Lilian Weng, the former OpenAI VP of AI Safety and cofounder of Thinking Machines Lab, has resigned effective July 29, 2026, citing seven months of health problems exacerbated by startup-level pressure. Her departure marks the fourth of six original cofounders to leave the $12 billion AI startup since its February 2025 launch, raising questions about workplace culture and talent retention at one of the most richly funded AI companies ever created.

Weng's resignation letter stood out for its candor in an industry where founder exits typically come wrapped in euphemisms about "new chapters" and "exciting opportunities." Instead, she was direct: she had been sick more in the past seven months than at any other point in her life, and the relentless pace of a startup had made recovery impossible. She described the experience of resting while fighting guilt about not contributing to the team as deeply unpleasant. Weng noted that she considered scaling back her responsibilities, but ultimately realized that any role where she could not give 100% effort would feel like failure.

Mira Murati, Thinking Machines' founder and former OpenAI CTO, responded with public grace, saying the experience of cofounding the company had been "precious" and that she was "glad Weng was putting her health first". That measured response contrasts sharply with the broader pattern of departures at the company, which has involved disputes over compensation, compute constraints, and unclear product direction.

Why Is Cofounder Attrition at Thinking Machines So High?

The numbers tell a striking story. Murati launched Thinking Machines in February 2025 with six cofounders and a vision for enterprise AI that adapts to individual businesses rather than imposing one general-purpose model on everyone. Within months, the attrition began.

• October 2025: Andrew Tulloch departed for Meta, one of the world's largest AI research labs.
• January 2026: Barret Zoph and Luke Metz both left to return to OpenAI, with Zoph's exit reportedly turning acrimonious after he was fired for sharing confidential company information with competitors.
• July 2026: Lilian Weng resigned, citing health and burnout.

Only two of the original six cofounders remain at the company. The departures have involved more than just better offers elsewhere. According to reporting by Fortune and TechCrunch, the exits have been driven by money constraints, limited access to computing power, and a lack of clarity on products and business model.

The broader context amplifies the significance of Weng's departure. Thinking Machines raised $2 billion at a $12 billion valuation in a seed round, making it one of the most richly funded AI startups before it had shipped anything publicly. That valuation created enormous expectations, and expectations at that altitude tend to produce exactly the kind of relentless pace Weng describes. The competition for elite AI researchers is fierce: Meta, OpenAI, Google DeepMind, and others are all writing large checks to pull talent back or poach it.

What Has Thinking Machines Actually Shipped?

Despite the revolving door of cofounders, Thinking Machines is not standing still. On July 15, 2026, just two weeks before Weng's departure, the company shipped its first open-source model, called Inkling. The timing is notable: shipping a flagship product right as a cofounder exits for health reasons is an unusual kind of milestone.

Inkling is a 975 billion total parameter mixture-of-experts model, with 41 billion active parameters per query, trained on 45 trillion tokens spanning text, image, audio, and video. It carries an Apache 2.0 license, meaning companies can download, fine-tune, and commercialize it freely. The strategic pitch pairs Inkling with Thinking Machines' customization platform, called Tinker, and takes a revenue-sharing cut on enterprise deployments. It is a credible product play, and on independent testing, Inkling is the strongest open-weights release from a Western lab, though it still trails the Chinese open-source flagships on raw capability.

Weng herself acknowledged the milestone in her resignation letter, calling Inkling "an incredible milestone in the company's history". The irony is that she will not be part of the next chapter of the company she helped found.

How to Evaluate Startup Culture and Burnout Risk

Weng's departure offers a case study in how even well-funded, prestigious startups can create unsustainable working conditions. Her experience reveals several warning signs that apply beyond Thinking Machines:

• Health as a Leading Indicator: When a cofounder with a track record of sustained intellectual work openly admits to being sick more frequently than ever before, it signals systemic pressure rather than individual weakness. Weng spent nearly seven years at OpenAI leading the safety team and became one of the most cited researchers in the field; she does not leave positions lightly.
• Guilt as a Structural Problem: Weng described bearing guilt about taking sick leave while the team sprinted toward the Inkling release. This suggests a culture where stepping back, even for legitimate health reasons, is framed as letting others down rather than as necessary self-care.
• Valuation Mismatch with Product Maturity: A $12 billion valuation before shipping a product creates pressure that is difficult to manage. The company attempted to raise a second round at a $50 to 60 billion valuation in November 2025, but those negotiations failed by January 2026, returning the valuation to $12 billion. That whiplash likely intensified internal pressure.

Weng's letter included a phrase that resonated across the AI community: she hopes to find "an environment with a more predictable pace and clearer boundaries of responsibilities." That is not a critique of Thinking Machines alone; it is a statement about the current state of AI competition, where progress is measured in days and the pressure to ship faster never stops.

The broader question Murati now faces is not only how to replace a fourth cofounder. It is whether the operating culture that produced four exits in seventeen months can attract and hold the kind of talent Thinking Machines' ambitions require. The product is compelling. The cofounder attrition rate is not.

Weng's departure also highlights a tension in the AI industry: the best researchers and engineers are in high demand, and they have options. When a company's culture or pace becomes unsustainable, they leave. Thinking Machines has considerable resources, including a multi-year strategic cooperation with Nvidia and a multi-billion-dollar computing power agreement with Google Cloud. But resources alone cannot solve the problem of burnout.