Post #1232036

2026-04-01 11:55 UTC

@bodil I don't fundamentally disagree, but that raises another interesting point. Just how little those qualities have caused Claude Code any significant failures. Which is the other thing people say - if these tools can do anything of use where are the successful apps? Well kind of here, right? Claude Code is one of the most successful product launches of its kind, for quite some time. It works well enough to be considered a market leader. Apparently it's made of pure jank, but apparently, perhaps surprisingly, that doesn't matter so much? I mean, yikes. I don't like it all that much as a tooling approach, but it's coherent, performs well, and works reliably, when I have tried it. (I've put in a few dozen hours maybe) Me being one of nature's outliers forever, I think Claude desktop is a better product, whatever that's worth 🤷 - but I don't use that as a coding agent much either.

Replies (3)

@Orb2069@mastodon.online 2026-04-01 12:30

@cms It turns out when the user base has been conditioned to accept that the system randomly provides incorrect output ('hallucinates') when operating 'correctly', HOW it fails is not so important. Pull the lever and try again. @bodil

Open ##1232037
@bodil@social.treehouse.systems 2026-04-01 12:35

@cms I've been told for my entire career that code quality isn't important as long as the code does the job with acceptable frequency. And, well, if you write CRUD apps for companies providing non-essential services, I suppose that's generally correct. And if I find out the software for my car's lane keeping feature has so much as a single line written by Claude in it, I'll be fucking selling that car immediately with zero regard for the evident success of Claude's web UI.

Open ##1232041
@ratsnakegames@mastodon.social 2026-04-02 21:16

@cms @bodil Anthropic is operating at a massive loss that is offset by even more massive amounts of speculative capital. I do not consider that one of the most successful product launches.

Open ##1902576