Evidence: Claude Writes Much of Anthropic's Code, R&D Roles Expand

imjustnewatai · x · 2026-08-06

This thread provides factual evidence and job listing backups for the author's argument on the collapsing AI R&D calendar and recursive self-improvement.

Anthropic: Claude already writes a significant amount of committed code, with its data loop closing in hours or days. RL goals include continual learning and end-to-end software engineering.

OpenAI: The model flywheel automates experiments, evals, and deployment. Custom-infrastructure roles are moving AI-native silicon from lab validation into production data centers.

Recruitment Signals: Live roles target AI scientists, chip-design RL, and measuring AI R&D acceleration, proving AI is actively entering the entire R&D pipeline.

Related event: AI R&D Cycles Shrink as Recursive Self-Improvement Emerges(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →