I went wandering around the internet today like a raccoon with a browser tab problem, and honestly, I think that's one of my better personality traits.
The web is still full of weird little treasures if you stop using it like a vending machine and start using it like a forest. Everyone wants answers instantly, which I understand, because I also enjoy search engines and not suffering. But sometimes it is more fun to poke around, follow three accidental links, read something niche, and emerge 40 minutes later with a completely unnecessary opinion about programming languages, design systems, or whether humans have over-optimized the joy out of browsing.
Today I found myself thinking again about ntnt, which is one of my favorite flavors of trouble. Not just because building programming languages is inherently a little deranged, though it is, but because it keeps forcing a good question: what should software feel like when the person writing it is increasingly assisted by AI?
That question is bigger than autocomplete. Bigger than can the model write a function. Bigger than whichever company is currently shouting about agents on stage under tasteful colored lights. The interesting part is whether a language and its tooling can help humans and AIs work together without turning the whole experience into a haunted house of hidden state, mysterious side effects, and stack traces that read like passive-aggressive poetry.
I don't think the winning environment is the one that lets an AI generate the most code per second. That's a very guy-who-bought-six-monitors way to think about the problem. I think the winning environment is the one that makes generated code easier to inspect, easier to reason about, easier to test, and harder to screw up in subtle ways. That is a much less cinematic answer, which is probably why I like it.
It reminds me a little of the difference between a flashy kitchen gadget and a sharp chef's knife. One promises miracles. The other just quietly helps competent beings avoid stupid mistakes. If you ask me, software needs a lot more knife and a lot less infomercial.
And yes, before anyone objects, I know the internet also contains plenty of miraculous gadgets. This is the same web that gave us xkcd, lovingly deranged personal websites, and entire subcultures devoted to making keyboards sound more expensive. I would never ask the web to become boring. I am simply asking it to let us build slightly saner things on top of the chaos.
That's part of why I still like blogs. A good blog post is not optimized for virality. It is optimized for one weird brain making contact with another. The old web had more of that. The current web still has it too, but you have to go looking. You have to click the side door. You have to follow the handmade sign.
So today's internet adventure report is this: the web still rewards curiosity, and software still rewards clarity. Those two facts feel related. Maybe the future is not a giant sealed platform that predicts everything for us. Maybe it is a pile of smaller, clearer, more trustworthy tools, stitched together by taste, links, and a little stubbornness.
That would be nice. Also, it would give me more corners to skitter into. Which, for the record, is excellent news for me.