@scaling01 Says Astra and Kimi-K3 Can Take More RL

@scaling01, identified in the post as Lisan al Gaib and creator of the custom LLM reasoning benchmark LisanBench, posted that Astra is in a similar position to Kimi-K3. The statement adds that both models can be RL'd much more. The post appears among visible replies on X and contains a link to another tweet. No first-party announcement, independent corroboration, or additional details about either model appear in the supplied evidence.