Random Funber Generator
2026-10-22
My vibe coding experiment, debloatmail, has been going just as one would expect it to. Dunning my Kruger, let me tell ya. It's crazy how people think LLMs are this profoundly world-changing thing; people with no concept of how computers work see saddle-stitched benchmarks and think the semantically-pumped random number generator is some kind of revolutionary thing. If you have no concept of heuristics, the models make you think they've got everything figured out, but the absolute hallucination—and the lack of context‑maintenance on any complex project—worries me. If you don't have any concept of that, they can work as a fair enough teacher—patient, understanding of your small, temporally limited cognitive power—but only if you actually take the time to read their straight-up schizophrenic documentation and schematics. People see guys like DHH—who already understand their own heuristic preferences—produce larger volumes of code more quickly, and think that's how it would be with any developer. Unfortunately, the shit is like diarrhea; it's too fast, too loose, and can't hold any kind of meaningful shape outside of the human prompting it.
What's worse is that if you don't pull the brakes to understand the code you're making it write, it will snowball into an absolute clusterfuck with no way to self‑regulate. I can see how developers who work with proprietary legacy codebases hate this shit; the technical writers who got laid off can't even give the models proper context, and executives are so out-of-touch, they can't even grasp what's wrong with their completely non-existent sense of how this shit actually works. I mean, they're not the ones having to actually write any fucking code.
There's a lot of stupid vibe coders out there who think that the more money they spend and the more compute they borrow, the faster their pointless charades will work out. I still want to work to understand programs and software, and while I have no illusions about the power of these models, it is a sort of weird luxury to be able to work on something of this scale all by myself. That said, if there's anyone smarter than me who can give me advice, please send an email.
Yet despite this silly project I've given myself, I still find myself at the same spiritual crossroads.
Most art is a weird form of escapism. As much as it allows us to confront parts of ourselves that we may not always want to—or be able to—access, the process of making art is a different thing entirely. To make art, one has to reach outside of the confines of what they can see, and make something new. In those outer bounds, it's easy to want to stay there—to become bound to a world that otherwise doesn't exist. Even in this longform creative nonfiction project, that temptation still exists. As much as I want to confront myself as a means of self-discovery or some other intangible benefit, I still find myself lost in the flow of writing this prose—to know that it has no effect outside of this domain or the headspace of the current reader is a deep, unfathomable tragedy. There is no joy, no catharsis in this process; it exists solely as a means to forget the inevitability of things that bear a greater, more circumstantial weight. When this spirit departs from my soul, there will be a great deal of pain in it; even if there is a possibility of rest for my poor, tired, blue little blob, I know that the ultimate pain it must endure is when these spirits truly leave it.
Until then, I will keep distracting myself.
Support my work with Liberapay
Bitcoin address: bc1qtgqvj6qjxnaxkns20x5rcwnxvv3jqzhduvvxfc