Comment on Devs gripe about having AI shoved down their throats
Septimaeus@infosec.pub 5 hours agoFor example the tools for the really tedious stuff, like large codebase refactoring for style keeping and naming convention adherence, those tools have become a lot more powerful than what I remember from a decade ago.
While I’ve only experimented a little with some the more explicitly generative LLM-based coding assistant plugins, I was impressed (and a little spooked) at how good they often were at guessing what I was doing way before I finished doing it.
I haven’t used the prompt-based LLMs at all, but I’ve watched nearby devs use them for stuff like manipulating a bunch of files in a repeated pattern, breaking up a spaghetti method into reusable functions, or giving a descriptive overview of some gnarly undocumented legacy code. They seem pretty useful, but I don’t think I’ll be able to use them fluidly until I can host them locally.
MangoCats@feddit.it 3 hours ago
In the work I have done with Claude over the past months, I have not learned to trust it for big things - if anything the opposite. It’s a great tool, but - to anthropomorphize - it’s “hallucination rate” is down there with my less trustworthy colleagues. Ask it to find all instances of X in this code base of 100 files of 1000 lines each… yeah, it seems to get bored or off-track quite a bit, misses obvious instances, finds a lot but misses too much to say it’s really done a thorough review. If you can get it to develop a “deterministic process” for you (shell script or program) and test that program, then that you can trust more, but when the LLM is in the loop it just isn’t all there all the time, and worse: it’ll do some really cool and powerful things 19/20 times, then when you think you can trust it it will screw up an identical sounding task horribly.
I was just messing around with it and I had it doing a files organization and commit process for me, was working pretty good for a couple of weeks, then one day it just screwed up and irretrievably deleted a bunch of new work. Luckily it was just 5 minutes of its own work, but still… that’s not a great result.