Recursive Language Models organize an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →'s work around a programmable environment that can retain context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → outside the model's immediate prompt. In this interview, the approach's creator, MIT PhD student Alex Zhang, explains that the model uses code to inspect stored material, invoke tools and call another instance of itself on a smaller problem.
The intended benefit is control over how a large task is divided. Individual model calls can receive focused local problems while the broader context remains available in the environment. This is a proposal for how to organize model capabilities and memory; the interview's explanation does not establish a universal performance gain from adding recursive calls.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
