Thinking Machines Lab just launched their first product, Tinker Tinker is a distributed training API for fine-tuning language models 🔹Gives researchers & developers low-level control over algorithms and data while abstracting away the complexity of distributed training & deployment 🔹 Lets you fine-tune open-weight models like Llama and Qwen, including large MoE variants like Qwen3-235B-A22B The @Anyscale team had early access to experiment with the API and wrote up a simple end to end example

They’ve changed so much! DISNEY FAIRIES TINKER BELL AND FRIENDS DOLLS!