Hacker News new | past | comments | ask | show | jobs | submit
I don't know if people even need the best models anymore. Some people I've worked with have expressed a preference for older versions.

Things have gone very very fast. Two years ago, you couldn't reliably have AI write you a medium sized code project. It would spin in circles and apologize. Now, an old model can write it with only a little direction, and the programmer can have a reasonable preference against the newer model.

It seems to me open-weight models will put a floor on the pricing. They're slightly behind, but slightly behind last year meant nothing worked, now it means "just as good".

I agree - for the majority of people, there really is a hard limit on the technology. (Can it do what I want it to do in a reasonable amount of time with reasonable accuracy)

Many of my coworkers liked Opus 4.8, it was the moment they could actually use the models more reliably.

We're there, and moving forward doesn't really get us anything better. Most people are using them to format documents, spit out spreadsheets, pul together data points, and ask the models search questions.

We can already do that, it's only a matter of time before we realize we're paying 200/Mo for Jane in accounting to handle some of her work, when 5$ / Month would render the same benefit.

MOST people simply do not need frontier level intelligence.

The new models can pretty much run without human review. So yeah that flips the perspective: do we still need a human for that task, not which model can the human use.
I also expect AI to be more for machines and less for humans.

These providers are not selling chatbots that consumers can use as personal assistants anymore. Not even selling coding assistants. They want to manage the digital layer of every industry in the world.

yes indeed, and the models are already good enough, I dont think people fully understand yet how big of an impact this will be.
In general I think there's going to be diminishing returns on any given aspect of LLMs because not everything can be infinitely improved upon. If something can write a medium sized program near flawlessly, and I'm writing medium sized programs - then nothing else it can, or cannot, do matters.

Maybe bigger models will always just get progressively better at writing out bigger and bigger programs to the point that we're one-shotting fully functioning operating systems with wide driver support in the future, but again these features just don't matter if that's not what you want to do.

I think we're already limited by our own desires when it comes to coding.

The reason you can't one-shot something complicated is that you haven't generally considered all the important choices. You have to be asked for every choice, and complicated things have a lot of choices. But I think you can do something waterfall-esque, which used to be a big no-no in programming, if you just list out all the decisions meticulously.

I've had some success with doing that, just talking it over and thinking a lot about design decisions, making a document, and then letting it loose on the implementation. Inevitably, you find that there were contradictions or missing choices in the design, and you have to answer some questions.

So, not one-shotting, but it's my fault, not the AI's fault.

You can of course tell it to just fill in the gaps with whatever it thinks is reasonable, and then you'll have something potentially quite different from what you imagined.

The big improvement still coming down the pipe is that the models have better "taste" about filling in the things you didn't meticulously specify. A model from a year ago could make an amazing game, but you had to specify every detail precisely; now you can get similar quality results with a much more vague prompt.
That was the idea behind waterfall. You can look up how good that worked.
I think LLMs can be improved much more but most likely significantly diminishing economic returns on the investment for that improvement. Ultimately, it becomes too expensive for a few % more improvement at increasing cost. I believe many of the LLM providers are approaching that problem. So it's time for them to commercialize viable long term business, maybe keeping a team still trying to dig up improvements without blowing the bank.
Exactly. After horses, the Ford Model T [1] looked awesome. After a few decades, supercars started coming out. While it is nice to have a very fast Corvette, most people prefer to use a more basic SUV or sedan that is good enough for their day to day priorities.

[1]: https://en.wikipedia.org/wiki/Ford_Model_T

You beat me to it!

I was going to comment that most of us don't need a McLaren supercar (or a Hummer) for our grocery runs -- that Prius actually is Good Enough.

> put a floor on the pricing

ceiling