Simon Willison shares insights from OpenAI's internal RSI (Recursive Self-Improvement) day, highlighting how OpenAI's research team is using coding agents. The post notes a significant acceleration in AI spend per researcher in late July, speculated to coincide with internal access to the model later released as GPT-6 Astra.
Background
OpenAI has been increasingly adopting agentic AI workflows internally, with Recursive Self-Improvement (RSI) emerging as a key research initiative. GPT-6 Astra was anticipated as a major model release in late 2026.
- Source
- Simon Willison
- Published
- Sep 7, 2026 at 07:57 AM
- Score
- 6.0 / 10