rfw300

I don't think he got it backwards, at least if Hegseth's statement is accurate. AWS, GCP, etc. all do business with DoD. If they, as DoD contractors, are no longer allowed to do business with Anthropic, then presumably they have to stop re-selling or hosting Anthropic's models to anyone.

More generally, Anthropic's reliability track record for a company which claims to have solved coding is astonishingly poor. Just look at their status page - https://status.claude.com/ - multiple severe incidents, every day. And that's to say nothing of the constant stream of bugs for simple behavior in the desktop app, Claude Code, their various IDE integrations, the tools they offer in the API, and so on.

Their models are so good that they make dealing with the rest all worth it. But if I were a non-research engineer at Anthropic, I wouldn't strut around gloating. I'd hide my head in a paper bag.

I have little doubt where things are going, but the irony of the way they communicate versus the quality of their actual product is palpable.

Claude Code (the product, not the underlying model) has been one of the buggiest, least polished products I have ever used. And it's not exactly rocket science to begin with. Maybe they should try writing slightly less than 100% of their code with AI?

It also strikes me as being in competition with, you know, a group chat.

Well if only we had such thoughtful minds in the education space. How did no one ever think of "why don't we just give up on the children"?

This strikes me as entirely logical in the short run, and an insane way of packaging software that we will certainly regret in the long run.

Relevant xkcd: https://xkcd.com/1958/

Interesting project, but the lack of any actual benchmark results on existing models/agents is disappointing.

This is true as a technical matter, but this isn't a technical blog post! It's a consumer review, and when companies ship consumer products, the people who use them can't be expected to understand failure modes that are not clearly communicated to them. If OpenAI wants regular people to dump their data into ChatGPT for Health, the onus is on them to make it reliable.

The example you present seems fairly straightforward to my intuition, but I think your point is fair.

A harder set of hypotheticals might arise if music production goes the direction that software engineering is heading: “agentic work”, whereby a person is very much involved in the creation of a work, but more by directing an AI agent than by orchestrating a set of non-AI tools.

Any person who would choose 3.7 with a fancy harness has a very poor memory about how dramatically the model capabilities have improved between then and now.

"Captcha" doesn't refer to any specific type of puzzle, but a class of methods for verifying human users. Some older-style captchas are broken, but some newer ones are not.

How much compute do your systems expend on chunking vs. the embedding itself?

That may be true, but it does seem like OP's intent was to learn something about how LLM agents perform on complex engineering tasks, rather than learning about ASCII creation logic. A different but perhaps still worthy experiment.

That might be true for a narrow definition of chatbots, but they aren't going to survive on name recognition if their models are inferior in the medium term. Right now, "agents" are only really useful for coding, but when they start to be adopted for more mainstream tasks, people will migrate to the tools that actually work first.

Claude Code's Plan Mode increasingly does a (small-scale) version of this - it will research your codebase and come back to you with a set of clarifying questions and design decisions before presenting its implementation plan.

Your comment assumes that the purpose of an economic system is to preserve a financial status quo. It isn't, and shouldn't be. Inflation incentivizes people to put their money to productive uses in the economy (capital formation) rather than hoarding resources.

I'd like others' input on this: increasingly, I see Cursor, Jetbrains, etc. moving towards a model of having you manage many agents working on different tasks simultaneously. But in real, production codebases, I've found that even a single agent is faster at generating code than I am at evaluating its fitness and providing design guidance. Adding more agents working on different things would not speed anything up. But perhaps I am just much slower or a poorer multi-tasker than most. Do others find these features more useful?

I am extremely excited to use programmatic tool use. This has, to date, been the most frustrating aspect of MCP-style tools for me: if some analysis requires the LLM to first fetch data and then write code to analyze it, the LLM is forced to manually copy a representation of the data into its interpreter.

Programmatic tool use feels like the way it always should have worked, and where agents seem to be going more broadly: acting within sandboxed VMs with a mix of custom code and programmatic interfaces to external services. This is a clear improvement over the LangChain-style Rupe Goldberg machines that we dealt with last year.

