GLM 5.2 and the coming AI margin collapse • Hacker News Board

Re: GLM 5.2 and the coming AI margin collapse

#491

Post by andriy_koval » Wed, Jul 08, 2026, 5:46 PM UTC

sajithdilshan wrote: post #469

andriy_koval wrote: post #459

Earlier quoted context omitted.

meta researchers from eu will create new company and build model using knowledge they have.

In an imaginary world yes. Just because they know how, doesn’t mean that they can. It’s not like you can code a new model, you need compute power to train it, which they don’t have in EU. Also meta could just take the best mind in EU research centres to US with millions of dollars as yearly payments

I agree is that money is an issue, but not geo-boundaries. If EU decides to commit XXXB they may have a chance to catch up eventually.

Re: GLM 5.2 and the coming AI margin collapse

#492

Post by rdbl27 » Wed, Jul 08, 2026, 8:08 PM UTC

Dumblydorr wrote: post #278

nostrademons wrote: post #157

Earlier quoted context omitted.

Somewhere else in the comments here, someone else remarked "Individuals perhaps [move to the new models], but not organizations." That's illustrative. The mechanism by which organizations are forced to update their technology, move to more competitive suppliers, and cut costs is a recession. In one, every business that doesn't do so goes bankrupt, and what's left are the more efficient businesses that have adopted te…

“Brought on by government edict” No, brought on by a novel pathogen that killed 10 million people. It would’ve been much worse without government action.

The case of Sweden entirely falsifies the theory that "it would've been much worse without government action."

No government action whatsoever was taken in Sweden. None. Death rates were no better or worse than adjacent countries that intervened extensively.

Ergo, government intervention was irrelevant one way or another.

Re: GLM 5.2 and the coming AI margin collapse

#493

Post by Paradigma11 » Thu, Jul 09, 2026, 2:54 PM UTC

mynameisbilly wrote: post #429

jeremyloy_wt wrote: post #342

Earlier quoted context omitted.

And if the engineer bills for $100/hr or more, the trade off is worth it

Why is everyone still operating under the assumption the current token costs will remain so heavily subsidized? We could see $200-400/hr in token costs once these companies need to turn a profit

Because then people will just flock to open weight models that are even cheaper and Anthropic/OpenAI will go bankrupt. Then somebody will buy those companies for cents on the dollar and run a profitable token selling business for today's prices without doing any new research. New model releases will slow to a crawl but that's it.

Re: GLM 5.2 and the coming AI margin collapse

#494

Post by gnarbarian » Thu, Jul 09, 2026, 9:13 PM UTC

yogthos wrote: post #488

gnarbarian wrote: post #477

Earlier quoted context omitted.

if you give me a smarter model I will find a more complex problem for it to work on that will require an exponential number of sub-agents. the smarter the model the more capable it is to marshal resources to accomplish the goal. this means you don't need a smart person to be able to take advantage of this. a smart model will be able to do the same.

The question is that of cost. Sure, there's no downside to having a smarter model, but if you have to pay orders of magnitude to use it, then it's just a waste of money. If you can get around on a bicycle then you're not gonna buy a monster truck.

you don't know me very well lol.

You're right of course when you're talking about people. I believe there is a lot of low-hanging fruit for optimizing the performance of these models. that's actually what my startup is doing.

Re: GLM 5.2 and the coming AI margin collapse

#495

Post by ryukoposting » Tue, Jul 14, 2026, 3:50 PM UTC

KronisLV wrote: post #435

ryukoposting wrote: post #379

Earlier quoted context omitted.

> Not even doing anything crazy, just parallel long form work on 2 projects... and at most 3 parallel code review sub-agents. I had to read this sentence twice.

> I had to read this sentence twice. Heh, I mean I'm not running Gas Town: https://steve-yegge.medium.com/welcome-to-gas-town-4f25ee16d... so as far as AI stuff go, I'd assume that I'm not too much of an outlier. Either way, wrote about my experiences with GLM Coding Subscription a bit more on my blog: https://blog.kronis.dev/blog/z-ai-s-glm-5-2-is-a-great-model... I do suspect that there's plenty of people who'd use…

Reading Yegge's post made me crack a smile. He's beating the "we need Kubernetes for agents!" drum so damn hard and people aren't responding. I'm certain it's because 95% of the people he's talking to have no clue what Kubernetes actually does. They just know it's there.

Hell, every time I type "Kubernetes" I'm reminded of the fact that I've never heard someone say it out loud. I have no idea how to pronounce that word. Cube-r-neets? Koober-net-ees? Koober-nates? I'm a firmware guy so I'm pretty far removed from the server-touchers, but still.