Part of the sauce is not in the model, but in the agent itself. And for that matter, I think AMP an incredibly better agent that Claude Code. But then, Claude heavily subsidized subscription prices are hard to beat.

But then you pay for the less outrageously subsidized rates of API instead of the a bit less incredibly generous prices of the subscription.

Its not subsidized, in fact, they probably have very healthy margins on Claude Code.

Yeah. If you ignore the negligible fact that some investor may want a return on all that money that is going into capex I am pretty sure you can, Enron style, get to the conclusion that any of those companies have “healthy” margins.

Why do you think that?

DeepSeek had a theoretical profit margin of 545 % [1] with much inferior GPUs at 1/60th the API price.

Anthropic's Opus 4.6 is a bit bigger, but they'd have to be insanely incompetent to not make a profit on inference.

[1] https://github.com/deepseek-ai/open-infra-index/blob/main/20...

American labs trained in a different way than the Chinese labs. They might be making profit on inference but they are burning money otherwise.

Deepseek lies about costs systematically. This is just another fantasy.

What do you base your accusations on? Is there a specific number from the link above that you claim is a lie?

And how are 7 other providers able to offer DeepSeek API access at roughly the same price as DeepSeek?

https://openrouter.ai/deepseek/deepseek-v3.2

> they'd have to be insanely incompetent to not make a profit on inference.

Are you aware of how many years Amazon didn’t turn a profit?

Not agreeing with the tactic - just…are you aware of it?

Did Amazon really not turn a profit, or apply a bunch of tricks to make it appear like they didn't in order to avoid taxes? Given their history, I'd assume the later: https://en.wikipedia.org/wiki/Amazon_tax_avoidance

Anyway, this has nothing to do with whether inference is profitable.

It has everything to do with whether they make profit on paper,

vs give away the farm via free tier accounts, free trials,

and last but not least: subsidized compute to hook entire organizations………

Amazon was founded in 1994, went public in 1997 and became profitable in 2001. So Anthropic is two years behind with the IPO but who knows, maybe they'll be profitable by 2028? OpenAI is even more behind schedule.

How much loss did they accumulate until 2001? Pretty sure it wasn't the 44 billion OpenAI has. And Amazon didn't have many direct competitors offering the same services.

Because if you don't then current valuations are a bublle propped inflated by burning a mountain of cash.

That's not how valuations work. A company's valuation is typically based on an NPV (net present value) calculation, which is a power series of its time-discounted future cash flows. Depending on the company's strategy, it's often rational for it to not be profitable for quite a long while, as long as it can give investors the expectation of significant profitability down the line.

Having said that, I do think that there is an investment bubble in AI, but am just arguing that you're not looking at the right signal.

And that's OpenAI's biz model? :)

This was really useful; sometimes, by a glance, you'd see Claude looking at the wrong files or searching the wrong patterns, and would be able to immediately interrupt it. For those of us who like to be deeply involved in what Claude is doing, those updates were terribly disappointing.

We already do similar stuff for oil, uranium, and cobalt.

And this is something that would definitely net a good amount of money to municipal coffers in fines while educating drivers.

Traffic fines go to the central government, not local areas or police forces. On the one hand it takes away incentives to game the system (e.g designing suddenly speed limit reductions on otherwise fast roads), but it also means that enforcement is lacking as it becomes a cost for local governments and police forces

I don't know what these people from our now traditional daily lamentation session are coding where Claude can do all the work for them just with a few prompts and minimal reviews.

Claude is a godsend to me, but fuck, it is sometimes dumb as door, loves to create regressions, is a fucking terrible designer. Small, tiny changes? Those are actually the worse, it is easy for claude, on the first setback, decides to burn the whole world and start from zero again. Not to mention when it gets stuck in an eternal loop where it increasingly degenerates the code.

If I care about what I deliver, I have to actively participate in coding.

Kind of funny that some of those wait-6-month people are basically the same ones behind "no human driven cars being sold after 2025" and "computer vision is all you need"

Tensors are pretty old in physics; they are a central concept in Einstein's General Relativity.

You can find tensors even in some niche stuff in macroeconomics.

Yeah, Vibe code more github!

So far it feels they are vibe coding it day and night lol…probably with GitHub Copilot

> This is not a car for tech bros with no culture, no traditions, and no past.

Weird, because that's exactly what it looks like.