Mira Murati launches Thinking Machines Lab

Last fall, OpenAI's former CTO Mira Murati abruptly walked out the door and largely vanished from the spotlight. Now, she is back with a new company and a familiar roster of elite talent.

Murati just officially launched Thinking Machines Lab, a brand-new AI startup designed to challenge the dominant closed-lab playbook. While product specifics remain under wraps, Murati announced that the venture will focus on helping people adapt AI to specific needs, developing foundational capabilities, and fostering an open science culture by regularly releasing research and code.

The talent raid near OpenAI headquarters

To build an AI powerhouse, you need the brains who built the last one. Murati has wasted no time assembling an elite squad.

OpenAI co-founder John Schulman has joined Thinking Machines Lab as head of research. According to reports from The Verge, Schulman was actively scouting talent and holding recruitment meetings just blocks away from OpenAI's offices. Former OpenAI leader Barrett Zoph steps in as CTO, while Jonathan Lachman, who ran OpenAI's special projects group, has also jumped ship.

All told, Murati has already poached around 10 top researchers and engineers from elite AI outfits, including OpenAI, Character.AI, and Google DeepMind.

Human collaboration over fully autonomous agents

While frontier labs race toward fully autonomous digital workers, Murati's new venture is positioning itself around a different philosophy.

A company press release emphasizes that Thinking Machines Lab aims to build systems that help humans work alongside AI, rather than replacing them with entirely autonomous software. The stated mission centers on making AI customizable, widely understood, and grounded in solid foundations.

Crucially, the lab is also promising public transparency. By pledging to regularly share technical research papers and open code, Murati seems keen to resurrect the collaborative scientific ethos that many frontier labs abandoned in their commercialization push.

Why it matters

For builders and enterprise teams, this lab could fill a massive void. As leading models turn increasingly opaque, developers are often forced to choose between rigid proprietary APIs and open-source models that lag behind the cutting edge. If Murati's team can deliver state-of-the-art foundations paired with open research and real customizability, it gives teams unprecedented control over their own AI stack.

For the broader industry, OpenAI's talent moat continues to spring leaks. Just as Anthropic was born from early OpenAI departures, Thinking Machines Lab represents another formidable offshoot led by the very people who built GPT-4 and ChatGPT. OpenAI now faces fierce competition from its own alumni on multiple fronts.

We think focusing on human collaboration and customization is a brilliant tactical pivot. Most businesses today do not need an unpredictable autonomous agent; they need reliable, adaptable intelligence integrated into existing workflows. That said, championing open science while competing at the frontier is notoriously expensive. Without massive cloud subsidies or clear early revenue engines, it remains to be seen whether a transparent research lab can outlast deep-pocketed tech giants.