I generally despise AI.
Even ignoring the ethical and existential concerns, if I could trade every second I’ve had to spend reading/conversing about AI for time spent actually working, I’d probably be more productive. And happier.
But I was recently forced to partake, as we had a security vulnerability report (AI-discovered, I’m quite sure) that was way over my head, and I needed to tap into the Pluribus for some help. The end result was craftcms/url-validator, a new package that helps protect against SSRF vulnerabilities for applications that support uploading files via user-provided URLs. Just the sort of thing I always pictured myself worrying about when I grew up.
That experience opened my eyes to the potential, a bit, I guess.
I love to code, but if I’m being honest, there are definitely aspects of it that I don’t love. Such as:
- Resolving merge conflicts
- Line-by-line code reviews with all potential breaking change/security side effects in mind
- Debugging complex bugs, especially when third party plugin/project-specific code is involved, which I have zero incentive to commit to memory
- Debugging npm issues
- Writing tests
I’ve started leaning on it more for those sorts of things. And I have to admit, it’s pretty nice to not spent so many brain cycles on shit work.
So I guess I’m kind of coming around. Slowly. Contentiously.
I’m still mad about every app that tries to cram it down my throat at every opportunity. And fuck everyone using it for content creation. That’s a human’s job.
Clive on Mastodon ∙ Wednesday
@brandon much in sympatico, Brandon. For writing, my concentration these days, I use best llm levels to dig into research, where it's effective if you treat it as a grad assistant, keep asking back on any corners, but especially on ideas that come to you, for these are exactly what it won't find on its own by the nature of the algorithm, but will shift very nicely into with the conversation. Writing itself, Abso ...