The one thing I would have wanted from machine learning was my IDE detecting when I repeatedly do similar things and offering things the IDE already has to make it easier or to generate runners/tiny GUIs for doing the thing
Instead we get randomly generated PRs with plausible sounding descriptions.
@oli To me it feels like the current wave of machine learning applications is driven by very unimaginative people who look for ways to cheat the system by appearing to be doing the work expected of them but not actually having to do it.
Your suggestion is instead focusing on being able to actually allow achieving better efficiency or quality of work by people who care about their actual output, not just the appearance of being busy.
@frizi well, it's not clear where the line is. Regex search and replace is great for refactorings. So are tools that understand the language (rust analyzer's refactoring things like changing tuple variants to struct variants).
And there may be a use case for not writing regexes but just writing plain text instructions and reviewing the diff.
But... so far I only saw what you described.
Which may likely be bias because of all the folk using it responsibly don't show up with bogus PRs
oli
frizi 🦀🏳️⚧️