Recommended Free Tools
Yes—the January 14, 2026 report was accurate. Thinking Machines Lab said it had parted ways with co-founder and CTO Barret Zoph, and OpenAI soon announced that Zoph, fellow Thinking Machines co-founder Luke Metz, and former Thinking Machines employee Sam Schoenholz were returning to OpenAI. The departures were a major leadership blow, but they did not by themselves prove that Mira Murati’s heavily funded AI company had failed.
What happened on January 14, 2026?
The sequence of public announcements was unusually compressed:
As an Amazon Associate I earn from qualifying purchases.
- Thinking Machines Lab announced that it had “parted ways” with Barret Zoph. Murati also named Soumith Chintala as the company’s new chief technology officer.
- Less than an hour later, OpenAI announced the return of Zoph, Luke Metz, and Sam Schoenholz. Zoph and Metz were identified as Thinking Machines co-founders; Schoenholz was a former OpenAI employee who had been working at the startup.
The immediate facts are documented in TechCrunch’s January 14 report and WIRED’s contemporaneous account. Murati’s public statement was brief: it confirmed Zoph’s departure and Chintala’s promotion, but did not explain why Zoph left or discuss the other returning employees.
Who left Thinking Machines?
Barret Zoph
Zoph was a Thinking Machines co-founder and CTO. Before joining Murati’s startup, he spent about six years as a Google research scientist and held a senior OpenAI research or post-training leadership role, with coverage describing his former title somewhat differently. He returned to OpenAI after leaving Thinking Machines. (TechCrunch)
#1 Best Overall
Luke Metz
Metz was another Thinking Machines co-founder and previously served on OpenAI’s technical staff. WIRED reported that his earlier work included projects associated with ChatGPT and the o1 reasoning model. He returned to OpenAI with Zoph. (WIRED)
Sam Schoenholz
Schoenholz was not one of the two co-founders named in the January headline. He was a former OpenAI employee who had joined Thinking Machines and was included in the same OpenAI announcement, making the personnel move broader than a two-founder departure. (TechCrunch)
Why did Zoph leave? The public record contains competing accounts
OpenAI’s account
OpenAI executive Fidji Simo said the hires had been in progress for several weeks. According to an internal memo described by WIRED, Zoph had told Murati he was considering leaving; OpenAI did not share Thinking Machines’ concerns about Zoph; Zoph would report directly to Simo; and Metz and Schoenholz would work under Zoph, although some assignments were still unsettled. (WIRED)
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsAllegations reported by WIRED
WIRED described an allegation circulating on X that Zoph had been fired for “unethical conduct.” A source close to Thinking Machines separately alleged that he had shared confidential company information with competitors. WIRED said it could not verify those claims with Zoph. A follow-up report added allegations of serious misconduct and internal concerns about confidential information, while also describing a separate account focused on disagreements about the company’s product, technology, and future. These are reported allegations, not established facts. (WIRED follow-up)
The strategic-disagreement account
The New York Times reported that Zoph, Metz, and Schoenholz had become unhappy with Thinking Machines’ direction, that some people viewed the startup as lagging OpenAI and other rivals in product releases, and that the founders discussed possible strategic changes, including a potential Meta deal. The Times also reported that the three pushed for Zoph to have greater control over technical direction, a change Murati rejected, and that Zoph had been communicating with OpenAI CEO Sam Altman about returning before his departure. Those claims came from people familiar with the events and should be read as attributed reporting, not as independently verified findings. (The New York Times)
What OpenAI gained
OpenAI recovered a senior technical leader and two people already familiar with its research environment. Zoph’s prior leadership experience made him the most prominent returnee, while Metz’s work associated with ChatGPT and o1 and Schoenholz’s previous OpenAI experience gave the move additional depth. OpenAI publicly disclosed reporting lines, but not a complete set of new job descriptions. The episode therefore shows the value of experienced research teams in the AI talent market without proving that any particular product or model resulted from the hires.
Rank #3
Was this part of a larger exodus?
Yes, later reporting broadened the story beyond the January announcement:
- Co-founder Andrew Tulloch left for Meta in late 2025, according to WIRED.
- The New York Times reported that about nine additional employees had either joined OpenAI or received offers. That figure combined departures with offers and came from sources, rather than a public company roster.
- On August 3, 2026, Axios reported that co-founder Lilian Weng had also left and was reportedly returning to OpenAI. Axios described Weng as the fourth Thinking Machines co-founder to leave within a year. (Axios)
These reports do not establish that every departure had the same cause, or that all of Thinking Machines’ founders had left.
What Thinking Machines was building
Thinking Machines positioned itself as an AI research and product organization focused on adaptable, human-directed systems rather than only a conventional consumer application. By January, its principal public product was Tinker, a service that lets developers customize AI models with their own data. WIRED described Tinker as the company’s main product, and the company’s site remains available at thinkingmachines.ai.
The New York Times described an October release that helped developers adjust models for specific tasks. Product delays and release speed were central to some later accounts of internal frustration, but there is no public benchmark in the cited reporting that proves Thinking Machines objectively lagged every rival.
How much money and infrastructure did it have?
| Item | What was reported | Qualification |
|---|---|---|
| Seed financing | $2 billion | Announced in July 2025; reported by TechCrunch |
| Valuation | About $12 billion | Reported valuation, not a confirmed $50 billion financing |
| Possible new fundraise | More than $4 billion at a $50 billion valuation | WIRED described talks; this was not reported as completed financing |
| Nvidia relationship | At least one gigawatt of Nvidia-powered compute beginning in 2027 | Axios reported a multiyear partnership; the investment amount was undisclosed |
Reported investors in the seed round included Andreessen Horowitz, Accel, Nvidia, AMD, and Jane Street. Funding and a large compute commitment provide capacity, but neither independently demonstrates operational health.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Does the talent loss mean Thinking Machines is failing?
It is more accurate to call the episode a serious leadership and retention crisis than a proven corporate collapse.
Best Value
Why the departures were a major setback
- Two co-founders left in the same news cycle, including the CTO, who was replaced immediately.
- Tulloch had already left for Meta, and later reporting added Weng and possible additional staff losses.
- Anonymous-source accounts described disputes over product direction, technical authority, strategic options, and confidentiality.
- Founder turnover can affect product velocity, recruiting, culture, and investor confidence even when funding remains available.
Why failure was not established
- The company remained backed by a reported $2 billion seed round and roughly $12 billion valuation.
- Chintala assumed the CTO role, allowing the company to present a succession plan immediately.
- Axios reported that headcount grew from about 30 employees in 2025 to roughly 120 by March 2026, with more people joining from major AI labs than leaving for rivals, according to a source close to the company.
- The reported Nvidia partnership promised substantial future computing capacity.
The strongest defensible conclusion is that Thinking Machines suffered an unusually visible founder-retention crisis while remaining a funded, operating AI company. OpenAI gained experienced technical personnel; Thinking Machines still had capital, a product, hiring activity, and infrastructure plans. Neither side of that comparison should be mistaken for proof that the other company had won or lost the AI race.
How the story stands after the January headline
The January 14 event was the starting point of a continuing talent story, not the whole story. By August 2026, the reported departure of Lilian Weng and the earlier move by Andrew Tulloch made the retention problem broader. At the same time, the company continued developing Tinker, expanding its workforce, and pursuing large-scale compute. For readers assessing the business, the right frame is a high-value startup dealing with founder turnover and strategic uncertainty—not a confirmed shutdown or a settled verdict on its technology.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
