I have developed a reflection webui that gives reflection ability to any LLM as long as it uses openai compatible api, be it local or online, it worked great, not only a prompt but actual chain of thought that you can make longer or shorter as needed and will use multiple calls I have seen increase in accuracy and self correction on large models, and somewhat acceptable but random results on small 7b or even smaller models, it showed good results on the phi-3 the smallest one even with quantization at q8, I think this is how openai doing it, however I was like lets prompt it with the fake reflection 70b prompt around. but let also test the 01 thing, and I gave it the prompt and my code, and said what can I make use of from this prompt to improve my code. and boom I got warnings about copyright, and immediately got an email to halt my activity or I will be banned from the service all together. I mean I wasn't even asking it how did 01 work, it was a total different thing, but I think this means something, that they are trying so bad to hide the chain of though, and maybe my code got close enough to trigger that.