
Claude's Edge and Post-Training Art

The $1B Al company training ChatGPT, Claude & Gemini on the path to responsible AGI | Edwin Chen
Topic 3 of 8
Published Dec 7, 2025 · Clip 13:32–17:38 (4:06)
Claude's edge in coding and writing stems from superior data choices, including human data gathering methods, synthetic data balance, and post-training 'taste' in objectives like visual design vs. efficiency. Labs face infinite decisions on data mixes, benchmarks vs. real tasks, and values shaping model behavior. Edwin notes an art to post-training where company principles determine model sophistication, leading to differentiated behaviors beyond commoditization.