pull down to refresh
That's fair, and you've obviously cut your teeth by doing the hard work of turning slop into real software
I just think AI is increasing what noobs can accomplish, remembering that only pro's like yourself can confidently present a project to a paying customer with assurance of reliability
I'm not worried about paid projects. If you mess those up and your customer is half-competent, you'll pay for your own fuckups. Damage clauses are no joke.
In a professional setting, I'm much more worried about FOSS, and specifically, open source libraries that are adapted into everyone's apps by LLMs (also for those paid projects.) Not only is it a cesspool of vulnerability in the delivery mechanism (all those npm malware infections of late) but underlying libraries that were once carefully crafted are now getting sloppy. A minor update can fuck you up. If you don't know what you're doing, you don't know to even ask the bot to assess your dependency.
Now, if anyone vibe-codes something for themselves that solves a problem they have, I have no issues with that at all. I encourage that. But everyone wants to be famous now from the work of an LLM that itself has no fucking clue (because it is not some sentient being, despite what the CEOs claim.) Thus, the slop gets published and marketed. And when you become a user of said product because everyone and their dog was telling you on X and nostr (and SN!) how fucking awesome this new app is, you're at risk too. And if 99% of devs don't check before they install, think about which users are going to get rekt: all of 'em.
Excellently laid out concerns and I do agree with you on those, noobs wouldn't see the shit lurking deep in the trenches like you. Good points well made
It removes barriers, we agree on that. Capability is a big word though as it depends on the specifics of the goal you're trying to achieve. If we're talking nothing-at-stake, zero-requirements throwaway stuff, then yes, everyone is super capable now. Has been since GPT-2.
I know because I helped a fair share of those "noobs" by completely rewriting the GPT generated prototype code to their ideas into productized software. Back then, not a single line survived. Nowadays, maybe some lines will survive. But more often than not, even the boilerplate is poor.