Former OpenAI CTO Mira Murati has unveiled the first product from her new company, Thinking Machines Lab — a platform called Tinker, designed to let developers fine-tune advanced AI models without the need for complex infrastructure.
Tinker allows supervised learning-based fine-tuning on systems such as Meta Llama and Alibaba Qwen, making it easy to adapt existing models to specific use cases.
The platform is already in use at Princeton, Stanford, and Berkeley, where researchers employ Tinker to build specialized AI systems that solve mathematical proofs, analyze chemical data, and support complex scientific reasoning.
Thinking Machines Lab is now accepting applications for early free access, while paid versions will be available in the coming weeks.
With Tinker, Murati signals a clear direction — the future is not about building the largest model, but about efficiently enabling the creation of countless smaller, intelligently customized models that solve concrete tasks in science, education, and industry.
In brief: Tech World Highlights:
- Meta has recruited another prominent AI researcher from OpenAI, as Yang Song leaves to become lead researcher at MSL under Shengjie Zhao (also a former OpenAI employee).
- Tencent has open-sourced HunyuanImage 3.0, a text-to-image generation model claimed to rival the best closed industrial alternatives.
- Google released updates for its Gemini 2.5 Flash and Flash-Lite models, improving agent tool performance, efficiency, and instruction-following.
- Exa introduced exa-code, a tool that helps AI coding assistants find hyper-relevant web context, significantly reducing LLM hallucinations.
- OpenAI’s new Apps CEO Fidji Simo is reportedly recruiting a new lead to oversee ChatGPT monetization and advertising efforts.
AI Trending Tools:
