A Year of Tinker: Thinking Machines on Turning Expertise into Model Weights You Own
miramurati · x · 2026-10-16
Thinking Machines Lab published a one-year retrospective of Tinker, its post-training API that reduces fine-tuning to four API functions (sample, forwardbackward, optimstep, savestate) while users keep their data and training loop and Tinker manages GPUs and checkpoints.
Co-founder Mira Murati highlighted watching people turn their own expertise into model weights they own, arguing the custom-model market stayed small not because fine-tuning lacks value but because no post-training platform was versatile, reliable, fast, and cheap enough.
Related event: Thinking Machines Marks Tinker's First Year of Accessible Fine-Tuning(3 posts)→
More from coding & agent
- Dev argues coding models now beat nearly all engineers in every branch of software — zack_overflow · 2026-10-16
- 75-80% of my AI requests now go through Dots: a power user's month-long review — nickbaumann_ · 2026-10-16
- sleepy-llm: A Reverse Proxy That Auto-Wakes and Shuts Down Your Local LLM Server — ac101m · 2026-10-16
- Nous Research's Open-Source Hermes Agent Makes TIME's Best Inventions of 2026 — NousResearch · 2026-10-16
- Law firm tests show Claude beating paid legal AI tools like Legora — jkubicki · 2026-10-16
- Yacine mocks the idea that you can skip reviewing LLM code on web apps — yacineMTB · 2026-10-16