On 21 October 2026, Wired reported on Flint, Recursive’s open-weight coding model, in a story about recursive self-improvement — using AI to build AI.
Writing a short program could involve asking a coding model such as Flint for a draft, then checking whether the code works. Access to a model’s weights could also allow its behaviour to be examined beyond the suggestions it produces.
Will Knight, Wired’s senior writer covering AI, said he had tested Flint for several days and described it as “impressive”. He wrote that the model suggested fully recursive self-improvement was “getting closer”.
OpenAI has ruled out fully autonomous self-improvement for now and urged global standards. Recursive argues that the benefits of using AI to build AI outweigh the risks, Wired reported.
