JacobAsmuth

Conspiracy theorist discovers seasonality, 2026. Colorized.

Any news on grannybench?

You'd prefer public questions and answer sheets that anyone can train on?

You can check out the token usage on AA. It used ~30% fewer tokens than Opus.

What are you confused about? This model is the same price and speed and token usage per task of Argon but scores better in ~every test.

I'm mildly optimistic about Anthropic: positive free cash flow, growing margins, and incredible revenue growth are heartening. Its clear now that their delayed IPO was due to continually improving financial metrics and the desire to balance the time value of an earlier IPO with the gains from presenting a financially healthier company which is growing quickly and spending heavily to do so.

OpenAI seems to be in a much tougher spot and I'm not sure how they'll fare in the future.

But honestly, neither of those really matter to me and its why I didn't mention them in my above comment. Anthropic + OpenAI revenue is ~$120B/yr, and even if those companies go to zero because open models commoditize their business model, that demand for inference isn't going to go away, someone will need to absorb it. In fact I would expect that an open model good enough to drive demand for Opus 5.5 all the way down to zero would be so good that inference on that model would drastically exceed what we're seeing. I'm more concerned with one layer up, the datacenter builders and the semiconductor companies, and there I see fortress balance sheets, great governance, and massive demand. Open Models are great for these companies, not a threat at all. I don't see any credible threat to AWS+Azure+GCP datacenter buildout not paying for itself.

Thank you!

Any news?

The S1 was filed yesterday. https://www.sec.gov/Archives/edgar/data/2130357/000162828026...

Operating cash flow of $3,929M and $482M of CapEx gives $3,447M of Free Cash Flow in H1 2026

1. One of Warren Buffet's final acts before stepping down was buying billions of dollars of Google stock. Of course he understands fundamental analysis, that's why he's buying in to one of the companies that have solid financials.

2. Burry has repeatedly made claims about semiconductors and semiconductor manufacturers that even the most rabid anti-ai industry expert would tell him were false. He also has made dozens if not hundreds of predictions of impending doom that simply did not come true and in fact had to shutter his fund last year due to massive losses.

People are calling it a bubble because they dont understand the financials.

They state their operating cash flow and their capex. FCF is defined as OCF-Capex

https://www.sec.gov/Archives/edgar/data/2130357/000162828026...

Operating cash flow of $3,929M and $482M of CapEx gives $3,447M of Free Cash Flow in H1 2026

https://www.sec.gov/Archives/edgar/data/2130357/000162828026...

Operating cash flow of $3,929M and $482M of CapEx gives $3,447M of Free Cash Flow in H1 2026

The S1 that was just filed.

https://www.sec.gov/Archives/edgar/data/2130357/000162828026...

Operating cash flow of $3,929M and $482M of CapEx gives $3,447M of Free Cash Flow in H1 2026

Yup. gorgetbench, grannybench, and tebufranksalso bench are all completely unsaturated.

Okay so you agree that big companies can keep funding this buildout, NVidia (and everyone in their supply chain) is backlogged 18+ months, and there's essentially zero risk of insolvency from any of the major customers, all of whom are committed and able to increase their purchasing (pursuant to their contractual obligations) of AI hardware over the next 3 years?

No one serious cares if there's a bubble in the stock market because the financials are solid and when those play out then whatever "bubble" you imagine exists today will be small compared to realized revenues in a few years.

Well then the fact that those hosting partners have 10+ year commitments should make you pretty confident then, right?

It's sad how far down you have to scroll to get the real answer. It's because Anthropic margins on inference are absolutely bonkers and it's easy for them to drop prices by 80% and STILL have positive free cash flow

AKA worldcoin

See these kinds of comments bother me because it seems like you haven't actually looked at the data. It reads "very informed" but actually is essentially contentless. It's the same reason I didn't like reading AI output before tool-calling/web search became a thing.

Not a chance https://news.ycombinator.com/user?id=bazzargh replies to this

Thanks

Is 4 correct or incorrect? What data do we have so far in this regard?

Can you give an example of a search query with a terrible AI Overview? Or even better, a class of search queries that produce terrible AI Overviews? Would really appreciate it!

What's pillowfort?

Yeah they read the news this morning and whipped up a quick Torch->XLA translation layer and coordinated with four different inference engines (who also all wrote their own blog posts) and shoved this out the door quick to stay on top of the news cycle

Products can also be a moat.

The CCP may be forcing them to do this.

Well except that baseten serves deepseek faster than this already at a lower price. So this seems to mostly help startup inference providers that don't have their own inference engines.