Add our position on AI use in coding Composr

AI doesn’t lack understanding — it lacks context. Give it examples, rules, and structure, and it will reflect them. Give it ambiguity, and it will reflect that too. Input = Output.

I think context and understanding go hand-in-hand. They're different ways to explain a similar principle. The difference is machine versus human.

AI needs context (data) to reflect what you want. Similarly, humans need understanding (knowledge, and knowledge of how to process that knowledge) to do the same thing. Some humans don't need "raw data" (or examples) to do what is needed, but a lot of us do. AI almost always needs data to do what is supposed to be done since its scope is much more limited than the human brain.

And even if context is provided to AI (and assuming the model allows enough tokens to handle all the context)... in the case of Composr, there is so much context that would be necessary to provide to AI that there are many points of failure; it is very easy for AI to make mistakes and misinterpret / hallucinate something (which starts a chain reaction of additional misinterpretations, leading to a faulty output). So even if you could give it context, the failure chance is still too high for me to say, "sure, you don't need to check it."

I will clarify a bit though on context in the docs. Thanks for the feedback.

I've also been looking into the possibility of being able to train a coding AI model on Composr's repository. As far as I know, that is not possible at this time, at least on my system. It's also something that is very low on my priority; just a fun thought.