This is untrue simply based on the so many past instances of Gemini, OpenAI making their products cheaper. The ratelimits for GPT 5 are pretty high. The API costs have decreased by 50% over and above o3's reduction which was also massive.

This is not even considering the fact that the performance has also increased.

That's what the post means. OpenAI doesn't lose money on each request but rather gains it. To recuperate the fixed costs in R&D.

I thought Uber ridership has been increasing worldwide, if you have data that contradicts this please share it.

Because most of the other entrants are going bust. Saw Revel is the latest to bite the dust and exit ridesharing.

Cool but that just means Uber is hard to beat. I think Uber is doing just fine.

Almost no business works like this - every additional request does not make OpenAI lose money but rather gain it.

The fixed cause due to R&D is what makes it unprofitable but not each request. Your line of thinking is bit ridiculous because OpenAI is never going to lose money per request.

> The fixed cause due to R&D is what makes it unprofitable but not each request. Your line of thinking is bit ridiculous because OpenAI is never going to lose money per request.

We don't know this for sure. I agree that it would be insane from a business perspective, but I've seen so many SV startups make insane business decisions that I tend to lean towards this being quite possible.

Nope, it costs billions to train and run those models, they are operating at a loss.

You are describing a subsidy.

If your kid makes $50 with a lemonade stand, she thinks she made $50, because she doesn't account for the cost of the lemonade, table, lawn, etc. You're subsidizing your child.

I agree its subsidised but crucial point being that each API doesn't cost them but gives them profit. If R&D were to be stopped now they would be profitable.

No it gives them income. Profit is when all costs are subtracted.

Correct, I misspoke.

What does this mean? "Rideshare became more of a commodity"

>And now, everyone is working on the Starship. While its cool, its absolutely impractical. It would take 8 -20 launches to refuel it in orbit for an actual mission, and the engines are basically on the verge of blowing up when operating due to how precise the system has to be for that high of a pressure ratio.

When he does achieve it I imagine you will create a new post saying something like "he didn't achieved Starship because of the smart people working on it, but because he had the money to throw at test over and over again".

Mostly the priorities are not competing. You can and should build more homes but the end goal is to improve utility of land which can further drive prices up thus helping both cases.

You can first build more homes - prices go down and more people come into the city and prices start going back up because of increased economic activity.

> You can and should build more homes

Yes. The entry point to ownership lies within a state of affordability.

> but the end goal is to improve utility of land

For who? If it's anyone other than typical wage earners, this goal competes with home ownership by pressuring ownership costs upward and away from affordability.

For context, home purchases are mostly (if not fully) impossible for typical wage earners.

For the city or country. Home ownership is not impossible, I know you meant it as hyperbole.

We shouldn’t mind if home ownership becomes costly if more economic activity results out of it. More economic activity means more money in the hands of wage earners (just that proportionally more money is spent on housing which is okay).

> Home ownership is not impossible, I know you meant it as hyperbole.

I did not. 25% of US households make under $40/yr.

(based on 2022 data that 25% of all US households make ~$35k/yr or less. +10% for minorities.)

Because they can get so dominant as to alienate normal users. Like in Reddit.

But I agree with your larger point and I think it is a valid point.

I wrote about why higher level abstractions may not work all the time.

https://simianwords.bearblog.dev/so-you-want-to-create-a-new...

TL;DR the overhead in learning the new abstraction and dealing with the inevitable edge cases (known or unknown) must be taken into account.

All abstractions sound good until you account for these things.

The new language can work if the edge cases of this language are known and minimal.

An idea worth exploring: if specialized models on datasets can be trained quickly, it can be used as tools by bigger models.