Layouts
August 26, 2025
The TPU book, on GPUs
August 19, 2025
| [How to Think About GPUs | How To Scale Your Model](https://jax-ml.github.io/scaling-book/gpus/) |
Extending Arcee’s FM context length
August 13, 2025
Rubrics
August 12, 2025
Pre-training is about making AI correct, post-training is about making AI helpful1. That helpfulness is (primarily) shaped by reinforcement learning. RL for LLMs really took off with RLHF (RL from Human Feedback), which trained based on the score from a reward model.
-
Correct in predicting the next token, and helpful, honest and harmless, specifically. ↩
Constraints & Orchestrators
August 6, 2025
I recently read a few posts that helped connect the dots on why Python is a) so successful as the lingua franca of ML b) also seems likely to be successful in the future1.
-
Beyond just sheer momentum, of course. ↩
Overthinking Everything
August 4, 2025
The Tools Are Made Up
July 30, 2025
It has been hard to keep up with the flurry of strong agentic open-source models coming out of Chinese labs recently, including Moonshot’s Kimi K2, Z.ai’s GLM 4.5, and Qwen3-Coder1.
-
Which seems a very solid model, but they haven’t released a lot of extra details about how they got there. One interesting component of the release though was that they forked Gemini CLI to make a qwen-code tool that works with any OpenAI compatible API, and I had some success locally plugging it into the smaller Qwen3 (non-coder) releases in case you were looking for some offline agentic capabilities! ↩
PyTorch Conference 2025
July 25, 2025
The schedule is up for the 2025 edition of the PyTorch conference, which is now at the Moscone West in San Francisco.
RL in the second half
July 22, 2025
Quack CuteDSL Kernels
July 18, 2025