![](\"https:\/\/fortune.com\/img-assets\/wp-content\/uploads\/2026\/07\/GettyImages-2032109653-e1785437635643.jpg?w=2048\")
<\/p>\n<\/p>\n
Andrew Ho left OpenAI after eight months on Wednesday, as the former researcher announced on X that he was starting a company to sell high-end reinforcement learning datasets to frontier AI labs.<\/p>\n
Around 5 a.m. the next morning, still awake, Ho posted some advice for the colleagues he\u2019d left behind: take the money while you can. The frontier labs, he wrote, are overvalued.<\/p>\n
\u201cI would strongly recommend taking liquidity if you\u2019re eligible for tender offers,\u201d Ho wrote. \u201cIt seems somewhat implausible that the valuation is going to, like, 2x after the IPO, but it does certainly seem plausible that it could go down by 50%.\u201d<\/p>\n
By the time he woke up, the posts had drawn hundreds of thousands of anxious eyes, having arrived in the middle of a tech selloff that had pushed the Nasdaq 100 into correction the day before. \u201cIt was kind of remarkable,\u201d Ho said in an interview with Fortune<\/em> Thursday.<\/p>\n He said he has about $700K in shares that he\u2019s not legally allowed to sell until after the IPO and lockup period. \u201cSo I\u2019m just stuck,\u201d he said.\u00a0<\/p>\n So even if he\u2019s talking against his own book, Ho is publicly standing on business. Like most in the industry, he believes that demand for inference will rise dramatically;\u00a0 that there will be a compute crunch soon; that the data center buildout is necessary. But what he\u2019s \u201cparanoid\u201d about is that competition from cheap models is forcing frontier labs onto a faster and faster treadmill of spending, and that revenue will never quite catch up.\u00a0<\/p>\n What Ho describes is what analysts, with a little flourish, like to call a \u201cRed Queen\u2019s race,\u201d where you have to run as fast as you can just to stay in place. Frontier labs are having to pay more and more for each cycle of training, and the advantage it buys is temporary; cheaper competitors like Moonshot\u2019s Kimi can close the gaps for a fraction of the cost through distillation.\u00a0<\/p>\n Revenue, he added, will almost certainly go up with each model; but the question is, how much? With all the debt frontier labs are taking on, he said, \u201cIf you miscalculate even by just a very fine amount, that can be the difference between life or death for a company.\u201d\u00a0<\/p>\n Most investors and peers at his lab feel okay about this, he added, because they think they\u2019re on the cusp of a breakthrough: RSI, or recursive self-improvement. That\u2019s the idea that an AI system can get good enough at AI research to improve itself, running its own experiments and training its own successors. If each version can build itself a slightly better version, the curve for intelligence could go parabolic, as well as the cost of compute.\u00a0<\/p>\n But Ho doesn\u2019t buy it. Research, he said, isn\u2019t limited by models\u2019 intelligence; rather, it\u2019s limited by a lack of \u201cresearch taste.\u201d The models aren\u2019t good at proposing experiments, or at recognizing which results are important and which aren\u2019t. \u201cYou can\u2019t reason your way to phenomena,\u201d he said.\u00a0<\/p>\n Models, Ho added, are extraordinarily good at some things and just bad at adjacent ones, and though the pattern seems random, it\u2019s not. Capabilities have advanced fastest where results are equal to check; a mathematical proof is valid, or it isn\u2019t. A program can compile data, or it can\u2019t. That\u2019s where the billions of dollars of spending have gone, and it\u2019s worked out so far. But everything else that\u2019s harder to verify is still stuck, he said.\u00a0<\/p>\n And that\u2019s why he left. The low-hanging, verifiable fruit has been picked through, so the next round of gains, as Ho sees it, will come from messier tasks, which is the exact business he\u2019s starting: creating datasets for the judgement-heavy work models still can\u2019t do. His first products will focus on long-horizon scientific reasoning and statistical analysis.<\/p>\n This view puts Ho at odds with the case circulating Silicon Valley social media this week, most prominently from podcaster Dwarkesh Patel, who argued that compute could get 10x more expensive\u2014and thus the labs could make 10x more the revenue\u2014because of RSI. But it puts him more squarely aligned with Wall Street, which sold off 10% of Meta on Wednesday and 8% of Google last week out of fears that the companies funding the AI buildout won\u2019t get their money back.<\/p>\n Ho, now speaking more in investor-brain (he has a little background in finance), says the \u201cpeople making out really happily here are Nvidia and Micron,\u201d since they\u2019re the ones selling the chips to everyone in the race.<\/p>\n Labs themselves have two ways out of the Red Queens\u2019s race, Ho added. But both are hard: Labs could build their own chips and break Nvidia\u2019s pricing power. Or they could move up into the application layer and start building and selling more products themselves, like Meta buying Cursor. Right now, Ho argues, the labs capture a fraction of the value their models create in the economy, and that gap is \u201cvery undercapitalized at the moment.\u201d<\/p>\n For now, Ho is focusing on fielding the heavy interest in his day-old company, which doesn\u2019t have a name yet, and trying not to think about his locked-up OpenAI equity.<\/p>\n \u201cI see this vested equity, I\u2019m thinking about it, I\u2019m just sort of stressing about what\u2019s going to happen to it,\u201d he said. \u201cIt\u2019s this huge proportion of my net worth.\u201d<\/p>\n Most of his former colleagues, he suspects, don\u2019t dwell on it.<\/p>\n \u201cYou\u2019re going to work,\u201d Ho said. \u201cDon\u2019t think too hard about these questions. Which is maybe for the best.\u201d<\/p>\n<\/div>\n #stuck #Meet #OpenAI #researcher #sitting #700K #equity #overvalued<\/p>\n","protected":false},"excerpt":{"rendered":" Andrew Ho left OpenAI after ei… <\/p>\n","protected":false},"author":1,"featured_media":1230,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[2360,428,373,2361,636,792,1952,480,2362,195,2359,143,2358],"class_list":["post-1229","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-finance-news","tag-700k","tag-anthropic","tag-debt","tag-equity","tag-google","tag-meet","tag-nasdaq","tag-openai","tag-overvalued","tag-research","tag-researcher","tag-sitting","tag-stuck"],"yoast_head":"\n\nHo doesn\u2019t buy the RSI dream<\/strong><\/h2>\n