This is Experimental

INFO

🙋‍♂️ This content is fully written by a human

I'm not going to pretend that this is the latest and greatest technique that's proven to get more effective use out of your AI.

This is mostly for me to play around with different prompting techniques.

What is really motivating this, is to see if I can get effective use out of smaller models by using various prompt/context decomposition techniques.

See also the associated project ThunderJar which is intended to measure how different models/harnesses/prompts behave given the same task.