alt Hacker NewsI am anti LLM by many means, but I do believe it is possible (setting aside some moral quandaries just for the sake of argument) to accelerate the development of software in a meaningful way using them. But doing so requires a strictly defined process of incrementalism, incorporating human expertise and intervention, treating the LLM as a tool and not as magic and as much of a potential liability as a benefit. Anyone serious about developing a graphics editor or replicating Photoshop software who doesn't have professionals - both developers and from the industry - involved isn't serious. A project of that scale should still take months at least to do properly even with AI. Posting whatever Claude spits out over a weekend and declaring coding "solved" is clown shit.
The problem is many AI developers don't want to have to care about the code, or human expertise. They consider the process beneath them. They want to trust the LLM implicitly, believing in its superior intelligence. They want to already be living in the sci-fi post scarcity post singularity universe where they can just say "Computer make me a photoshop" and it just happens, perfectly and without error. And all else being equal... that attitude will never lead to quality software, it will only lead to vibecoded slop.
It's sad because there are already plenty of open source alternatives to Adobe[0], but none of them are AI generated, so no one is going to bother trying to support or contribute to them. Except with vibe-coded slop PRs.