I’m currently using the Black Formatter extension, but it’s not automatically resolving issues related to Flake8’s rule (E501 – line too long). It seems that LLM models struggle to accurately assess line length when suggesting or formatting code, resulting in persistent linting warnings.
I’d like to know if anyone else is experiencing this problem and if you’ve found any solutions or workarounds.
Does anyone have suggestions or experiences to share?
Not using Black Formatter, but others (depends on language), but best is to use a formatter that lets you set this.
AI does not count characters 
it can assign higher probability to shorter lines if you mention this in a rule but it would create other issues likely.
Instead AI works with code it sees (training data, prompt, files you attached etc..) and generates output based on that.
Personally I rely on formatters (actually not even those from extensions, but langauge based formatters).