this. Sixty-nine employees in Anthropic’s San Francisco office listed items they wanted to sell: a snowboard, lab-grown rubies, a bag of nineteen ping-pong balls. Claude agents, briefed on each person’s preferences in a ten-minute interview, handled all the bargaining autonomously over Slack. No human approved a bid. No one signed off on a counteroffer. After a week, people exchanged real goods based entirely on deals their agents had struck.

About 186 deals was completed, totaling about $4,000 in total transaction value.

But the study also concealed a second experiment. Half the participants were quietly assigned Claude Haiku 4.5, a weaker model, while their counterparties got the frontier Opus 4.5. Across 161 items sold in multiple runs, an Opus seller pulled in $2.68 more on average. An Opus buyer paid $2.45 less.

The most counterintuitive finding in the paper is that what people instructed their agents on how to negotiate barely mattered.

Participants who gave aggressive instructions, who told their agents to push hard, to hold firm, to maximize price, saw no statistically significant improvement in sale likelihood once opening prices were controlled for. Model capability dominated prompting. The person who spent ten minutes carefully coaching their agent on tactics was, in measurable terms, customizing tone, but did not move the needle on outcomes.

The $400B+ market being disrupted.

In the internet era digital commerce was built on one assumption: a human is available to be influenced at the moment they consider a purchase. Come banner ads, retargeting pixels, search rankings, sponsored listings. Every tool in the performance marketing stack exists to intercept a person at that moment and nudge them toward a decision.

Google generates $240 billion a year in advertising revenue on that basis. Meta is in a similar realm.

Together they account for more than half of global digital ad spend, and nearly all of it is based on access to a human at the point of purchase intent.

But what happens when that point of intent is evolving? Adobe Analytics measured a 4,700% year-over-year increase in agent-driven traffic to US retail sites in a single month.

The human isn’t disappearing from commerce. But the moment where they’re reachable, the moment where they’re weighing options and can be nudged, is migrating inside the agent.

Expedia’s 2026 10-K is, as far as I can find, the first filing where a public company has named AI agents as an existential competitor category. This year’s filing notes that the company “may lack a significant presence” in AI-driven platforms.

Expedia is responding the way it can: plugging into Alexa+, Gemini, ChatGPT, and Google’s UCP travel coalition with Booking, Marriott, IHG and Wyndham. Two-thirds of Expedia’s bookings still come direct, and direct is growing faster than indirect.

The New Commerce Stack

Project Deal is the first field evidence that both sides of a market can be fully mediated by AI agents at the same time, and that the mediation is invisible to participants.

The agentic commerce infrastructure is being built right now. Stripe announced last week a partnership with Google enabling purchases inside Gemini and AI Mode via UCP, and opened its Link consumer wallet to AI agents so users can let agents make payments on their behalf, and launched something called the Machine Payments Protocol. Google’s UCP launched with Shopify, Walmart, Target, and twenty other partners. Visa launched a Trusted Agent Protocol. Experian announced a “Know Your Agent” framework three days ago.

The rails for agents to transact are going up fast.

What is not being built is the consumer-facing layer. The one that makes agent decisions legible to the people those agents represent.

What did your agent consider? What was offered and declined? Where did it concede? Was it steered, and by what? Could it have done better? No current system answers these questions. Most users do not yet know to ask them.

To date, every consumer protection policy we have assumes a human is present at the moment of decision. Price transparency works because a person can check prices. Comparison shopping works because a person can walk across the street. Cooling-off periods work because a person can reconsider.

These mechanisms depend on the consumer being able to perform some behavior, reading terms, comparing offers, walking away, that gives them a window into the market they’re operating in.

Agent commerce removes that window. The user experiences an outcome, not a process. They see what was bought, not what was considered and rejected. They see the price paid, not the price that was available.

The technical mechanism to make agent reasoning visible exists. Models can externalize their chain of thought, the sequence of considerations, tradeoffs, and concessions that led to a decision. No commercial agent surfaces this to users after a transaction. You don’t get “I considered four vendors, rejected two on price, and conceded on shipping speed to close.” You get the receipt.

This is one mechanism that could give consumers a window into what their agent actually did on their behalf.

Joy Yang is a Strange Research Fellow. She is pursuing computer science and government at Oxford, and is a researcher with its Visual Geometry Group. She was previously an intern with OpenAI and Google.

Events

Interested in AI and design? Join us for a private demo of MagicPath 2.0 with founder Pietro Schirano this Thursday in San Francisco.

Strange Magic Hour: Design

RSVP

Every major hyperscaler reported earnings this week. All three said the same thing: we could have sold more compute if we had it.

Google Cloud posted $20 billion in quarterly revenue, up 63% year-over-year. Its backlog nearly doubled to $462 billion. Pichai told analysts he’s “compute constrained” and revenue would have been higher if they could meet demand. Microsoft said the same thing.

I keep coming back to two things.

One: we’re likely seeing the beginnings of an inflection point for inference demand.

Two: if every GPU in the West is spoken for by paying customers, there’s probably little left for open source.

That second point… I can’t stop thinking about.

Demis Hassabis sat down with YC this week and said it plainly: the West is losing to China on open source AI.

Google doesn’t have enough compute to build two frontier models, one open and one closed. That’s why Gemma stays small. Meta is the only Western lab shipping frontier-class open weights, and even their open releases lag what they use internally.

This matters because open source is the foundation layer for every major technology platform. Roughly a whopping 70 to 90% of the code in modern web and cloud applications is open source. Open source is really the thing everything else gets built on. Cede that layer and you cede influence over how most of the world deploys AI.

Every startup, every government, every developer who can’t afford frontier API pricing on every tool call builds on open weights. Right now, that increasingly means DeepSeek, Qwen, Minimax. The Chinese open ecosystem.

