He refused a reported $1.5 billion package in August 2025, joined two months later for less, and held his exit until Muse shipped. Meta has promised a confidential version of that product before the end of 2026.
Reality
- Evidence64
- Adoption
- Not scored
- Hype gap+15
- Incentives50
- Confidence62
Nvidia would supply roughly half of Thinking Machines' $5bn-to-$6bn round as an investor
The Information now puts the raise at $5bn to $6bn on at least $40bn pre-money, with Nvidia writing roughly half the cheque, which makes this a compute-supply arrangement for anyone buying Tinker cluster time.
Reality
- Evidence38
- Adoption42
- Hype gap+20
- Incentives85
- Confidence45
buildOne report1 publisher Half the BIRD training examples in a 2,500-row audit carried the wrong reference query
Thinking Machines and four academics trained one model past the usual agent pipelines on text-to-SQL, but the part worth copying is the audit they ran first, which put annotation errors in 61.1% of the benchmark examples they checked.
Reality
- Evidence57
- Adoption20
- Hype gap+14
- Incentives70
- Confidence46
OpenAI's roadmap now fits inside one president's span of control
As senior leaders left, Greg Brockman took over ChatGPT, Codex, enterprise and the infrastructure build, according to The Verge. That puts the consumer-versus-enterprise tradeoff inside a single office, months before an IPO.
Reality
- Evidence34
- Adoption
- Not scored
- Hype gap+26
- Incentives58
- Confidence44
buildOne report1 publisher Thinking Machines will pay $50,000 in its own credits to see if Inkling breaks under fine-tuning
The lab named tamper resistance an open question 40 days after Inkling's weights shipped. The award is platform compute, priced by the grantor and spendable only on the approved study.
Reality
- Evidence58
- Adoption24
- Hype gap+28
- Incentives79
- Confidence52
Meta hires another ChatGPT alumnus, and the whole industry pays the retention bill
Luke Metz joins Superintelligence Labs after a three-lab loop. The $14.3bn Scale stake and billion-dollar packages set a price rivals now have to match from capital rather than revenue.
Reality
- Evidence32
- Adoption52
- Hype gap+28
- Incentives58
- Confidence36
buildOne report1 publisher Dated AI forecasts have no scoreboard, and building against one costs weeks you can count
A developer graded 18 expired AI predictions against their own words. The best mark he shows is a C+, and the one he built on took six weeks out of his own roadmap.
Reality
- Evidence30
- Adoption38
- Hype gap+22
- Incentives62
- Confidence32