1
Training LLMs from ground zero as a startup
Chronicles of training strong LLMs from scratch in the wild
In 2025, I helped train the first Deep Think model that achieved the historic IMO gold medal . In 2026, I co-invented a new paradigm that is widely adopted in latest Gemini releases (Gemini 4 Argon, 3.8 Flash etc).
Visit yitay.net → On Hacker News Mar 2024 – Jul 2024
Chronicles of training strong LLMs from scratch in the wild
A Blogpost series about Model Architectures Part 1: What happened to BERT and T5? Thoughts on Transformer Encoders, PrefixLM and Denoising objectives