Research acceleration: The view inside OpenAI

OpenAI published two pieces on recursive self-improvement on the same day, a post on research acceleration and Jakub Pachocki's essay An Alien Mind. Simon Willison points out that the first one uses RSI without ever expanding the acronym, which tells you who it was written for.

An acronym you do not define is one you expect the reader to already carry. Willison reads RSI as the term taking over from AGI inside the lab, and the writing supports that reading. It is being used as a settled house word rather than introduced as a claim.

For anyone shipping on their API, the practical content here is zero. No model. No endpoint. No pricing or deprecation. What it does give you is the vocabulary the lab is standardizing on, and lab vocabulary tends to arrive in product copy and changelogs a couple of releases later.

I am filing both under positioning until something with a version number behind it uses the same word.

https://simonwillison.net/2026/Sep/6/research-acceleration-the-view-inside-openai/

#dev