alach11

News just broke today that Google is planning on doing this: https://news.ycombinator.com/item?id=48986351

the winner will be whoever burns their models to ASICs fastest

As of today, that appears to be Google!

https://news.ycombinator.com/item?id=48986351

If anyone can make this work, it's Google. They have all the right pieces: scale, chip design experience, and high-volume low-intelligence usage (their AI summaries for searches). This is sort of a bet against model capabilities accelerating, right?

With the administration's track record of reactive (if not capricious) actions, I doubt any leading AI Labs are going to flout this order, even if it is "voluntary". Nobody wants to be designated as a supply chain risk.

The honest answer to that question, in June 2026, is that we do not know

The honest reading of those numbers is not that defense is winning on economics

The honest 2026 answer is in three parts.

The honest answer is that we do not know, because no one has tried

Firstly, I appreciated the article and especially the visuals. But I had the same reaction as the GP commenter. It was hard to read. I'm sick of this punchy, repetitive, LLM-generated prose.

Everyone loves to say this when the death of Stack Overflow is discussed, but it always was that way. Strict moderation, love it or hate it, was part of the platform. And it could have kept going that way for many more years if not for LLMs 99.9% obviating the need for a coding Q&A forum.

It's a tall order to live up to the impact of Rerum novarum, the encyclical by the former Pope Leo that greatly guided thinking out of the industrial revolution. Personally, I'm excited to read this. If we take the claims of most AI labs at face value, they believe their work will fundamentally change the relationship between humans and the economy. More involvement from faith leaders is a good thing.

Isn't a large user base and the data collected from those users a moat of sorts?

If Anthropic actually cared about humans, they would have the best customer support (staffed by humans, for humans)

I know Anthropic support is slow from firsthand experience, but it has to be pretty difficult to scale support 10-80x per year. And even more so when you have a long-tail of very low revenue usage in the form of $20/month subscriptions.

Snake and DOOM were two of our early tests (for filter functions and MCP) when we stood up Open WebUI for internal chat/agent use. Sometimes games are the best way to limit-test new tech.

I ran an internal (oil and gas focused) benchmark yesterday and found Opus 4.7 was 50% cheaper than Opus 4.6, driven by significantly fewer output tokens for reasoning. It also scored 80% (vs. 60%).

On my private internal oil and gas benchmark, I found a counterintuitive result. Opus 4.7 scores 80%, outperforming Opus 4.6 (64%) and GPT-5.4 (76%). But it's the cheapest of the three models by 2x.

This is mainly driven by reduced reasoning token usage. It goes to show that "sticker price" per token is no longer adequate for comparing model cost.

A significant part of Anthropic's cachet as an employer is the ethical stance they profess to take. This is no doubt a tough spot to be in, but it's hard to see Dario making any other decision here.

What I don't understand is why Hegseth pushed the issue to an ultimatum like this. They say they're not trying to use Claude for domestic mass surveillance or autonomous weapons. If so, what does the Department of War have to gain from this fight?

I believe Nat Friedman said "pessimists sound smart, optimists make money." It's certainly much easier to give a snarky/negative take and shoot an idea down than think creatively about how to make it work. Also, negative people are perceived as smarter!

https://www.sciencedirect.com/science/article/abs/pii/002210...

I have to imagine governments are closely monitoring prediction markets as part of their intelligence apparatus. But then you just add another layer of subterfuge. Imagine a D-Day prediction market... "Will the Allies Land in Normandy, Pas-de-Calais, or somewhere else?" The US might buy a major position on Pas-de-Calais the night before as a decoy!

I really wish these models were available via AWS or Azure. I understand strategically that this might not make sense for Google, but at a non-software-focused F500 company it would sure make it a lot easier to use Gemini.

Usually the first day or two are readily solvable in Excel with just regular spreadsheet formulas.

This is the biggest news of the announcement. Prior Opus models were strong, but the cost was a big limiter of usage. This price point still makes it a "premium" option, but isn't prohibitive.

Also increasingly it's becoming important to look at token usage rather than just token cost. They say Opus 4.5 (with high reasoning) used 50% fewer tokens than Sonnet 4.5. So you get a higher score on SWE-bench verified, you pay more per token, but you use fewer tokens and overall pay less!

Just curious - are you using Open WebUI or Librechat as a local frontend or are all your workflows just calling the models directly without UI?

This is a really impressive release. It's probably the biggest lead we've seen from a model since the release of GPT-4. Seems likely that OpenAI rushed out GPT-5.1 to beat the Gemini 3 release, knowing that their model would underperform it.

"Quantity has a quality all its own". It's categorically different to be able to do harm cheaply at scale vs. doing it at great cost/effort.

Can you cite your source for inference being at a loss? This disagrees with most of what I've read.

This is almost certainly the issue. It's very unintuitive for users, but LLMs behave much better when you clear the context often. I run /clear every third message or so with Claude Code to avoid context rot. Anthropic describes this a bit with their best practices guide [0].

[0] https://www.anthropic.com/engineering/claude-code-best-pract...

Computer use is the most important AI benchmark to watch if you're trying to forecast labor-market impact. You're right, there are much more effective ways for ML/AI systems to accomplish tasks on the computer. But they all have to be hand-crafted for each task. Solving the general case is more scalable.

I'm really interested in the progress on computer use. These are the benchmarks to watch if you want to forecast economic disruption, IMO. Mastery of computer use takes us out of the paradigm of task-specific integrations with AI to a more generic interface that's way more scalable.

Thus far I didn't have to worry about ChatGPT having bad incentives when giving me advice on product purchases. Now that "Merchants pay a small fee on completed purchases", will the model steer me towards ACP-supported retailers at a higher rate?

using AI is clearly doing more harm than good

How do you know this? Wouldn't we expect the benefits of AI in the legal industry to be way less likely to make the front page of HN?

I can't fathom why Microsoft has such limited functionality in Copilot for Excel. Some of the ideas in projects like this (and similar) seem so obvious. Are they just being conservative with token cost?

I'm going to make an unpopular suggestion. Have you considered using a service that will print and ship to you, like CraftCloud?

Depending on volume, your total cost would likely be lower. I know you mentioned privacy concerns so this may not be an option. But it significantly simplifies your work, letting you focus on the parts themselves.

Florida did drug testing as a condition for welfare benefits... and it cost more than they saved

It's more complicated than that. Of the 6352 people who applied for TANF, 2306 dropped out during the process. Then of the 4046 TANF applicants remaining, only 2.6% tested positive for drugs. The vast majority of media coverage focused on the 2.6% being less than the ~8% drug-use rate in the general population.

What we don't know is of the people who dropped out, was this due to unintended reasons (privacy concerns, the inconvenience of the drug test, missing deadlines) or due to the intended reason (people self-selecting out because they knew they would test positive and become ineligible for 12 months). We'll never know the real breakdown, but it's misleading to say "it cost more than they saved".