Writing
The author’s technical notes and essays on production ML, agentic systems, and the craft of building AI.
- Looking Ahead: What Multi-Token Prediction Teaches Us About Learning
The paper 'Better & Faster Large Language Models via Multi-Token Prediction' (Gloeckle et al., DeepMind, 2024) shows that training language models to forecast multiple future tokens simultaneously improves both quality and speed. Here's what it means for how we think about AI training — and why looking ahead matters more than you'd expect.