๐The Next Web (TNW)โขStalecollected in 38m
Mira Murati returns with new AI venture
๐กFormer OpenAI leader Mira Murati is back with a new venture; watch for her approach to AI governance.
โก 30-Second TL;DR
What Changed
Mira Murati returns to the industry after 18 months
Why It Matters
Murati's move signals a shift in leadership talent toward independent labs focusing on safety and governance. Her track record suggests this new venture will likely influence future AI policy.
What To Do Next
Monitor Thinking Machines Lab's official channels for their first product whitepaper or API documentation.
Who should care:Founders & Product Leaders
Key Points
- โขMira Murati returns to the industry after 18 months
- โขNew venture named Thinking Machines Lab
- โขFocus on AI governance and undisclosed product development
๐ง Deep Insight
Web-grounded analysis with 19 cited sources.
๐ Enhanced Key Takeaways
- โขThinking Machines Lab was co-founded by Mira Murati alongside other prominent OpenAI alumni, including John Schulman, Barrett Zoph, and Lilian Weng, assembling a team of around 30 researchers and engineers from leading AI companies.
- โขThe company secured a record-setting $2 billion seed funding round in July 2025, achieving a $12 billion valuation with investments from Andreessen Horowitz, Nvidia, AMD, Cisco, and Jane Street.
- โขIts first product, Tinker, launched in October 2025, is an API and managed cloud platform designed to simplify the fine-tuning of open-source AI models, such as Meta's Llama and OpenAI's gpt-oss models, utilizing techniques like Low-Rank Adaptation (LoRA).
- โขThinking Machines Lab has established significant strategic partnerships, including a multi-year deal with Nvidia to deploy at least one gigawatt of their next-generation Vera Rubin chips, and a multimillion-dollar cloud deal with Google for AI infrastructure.
- โขThe company's core philosophy emphasizes building multimodal AI systems focused on human-AI collaboration, interpretability, and customization, aiming for real-time, nuanced interactions rather than traditional 'turn-based' conversations.
๐ ๏ธ Technical Deep Dive
- Product: Tinker, an API and managed cloud platform for fine-tuning large language models.
- Functionality: Streamlines the customization of AI models, making it accessible without requiring massive GPU infrastructure.
- Supported Models: Users can modify a range of open-source models, including Meta's Llama, Alibaba's Qwen, OpenAI's gpt-oss models, DeepSeek V3.1, and Moonshot AI's Kimi K2 Thinking.
- Techniques: Utilizes efficient post-training techniques such as Low-Rank Adaptation (LoRA).
- Architecture (Tinker): Allows researchers to define their own training logic in Python, while Tinker automatically manages distributed computing tasks like setting up GPUs, connecting multiple machines, handling errors, and scaling resources, thereby improving GPU utilization and shortening iteration cycles.
- Interaction Models (Research Focus): The company is developing advanced "interaction models" designed for real-time human-AI collaboration across voice, video, and text. These systems process information in 200-millisecond chunks for continuous responsiveness, employing a main interaction model for live conversation and a secondary, slower reasoning model for deeper thinking and complex tasks in parallel. These models are intended to react to visual changes, count workout reps, translate live speech, and proactively interact with users.
- Compute Infrastructure: Secured access to significant compute power through partnerships, including a commitment to use one gigawatt of Nvidia's Vera Rubin chips and Google Cloud's infrastructure built on Nvidia Blackwell chips.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
Thinking Machines Lab's focus on "interaction models" could establish a new paradigm for human-AI collaboration.
By developing AI systems that process multimodal inputs in real-time and maintain continuous, nuanced conversations, the company aims to shift the industry away from purely autonomous agents towards more integrated human-AI partnerships.
The company's emphasis on open science and customizable models will democratize advanced AI capabilities.
By publishing research, code, and offering tools like Tinker for fine-tuning open-source models, Thinking Machines Lab aims to make frontier AI more understandable and accessible to a wider community beyond top research labs.
Thinking Machines Lab will likely become a significant player in the enterprise AI market, particularly for applications requiring interpretability and auditability.
Their stated focus on building AI systems that produce reproducible results and can be customized for specific business applications, combined with their 'researcher-first' approach for Tinker, aligns well with enterprise needs for reliable and transparent AI deployments.
โณ Timeline
2023-11
Mira Murati briefly served as interim CEO of OpenAI during the board crisis.
2024-09
Mira Murati resigned as CTO of OpenAI to pursue her 'own exploration'.
2025-02
Thinking Machines Lab was founded by Mira Murati and other OpenAI alumni, established as a public benefit corporation.
2025-07
Closed a $2 billion seed funding round at a $12 billion valuation, led by Andreessen Horowitz.
2025-10
Launched its first product, Tinker, an API for fine-tuning language models.
2026-03
Announced a strategic partnership with Nvidia for compute power, including a commitment to deploy one gigawatt of Vera Rubin chips.
๐ Sources (19)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: The Next Web (TNW) โ