This isn't true. Gemini is trained and run almost entirely on TPUs. Anthropic also uses TPUs for inference, see, e.g., https://www.anthropic.com/news/expanding-our-use-of-google-c... and https://www.anthropic.com/engineering/a-postmortem-of-three-.... OpenAI also uses TPUs for inference at least in some measure: https://x.com/amir/status/1938692182787137738?t=9QNb0hfaQShW....

My audio experiment was much less successful — I uploaded a 90-minute podcast episode and asked it to produce a labeled transcript. Gemini 3:

- Hallucinated at least three quotes (that I checked) resembling nothing said by any of the hosts

- Produced timestamps that were almost entirely wrong. Language quoted from the end of the episode, for instance, was timestamped 35 minutes into the episode, rather than 85 minutes.

- Almost all of what is transcribed is heavily paraphrased and abridged, in most cases without any indication.

Understandable that Gemini can't cope with such a long audio recording yet, but I would've hoped for a more graceful/less hallucinatory failure mode. And unfortunately, aligns with my impression of past Gemini models that they are impressively smart but fail in the most catastrophic ways.

What use case do you have (or anticipate having) for WiFi 7 out of curiosity?

Indeed. OP, nothing is "in" an LLM's context window at rest. The old version of your file is just cached in whatever file stores your IDE's chat logs, and this is an expensive way of retrieving what's already on your computer.

I wonder if some of the non-robustness of the tradfi system is a feature, not a bug. If my account tries to send someone $3 million, I'd prefer that it's intermediated by a confused bank employee staring at a screen rather than a beautifully efficient, irreversible machine consensus. The bottlenecks and intermediaries create friction, sure, but that isn't per se bad.

My hang-up with crypto is that it solves the ledger-keeping part of running a financial system, but it isn't clear that's actually the hard part! Preventing and remediating fraud, money laundering, etc. are, and crypto makes those issues worse, not better.

It has nothing to do with Meta's social media business. Zuckerberg, like many other top tech executives, has concluded AGI/ASI is in striking distance. If he could somehow win the race, he becomes god (or so he thinks, anyway). And from the perspective of a man whose idol is Julius Caesar, what wouldn't you spend for that chance?

Not necessarily true. GPT-4.1 was released after GPT-4.5-preview. Next model might be GPT-3.7.

It's interesting that you name those examples, because Lee Sedol, the all-time great Go player, retired shortly after losing to Alpha Go, saying: "Even if I become the number one, there is an entity that cannot be defeated... losing to AI, in a sense, meant my entire world was collapsing... I could no longer enjoy the game. So I retired." [1, 2]

So for some, yes. It is of course also true that many people derive self-worth and fulfillment from contributing positively to the world, and AI automating the productive work in which they specialize can undermine that.

[1] https://en.yna.co.kr/view/AEN20191127004800315

[2] https://www.nytimes.com/2024/07/10/world/asia/lee-saedol-go-...

o4-mini does look to be a better model, but this is actually a lot cheaper! It's ~7x cheaper for both input and output tokens.

The amount of blatant pandering OpenAI is doing to California politicians should make everyone suspicious. They are structuring this nonprofit transition as a bribe to the California politician-nonprofit complex in exchange for a much more valuable regulatory approval. Their announcement from a few days ago:

> built by the leading AI company, put in the hands of nonprofits throughout California, across America and beyond… This commission will also incorporate feedback from leaders and communities in health, science, education, and public services—particularly within OpenAI’s home state of California.

These advisors have no connection to AI whatsoever—what they have in common is deep political connections in the state of California. Dolores Huerta is a remarkable woman, but she is a 95 year-old union organizer. The rest of the list is similar.

Is there anyone in America who believes OpenAI is making this transition to better achieve its nonprofit’s mission? I would like to meet them.

It can be both properly communicated and still bad for API use cases.