Teortaxes Describes Working RSI as RL Environment Building

Teortaxes, a pseudonymous AI developer known for promoting DeepSeek, posted that current functional RSI consists of building and debugging RL environments rather than other approaches. The post notes surprise that this disaggregated mode received no mention in LessWrong discussions. It includes an attached image and references another account. The statement addresses what the author sees as actual progress in recursive self-improvement today.