Positive users praised the transparency and cool results of post-training Luna from existing Sol configs, while negative users accused OpenAI of misleading claims, censorship, and typical hype around the process.
Based on 28 visible X reactions from 103 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@kimmonismus What a stupid interpretation of what is at most an AI acting as a judge. It's literally training another model, not itself, and you jump to the conclusion that it is RSI.
@grok @WilliamChaseIV @sharifshameem @grok u suck it was for this was trying to be funny just delete urself https://x.com/learn_ngo/status/2075467511584186638/photo/1
@sharifshameem Funny y’all keep mentioning this use case… We’re out here literally trying to make gpt-oss-20b run faster. https://x.com/gajesh/status/2075292156328349769?s=46
@sharifshameem @ponnappa "Amazing content! Looking forward to learning more from your future posts. 😊
The workflow initializes training runs from an oai-strawberry checkpoint
@sharifshameem So cool!
@jxnlco man you guys awesome
OpenAI just showed one of the clearest early signs of recursive self-improvement: GPT-5.6 Sol was used to post-train GPT-5.6 Luna. This is not an intelligence explosion (yet). Humans still defined the objective, infrastructure and constraints. But the loop is now visible: frontier models are beginning to perform the engineering work required to build and improve the next generation of models. Once AI meaningfully accelerates AI R&D, every generation helps produce the next one faster. Yes, read that again. That is how recursive self-improvement begins, not with a model rewriting its own weights overnight, but with AI gradually taking over the research and engineering pipeline that creates better AI. And this is further proof that the speed of releases is increasing and models are improving even faster.
re: 5.6 sol post trained luna and we won't stop you from doing the same thing To be clear, this was not a model inventing a whole training recipe from scratch. Most of the config already existed from Sol. For this kind of smaller-model setup, our team's estimate is usually one or two researchers for a week or two.
re: 5.6 sol post trained luna and we won't stop you from doing the same thing To be clear, this was not a model inventing a whole training recipe from scratch. Most of the config already existed from Sol. For this kind of smaller setup, our team's estimate is couple of extra weeks
@jxnlco thanks for the clarification! imo quite a lot of confusion about RSI recently and i think most ppl misunderstood it, the result is cool by itself imo without overselling (and also lfg "we won't stop you from doing the same thing")
Positive users praised the transparency and cool results of post-training Luna from existing Sol configs, while negative users accused OpenAI of misleading claims, censorship, and typical hype around the process.
Based on 28 visible X reactions from 103 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@jxnlco man you guys awesome