Their argument doesn't show that at all though, they make an argument against an age bias when the article explicitly says age is being controlled for.

> Figure 1 suggests that we'd probably expect to see an outlier or two in that mess just through chance, we're expecting a few p<0.01.

That's a valid point, and a much better criticism than the one I'm calling out. That doesn't redeem the original comment though.

Not at all, it's more like 6 hours if you're in a somewhat remote place, but most taxi trips are way smaller than that even in rural area anyway.

> Not to mention that the population in this usa countryside is already way lower then any population density in the whole of France.

I don't think it matters, as there are plenty of rural places in the US that still have a higher density than where I live.

> Operating a taxi corp there would simply be not viable on a financial term.

Probably not a taxi “corp”, but here taxis have always been independent businesses with one or two cabs an providing taxi service on an area of a few hundred of square kilometers (a radius of 50km or so around the "office" of the taxi driver). And a large fraction of the trips are not on demand, but upon reservation several days in advance (you know you'll need to go to a medical appointment to the city next month, or you need a ride to the train station later this week).

What? In my country at least (France), taxis have been a thing even in the countryside for as long as cars existed (or rather, as long as the telephone has). And I'd be very surprised if that wasn't also the case in most of Western Europe.

In fact it's a much easier job when it comes to routing and knowing your passengers' destinations than doing it in a metropolis.

Prioritizing growth? When OpenAI had saturated the market by year one and has only been losing market share for the past 18 months! Again that's basically assuming Sam Altman is a complex idiot. No board in existence will let you move away from the most obvious business model in the Silicon Valley if this model was the obvious solution you think it is.

> They show that "LLM inference which is significantly costlier than web search." is unsupported

It's not unsupported, the price per million token of frontier model is public, as well as the cost of running open source models, which gives us a rough idea of how much those frontier models cost to run, and it's significantly higher than what it costs to run a web search.

> Imagine dismissing numbers

But you're just making those “numbers” up: there's simply no cost numbers at all in these papers. They are just about energy, and you jump from there forgetting that consuming a kWh of electricity on a server from 19 years ago is substantially cheaper in capex than doing so on a H100, let alone doing so on newer hardware.

This is obviously harmful for society but also obviously the best outcome for the platform who wants to sell your attention to advertisers.

> Take away the huge R&D and they become hugely profitable with a robust ad business like Google Search.

Is refuted by the sentence that comes immediately before :

> They have a billion weekly active users, almost all free (Not Google search free. Free free) with about ~50M subscriptions. Of course they're not profitable even without R&D. Low cost isn't no cost.

Please explain why have OpenAI been actively not trying to “hugely profitable with a robust ads business” for the past 3 and a half year then? Or at least make a profit that can cover some of there R&D expenses, instead of losing money from their operations?

> Just admit you're wrong about inference costs and move on.

Unfortunately I'm not. And again, none of your links above says otherwise.

That's the biggest myth of economics. It wasn't even true the pre-modern world except occasionally (famines were a real thing, but they were still kinda infrequent), and it's pretty much never true in an industrial civilization, let alone in an era of growing automation capabilities.

> Research and Development: $19.18 billion. Loss from Operations: $20.92 billion

OpenAI isn't profitable even if you discount R&D entirely. Google search on the other hand has had ridiculous margin from the beginning, allowing the company to be highly profitable while financing significant R&D investment. These two businesses models are as different as it can be.

If ads were the solutions for profitability these companies would have used ads as their source of income from the very beginning.

I'm not talking about R&D.

> there are a class of tokenmaxxing users well outside the median (e.g Agentic Coding).

It wasn't a thing before last year and OpenAI wasn't making profit before that either.

> https://cloud.google.com/blog/products/infrastructure/measur...

> https://epoch.ai/gradient-updates/how-much-energy-does-chatg...

None of these support your claim.

Again, when you're here saying that the CEOs of the two biggest AI companies are basically idiots, nobody will take you seriously.

I too can have $50B revenues by selling dollars for 50 cents each, and in the process I'll make a smaller loss than they do.

Whatever you think about Altman and Amodei, they aren't clueless idiots and if ads was all they needed they would have done that from the beginning.

This kind of takes makes me cringe. Why don't you go back to LinkedIn?

I have no idea what you mean by “pair program with an agent”, but Opus have been able of autonomous coding since last November, and with any half-decent harness even local Qwen3.5 was able to do so 6 months ago.

Fable is a stronger model, which means it can solve harder tasks but it's also over-hyped, because only a small fraction of task is hard enough to be Fable-worthy.

Fable has only been out for a month but somehow everyone is supposed to have moved to a completely different way of working that supposedly only works for Fable and nothing else…

It's funny to see that Anthopic shills have been saying the exact same thing for the past two years now (and it was OpenAI fans before). It's amazing to see that Claude 3 Sonnet was "great" but now that even Qwen 9B is better than this version of Sonnet DeepSeek V4 is still not good enough despite being stronger than Opus 4.7 was.

> Let’s say you could buy an LLM that gets things right 98% of the time. But there’s another LLM that’s 100x the price but gets things right 99.9% of the time

If you think Fable makes 20 times fewer mistakes than DS4 you're delusional. It doesn't even do 20 fewer mistake than Gemma 4…

They are definitely smart enough to be useful, but dumb enough in their weak spots not to deserve the "general intelligence" qualifier.