RAGmacs - Pt. 1: Retrieving Deep Value in Emacs

How to use GPTel with LLM tool-use integration to retrieve introspection data from Emacs for better LLM results.

We're using GPTel with LLM tool-use integration to retrieve introspection data from Emacs and deliver high-quality results that are vastly better than the data embedded in LLMs weights. The tests are a success. This is just the beginning. The demos use manuals, completions, source code, docstrings, and other facilities to let the LLM traverse facts and more deductively arrive at useful conclusions. Source code for tools in the demo (subscribers exclusive!): github.com/positron-solutions/youtube/blob/master/lunch/09-r… Our Github Sponsors github.com/sponsors/positron-solutions GPTel for Emacs: github.com/karthink/gptel These techniques will scale across any language. By using these techniques within Emacs to make our Emacs usage more efficient, we are in a very tight feedback loop with very high headroom. Editors like Cursor don't really have anything to offer over programmable tools where the programmers employ LLMs to their problems rather than waiting for the IDE to come to them. Since Common Lisp is a good Lisp to reason about Lisp and other languages, it might just turn out that LLMs turn into a killer application for lisp-like languages. Since Lem is an Emacs, there's an intriguing possibility that we're witnessing a turning point in the making. #programming #llm #opensource TIMESTAMPS 00:00 Welcome 00:28 Retrieval Augmented Emacs 03:22 The Problems Being Solved 05:27 Not Even the Best Available 06:22 Why This is Huge