RSI is here, just disaggregated: DeepSeek using LLMs to design algorithms

teortaxesTex · x · 2026-09-10

Responding to a claim that DeepSeek isn't "RSI-pilled" — since the lab itself says data/environment pipeline engineering beats algorithmic novelty in post-training — teortaxesTex argues this is just data-driven RSI as stage one: V4.1 Flash swarms are synthesizing training data and DeepSeek staff are using internal and external LLMs to design algorithms. RSI, he contends, has already arrived, merely disaggregated across the pipeline.

Original post →

More from AGI Musings

AGI Musings channel →