Why Normal People Aren’t Using AI Agents

What we want is software that behaves predictably. Since LLMs don’t do that, we don’t want them or their “agents.”

For fuck’s sake.

So you’re decided not to use a useful tool, because you don’t like it. Gotcha.

I once thought the fediverse was full of cool, techsavy people. Turned out it’s full of old people yelling at clouds.

You have your definition of useful, I have mine. Maybe predictability is important to many others, but not to you.

Besides, “not liking it” sounds like a pretty legit reason for not using something, lol. Would you listen to music if you didn’t like it?

Music is NOT a tool.

Its more like not using a screwdriver and using only a hammer. And that would be fine - if you only have nails.

And this argument about different definitions is fundamentally dishonest under a comment that say broadly we don’t like AI. I have my definitions of interesting, useful and fun. But if I would make a post about, I would be torn apart here, lol.

If I’m trying to fasten something, and the screwdriver fucks up half the time, then yeah I’m going to nail it together instead. Or use glue, or anything but an unreliable screwdriver.

Thats the thing. AI doesn’t fuck up all the time and has gotten genuinely useful for some things in the last months.

But a huge amount of you doesn’t know this or doesn’t care, generating an insufferable, ignorant circlejerk, drowning out every interesting discussion you can have about this topic.

That is what people have said, every day, on repeat, for the last three years. Explain why it is better now instead of three months ago, when it was definitely not fucking up all the time and then I will copypaste that back to you for what was said three months ago for the same question.

Which standards are you gonna hold it to? The benchmarking involves using standardized testing, but they typically just feed those questions into the testing data, so the tests have to be constantly updated, and when they are, you will not be surprised to learn that the machines goes back to fucking up the answers all the time.

Experts in their fields do not use this shit because it is wrong on the things they know, but, oh wow, it must be totally right about the things I don’t know anything about? Every time.

Oh, just shut the fuck up.

www.theregister.com/software/2026/03/…/5226256