What's Next After RLHF? — Diogo Almeida, TypeSafe AI
OpenAI post-training pioneer argues Claude Code and ChatGPT are the same era, not successive ones
“the team I was part of basically invented post-training as a concept”
Diogo Almeida, co-author of GPT-4, ChatGPT, and InstructGPT, argues that the AI field is still in the 'ChatGPT era' and that agentic tools like Claude Code represent continuity rather than a paradigm shift. He frames the current moment as a fork between two extreme camps—one seeing AI as exponentially succeeding on every benchmark, the other seeing a bubble generating no real value—and positions his talk as an explanation of what actually comes next. The credibility of a post-training founder critiquing the era he helped create makes this a noteworthy signal on AI trajectory debates.