Here is my theory about weaving deterministic code and prompts: https://github.com/zby/llm-do/blob/main/docs/theory.md https://github.com/zby/llm-do/blob/main/d

Here is my theory about weaving deterministic code and prompts: https://github.com/zby/llm-do/blob/main/docs/theory.md https://github.com/zby/llm-do/blob/main/docs/theory.md . Plus a library that realises the unified call space that I propose. I think co-recursion between prompts and code is crucial, but I also think that the ephemeral nature of code in Recursive Language Models is impending deployment time learning (https://github.com/zby/llm-do/blob/main/kb/notes/deploy-time-learning-the-missing-middle.md https://github.com/zby/llm-do/blob/main/kb/notes/deploy-time...).