[crystal_revenge]

> There's a mass psychosis happening

There absolutely is but I'm increasingly realizing that it's futile to fight it.

The thing that surprises me is that people are simultaneously losing their minds over AI agents while almost no one is exploring playing around with what these models can really do.

Even if you restrict yourself to small, open models, there is so much unexplored around messing with the internals of these. The entire world of open image/video generation is pretty much ignored by all but a very narrow niche of people, but has so much potential for creating interesting stuff. Even restricting yourself only to an API endpoint, isn't there something more clever we can be doing than re-implementing code that already exists on github badly through vibe coding?

But nobody in the hype-fueled mind rot part of this space remotely cares about anything real being done with gen AI. Vague posting about your billion agent setup and how you've almost entered a new reality is all that matters.

[svara]

Yes, it's been odd to observe the parallels with the web3 craze.

You asked people what their project was for and you'd get a response that made sense to no one outside of that bubble, and if you pressed on people would get mad.

The bizarre thing is that this time around, these tools do have a bunch of real utility, but it's become almost impossible online to discuss how to use the tech properly, because that would require acknowledging some limitations.

[crystal_revenge]

Very similar to web3! On paper the web3 craze sounded very exciting: yes, I absolutely would love an alternate web of truly decentralized services.

I've been pretty consistently skeptical of the crypto world, but with web3 I was really hoping to be wrong. What's wild is there was not a single, truly distributed, interesting/useful service at all to come out of all that hype. I spent a fair bit of time diving into the details of Ethereum and very quickly realized the "world computer" there (again, wonderful idea) wasn't really feasible for anything practical (I mean other than creating clever ways to scam people).

Right now in the LLM space I see a lot of people focused on building old things in new ways. I've realized that not only do very few people work with local models (where they can hack around and customize more), a surprisingly small number of people write code that even calls an LLM through an API for some specific task that previously wasn't possible (regular ol'software build using calls to an LLM has loads of potential). It's still largely "can some variation on a chat bot do this thing I used to do for me".

As a contrast, in the early web, plenty of people were hosting their own website, and messing around with all the basic tools available to see what novel thing they could create. I mean "Hamster Dance" was it's own sort of slop, but the first time you say it you engaged with it. Snarg.net still stands out as novel in it's experiments with "what is an interface".

[neoromantique]

>As a contrast, in the early web, plenty of people were hosting their own website, and messing around with all the basic tools available to see what novel thing they could create

I'm hoping that the already full of slop centralized platforms now with LLM fueled implosion will overflow and lead to a renaissance of sorts for small and open web, niche communities and decoupling from big tech.

It's already gaining traction among the young, as far as I can see.

[AstroBen]

There's a good reason for that. The end result of exploring what they can actually do isn't very exciting or marketable

"I shipped code 15% faster with AI this month" doesn't have the pull of a 47 agent setup on a mac mini

[nosianu]

> The thing that surprises me is that people are simultaneously losing their minds over AI agents while almost no one is exploring playing around with what these models can really do.

I think we all do???

Even if I'm not coding a lot, I use it every day for small tasks. There is not much to code in my job, IT in a small traditional-goods export business. The tasks range from deciphering some coded EDI messages (D.96A as text or XML, for example), summarizing a bunch of said messages (DESADV, ORDERSP, INVOIC), finding missing items, Excel formula creation for non-trivial questions, and the occasional Python script, e.g. to concatenate data some supplier sent in a certain way.

AI is so strange because it is BOTH incredibly useful and incredibly random and stupid. Among the latter, see a comment in my history I made earlier today, the AI does not tell me when it uses a heuristic and does not provide an accurate result. EVERY result it shows me it shows as final and authoritative and perfect. Even when after questioning it suddenly "admits" that it actually skipped a few steps and that's not the correct final result.

Once AI gets some actual "I" I'm sure the revolution some people are commenting about will actually happen, but I fear that's still some way off. Until then, lots of sudden hallucinations and unexpected wrong results - unexpected because normal people believe the computer when it claims it successfully finished the task and presents a result as correct.

Until then it's daily highs and lows with little in between, either it brilliantly really solves some task, or it fails and that includes telling you about it.

A junior engineer will at least learn, but the AI stays pretty constant in how it fails and does not actually learn anything. The maker providing a new model version is not the AI learning.