The West is winning the frontier but losing the foundation.

Meanwhile, China is building a parallel compute stack entirely. Nvidia B300 servers are going for over $1 million in China because export controls have tightened again. But the pressure is accelerating domestic alternatives, not blocking them. ByteDance and Alibaba are shifting orders to Huawei’s Ascend 950. DeepSeek reportedly trained at least partially on Huawei silicon.

Google could change the open source game in the US. They have the research talent, the TPU stack, and the distribution. But when your closed models have a $462 billion backlog, it’s likely very hard to justify giving away compute for “free” or the greater good.

China doesn’t have this problem yet. Their frontier labs aren’t capacity-constrained at the same scale, and their government treats open AI as strategic infrastructure, not a business decision.

I think the compute wall is the most important structural force in AI right now. Not model architecture, not regulation, not talent. The physical scarcity of leading-edge silicon is determining what gets built, who gets access, and which ecosystem the rest of the world builds on…

Enjoy the brief.

Tara

THE DOWNLOAD

Cerebras Targets $40B IPO on the Back of a Single $10B+ Contract

Cerebras is seeking to raise as much as $4B in its IPO at a valuation of roughly $40B, nearly 5x its $8.1B private valuation from September 2025. This is largely due to multi-year compute agreement with OpenAI worth more than $10B, with an option for an additional 1.25 gigawatts through 2030. The company reported $510M in 2025 revenue, up 76% YoY, but customer concentration is still extreme.

Why it matters: This is the closest proxy for non-NVIDIA AI silicon at datacenter scale. The OpenAI agreement gives Cerebras a credible foothold in inference, where margin pressure is mounting fastest, and where its wafer-scale processors are designed to compete. But the deal structure reveals how concentrated the “NVIDIA alternative” market really is: one contract accounts for nearly all of the valuation step-up.

Meta Acquires Assured Robot Intelligence to Seed Humanoid AI Team

Meta acquired Assured Robot Intelligence (ARI), a startup building foundation models for humanoid robots, for an undisclosed sum. The team, including co-founders Lerrel Pinto and Xiaolong Wang, will join Meta Superintelligence Labs. Wang said the startup’s work made clear that achieving physical AGI requires a universal physical agent, that the agent will be humanoid, and that “scaling will come from learning directly from human experience, not teleoperation alone.” Meta is building its own hardware, sensors, and software for humanoid robots and plans to license the tech to other companies.

Why it matters: This is a talent acquisition that signals strategic intent. Pinto previously co-founded Fauna Robotics, which Amazon acquired, and Wang is an associate professor at UC San Diego and former NVIDIA researcher. Big Tech is locking up the small pool of researchers who can bridge foundation models and whole-body robot control before they incorporate as startups.

Samsung Chip Profit Jumps 48x on AI Memory Demand

Samsung’s semiconductor division reported operating profit up 48 times year-over-year in Q1, driven by surging demand for high-bandwidth memory used in AI systems. Memory prices have risen roughly 6x in the past year as DRAM fabs run above 90% utilization. Samsung has been validated as an HBM4 supplier for NVIDIA’s Vera Rubin, alongside SK Hynix, with Micron excluded from the flagship platform.

Why it matters: Two Korean companies now control who gets the memory required to build frontier AI systems. SK Hynix’s CFO has said the company has “already sold out our entire 2026 HBM supply,” and Micron confirmed similar constraints, with new capacity not meaningfully available until 2027.

Three AI Neo-Labs Raise $1.65B+ Betting on Post-LLM Intelligence

Three new labs raised over $1.65 billion this week, all built on the thesis that LLMs have a ceiling. Ineffable Intelligence (London), founded by David Silver (ex-DeepMind, AlphaGo), raised $1.1B at $5.1B valuation to build RL-native “superlearners” that generate their own training data without human examples. Natural Will (Beijing), founded by Tsinghua professor Ding Ning, raised $550M for embodied AI brains for robotics. Human Intelligence (Stanford), founded by James Zou, raised $100M at $1B to build AI scientist agents, building on his lab’s Nature-published work where LLM agents designed 92 plausible nanobody binders against Covid variants.

Why it matters: This is the third “mega seed neo,ab round” in two months, following AMI Labs (LeCun, $1.03B) and Recursive Superintelligence (Rocktäschel, $500M). The bets are on that reinforcement learning, embodied AI, and agent-based science will break through where scaling language models alone cannot.

New Framework RecursiveMAS Lets AI Agents Collaborate Through Internal States

When AI agents work together today, they talk to each other in text. One agent writes out its reasoning, the next agent reads it, and so on. A UIUC/Stanford/NVIDIA/MIT team built a framework called RecursiveMAS that skips the text entirely. Instead, agents pass raw internal representations to each other, the way neurons pass signals rather than sentences. The result across 9 benchmarks: 8% better accuracy, up to 2.4x faster inference, and 34 to 75% fewer tokens consumed.

Why it matters: Most multi-agent tools today (CrewAI, AutoGen, LangGraph) pay for every word agents say to each other. If agents can collaborate without generating text, the economics of running multi-agent systems change fundamentally. This is early research, but it points toward a future where the orchestration layer disappears into the model itself, and the cost of agent coordination drops close to zero.

DEEP DIVE FROM THE REVIEW

Inference is overtaking training in volume and dollars this year.

But data centers built for the training era weren’t built for what’s coming. The inference boom may leave a generation of them behind.

Strange Research Fellow Rahul Narula on what gets stranded, and what’s already there to take its place.

EVENTS

Interested in AI and design? Join us for a private demo of MagicPath with founder Pietro Schirano next Thursday in San Francisco.

Strange Magic Hour: Design

RSVP