yep. noticed this too. but I think this is due to short context LLMs have. now my tools, mostly browser LLMs copy-paste code, are analysing small sub-problem at

yep. noticed this too. but I think this is due to short context LLMs have. now my tools, mostly browser LLMs copy-paste code, are analysing small sub-problem at time. once LLMs will see whole repository + history of that repository, likely they will write abstractions that can be reused