[zkry]

Logically this makes sense: every model has a context size and complexity capacity where it will no longer be able to function properly. Any usage of said model will accelerate the approach to this limit. Once the limit is reached, the LLM is no longer as helpful as it was.

I work on full blown legacy apps and needless to say I don't even bother with LLMs when working on these most of the time.