aurareturn

Human brain can do calculations in the back of their head without realizing it. Verbalizing does not mean it's all they thought about.

Open weights is not a CCP thing. It's a business decision because Chinese labs do not have a good distribution strategy in the west/outside of China.

It's my long standing view.

There are already rumblings of CCP wanting to control export of top tier frontier models.

Chinese labs are gated by lack of compute, at least until they can build their own EUV machines.

At some point, they'll stop releasing models for free and will need a license.

Personally, I think Anthropic and OpenAI will introduce enterprise plans such as $2,000/month/engineer plans that will give 100x more tokens than the current max plans and then charge API pricing.

Yes, margins will go down a bit but the frontier labs still have some leverage.

It's probably true that open weight models are "good enough" for most software engineering tasks.

But the goal posts will move and I think frontier labs will find new areas that have a need for absolute frontier.

Depends on what the goal is and the skill of the engineer right?

Give a verifiable benchmark for the AI agent to hit such as reduce token processing time by 5% without affecting output.

How does it change if you’re building via subscription?

My point is that people will complain about GenAI but they actually love/enjoy AI generated content or projects as long as they fit what they're looking for.

What is unsustainable about Anthropic's numbers?

What are his numbers and analysis?

So he can get all of his conclusions and predictions wrong based on his numbers and analysis, and then you think his value is in his numbers and analysis?

HN reminds me a lot of Instagram.

People complain about AI generated content, then proceeds to give a dancing cat AI reel 1 million likes.

People here complain about vibe coded apps, then proceed to upvote this to the front page.

Arguably, it was cheaper.

By the way, we get way more value from our Anthropic subscriptions than any Chinese subscription.

Yes, but the trend is pretty clear based on the link I posted.

When was the last time Ed Zitron got a prediction/claim correct?

I disagree. Our devs use Claude for coding. When OpenAI had Astra, no one switched even though the company allowed OpenAI models.

If OpenRouter is a proxy, then their growth will pick up again in Q4: https://openrouter.ai/anthropic

In Q3, I think they struggled with the government ban on Mythos/Fable, Opus 5 not being so good, OpenAI models being more token efficient, and they decided to dedicate a higher percentage of compute to training.

If OpenRouter is a proxy, then their growth will pick up again in Q4: https://openrouter.ai/anthropic

In Q3, I think they struggled with the government ban on Mythos/Fable, Opus 5 not being so good, OpenAI models being more token efficient, and they decided to dedicate a higher percentage of compute to training.

Your posts here are being uploaded to the cloud and sold. Yet, you still write here. So people will still play Skyrim.

  if these models were available to mathematicians

$/token/intelligence is dropping exponentially. Soon, they will have access.

Why do I get this feeling that people who complain about these models not being accessible are the same ones who are anti-data center build out, thinks AI is in a bubble, and thinks AI labs are evil.

In other words, they want cheap, available access to highly intelligent models, but they believe everything about AI industry is wrong.

  Yeah you can dream, but only what we allow.

This was true before Claude by the way.

You still need humans to tell the AI what is important.

  Yet the creatives don't get into gamedev if they didn't have to spend any effort to make their art

The art will just change. The bar will be raised. What is considered as good art in 2021 is different in 2026.

In math, the speed and innovation will increase. The bar will be raised. Those mathematicians who do not want to be part of this new world will quit and go do something else, just like how some software devs have done so since AI agents. But not all will quit.

i think the author and most analysts miss one important bottleneck.

If you look at chip wafer volume from TSMC, they've increased their capacity by a "whopping" 20% since 2020 to 2025. Yes, that's right. All this talk about overcapacity and TSMC has only managed to increase total chip volume by 20% in 5 years.

You'd think that with all this talk of overcapacity, we've increased chip volume shipped by 10x. No. It's just 1.2x.

We're capped ASML and TSMC. No amount of math finesse can get around it. It's bottlenecked at the very top.

Even if Silicon Valley wants 10x number of chips, they can't. This buildout is physically capped by two behemoths.

  The code was never the hard part.

It definitely was. It was at least 50% of the hard part.

I remember there was a debate in my company on execution vs idea. Execution being writing code, the process that comes with it, and the idea being the understanding, market fit. Both sides had good arguments.

Hiring a software team to build an idea was extremely expensive and risky before ChatGPT. That's why most businesses in the world did not have a software engineering team.

Because AI cost fears here likely refer to companies responsible for AI buildout and use.

PCs dropping in volume can signal better AI demand.

There is absolutely no doubt in my mind that this is just a hit piece - IE, they were looking to paint Nvidia and Jensen in some bad light and found a vague lead. Didn't matter if the evidence isn't substantial. Just publish and smear first.

More like ChatGPT and AI is disrupting the search market.

Advertisers would have pulled out and Google would have lost revenue instead of increasing.

I'm not a believer in Google search long term, but advertisers continue to pay because they still get value - even if Google increases prices.

Looks like McCarthyism is back on the menu. Big media is out in force to push a narrative.