Post #3219977
2026-06-03 13:42 UTC
Replies (4)
-
@ChuckMcManis@chaos.social 2026-06-03 16:41
@lina@vt.social @sabik@rants.au @whitequark@social.treehouse.systems Perhaps it is more precise that some people saw 'claude' on a page of commits, looked at the actual code committed, and were horrified. The sin being not using claude, the sin being not checking what claude produced. That is just as bad as just checking in some random PR because "it seems to do what it says it does." rather than reading the code, agreeing that the code is correct, and *then* committing it.
-
@mirabilos@toot.mirbsd.org 2026-06-04 00:17
@lina@vt.social @sabik@rants.au @whitequark@social.treehouse.systems because that is nervertheless all of broken, plagiarism, exploitative, catastrophally harmful to the environment, and made with fashtech
-
@stilic@app.wafrn.net 2026-06-04 05:19
@lina@vt.social I mean… If one did a whole custom test framework with Claude and called it a day, it's not a light choice It implies that one will have to maintain yet another codebase just for testing To me, the way it was done made me stop trusting tridge and the project as a whole
-
@f4grx@chaos.social 2026-06-04 05:46
@lina@vt.social @sabik@rants.au @whitequark@social.treehouse.systems it is a reason to freak out when the usual way for ai agents to pass tests is to disable them. Test is the most important part of software. It's not grunt work for claude.