# Mira Murati’s Startup Thinking Machines Lab Loses Two Co-Founders to OpenAI
In a dramatic turn for the AI startup world, **Thinking Machines Lab**, co-founded by former OpenAI CTO Mira Murati, is losing two key co-founders—**Barret Zoph** and **Luke Metz**—who are returning to OpenAI.[1][3][4] This shake-up, announced publicly this week, highlights the intense talent wars in AI and raises questions about the stability of Murati’s high-flying venture.[1]
## The Sudden Departures: What Happened?
The news broke on Wednesday when Murati posted on X (formerly Twitter) about parting ways with Barret Zoph, Thinking Machines Lab’s co-founder and **CTO**. Her statement was brief: “We have parted ways with Barret Zoph. Soumith Chintala will be the new CTO of Thinking Machines. He is a brilliant and seasoned leader who has made important contributions to the AI field for over a decade, and he’s been a major contributor to our team.”[1][3] Notably, Murati made no mention of other departures in her post, fueling speculation about the full scope of the exodus.[1]
Just 58 minutes later, Fidji Simo, OpenAI’s CEO of applications, confirmed the bigger picture. “Excited to welcome **Barret Zoph**, **Luke Metz**, and **Sam Schoenholz** back to OpenAI! This has been in the works for several weeks, and we’re thrilled to have them join the team,” Simo wrote on X.[1][3] Zoph, previously OpenAI’s VP of research and a six-year Google veteran, is rejoining alongside Metz, another co-founder, and Schoenholz, a former OpenAI staffer who had briefly moved to Thinking Machines.[1][3]
Reports suggest the split with Zoph wasn’t amicable. According to Wired via TechCrunch, it wasn’t smooth, and details from an internal memo indicate Zoph informed Murati on Monday of his considerations to leave, only to be fired by Wednesday. OpenAI reportedly doesn’t share Murati’s concerns about him.[1][4] This rapid sequence underscores the friction in Silicon Valley’s AI talent poaching.[1]
## Background on Thinking Machines Lab
Launched in **February 2025**, Thinking Machines Lab emerged from OpenAI alumni, including Murati as CEO, Zoph, Metz, John Schulman (chief scientist), Andrew Tulloch, and Lilian Weng. The team grew to 30 top researchers from Meta, Mistral, and elsewhere, positioning it as a powerhouse.[2]
The startup quickly raised a staggering **$2 billion seed round** in its first five months, led by Andreessen Horowitz with backers like Accel, Nvidia, AMD, and Jane Street, valuing it at **$12 billion**—one of Silicon Valley’s largest seed rounds ever.[1][2] As a public benefit corporation, it gives Murati outsized voting control over the board, unlike peers like OpenAI or Anthropic.[2]
Thinking Machines focuses on **smarter, efficient AI models** via post-training techniques, contrasting the race for bigger models. Its mission: build multimodal systems for collaborative human-AI work, customizable to specific needs and values. The company critiques how training knowledge is siloed in top labs, limiting public AI use.[2]
## Key Milestones and Products
In **October 2025**, Thinking Machines debuted **Tinker**, its first product—a tool streamlining fine-tuning of open-source models like Meta’s Llama, Alibaba’s Qwen, OpenAI’s gpt-oss, DeepSeek V3.1, and Moonshot AI’s Kimi K2. Tinker leverages **LoRA** (Low-Rank Adaptation) for efficient customization without massive GPU needs or distributed computing complexities, rivaling full fine-tuning on small-to-medium datasets.[2]
Research from the lab showed LoRA’s prowess, and Tinker targets developers fine-tuning for tasks or industries. Looking ahead, as of November 2025, the company sought **$5 billion more** at a **$50 billion valuation**. Chief scientist John Schulman hinted at releasing proprietary models in 2026, adding multimodal features to Tinker, and tools for non-experts.[2]
## The Bigger Picture: Talent Volatility in AI
This isn’t Thinking Machines’ first loss. Co-founder **Andrew Tulloch** left for Meta in October 2025 (noted as November in some reports).[1][3] Earlier, OpenAI saw its own drama: co-founder John Schulman departed for Anthropic in August 2024 before joining Thinking Machines.[1] OpenAI faces its own exits, like VP Jerry Tworek and Phil Chen recently.[3]
**Talent mobility** is routine among AI giants, but losing **two co-founders**—including the CTO—less than a year in is a notable setback for a startup banking on its elite team.[1] For OpenAI, it’s a coup amid competitive pressures, including a rumored Apple-Google Gemini deal for Siri.[3]
## Implications for Thinking Machines and the AI Landscape
With Soumith Chintala stepping in as CTO—a veteran with over a decade in AI—Thinking Machines aims to stabilize.[1][3] Chintala’s promotion signals continuity, but the departures could dent morale and investor confidence, especially post its massive funding.[1][2]
Murati’s venture, once shrouded in mystery, carved a niche with Tinker and efficiency-focused research. Yet, this boomerang to OpenAI spotlights the gravitational pull of established players. As AI accelerates, expect more such flips—startups lure with equity and vision, but incumbents offer scale and resources.[1][2]
For investors and watchers, it’s a reminder: in AI, people are the product. Thinking Machines’ $12 billion valuation and Tinker momentum provide runway, but retaining stars like Schulman will be key to 2026 ambitions.[2] OpenAI bolsters its research bench, potentially accelerating frontiers.
This episode encapsulates AI’s high-stakes game: brilliance moves fast, loyalties shift, and startups must prove they can keep the dream alive beyond the hype.
(Word count: 812)
Original source: TechCrunch – Mira Murati’s startup, Thinking Machines Lab, is losing two of its co-founders to OpenAI