Ask HN: if the models are so good, why doesn't OpenAI start 100 companies?
The question: if AI can run a business, selling tokens looks like a strange strategy for the labs. The thread's best answers disagreed about whether the models are not good enough, or the labs are too short on cash.
An Ask HN thread of 665 comments put the question to OpenAI. One answer: the labs need revenue now, and selling tokens through long-running agents is the only thing that is provably growing.
A rival answer: the models are not good enough yet, and a fully automated business at scale would be a landmark nobody has delivered.
Cash first, then capability
One answer came as a numbered list from a commenter who argued that labs are already launching products, that selling tokens via long-running agents is "the only thing that's provably working with respect to revenue growth," and that "They need revenue yesterday." Higher borrowing rates mean capital is not what it was in the 2010s, so their runway is shorter than outsiders assume.
Another commenter backed the point from the other side: OpenAI has not yet amortised its current investments and starting 1,000 businesses means more debt and uncertainty.
"The real reason is the models just aren't good enough"
One commenter argued that the G in AGI implies not being specialised, so if the labs really had it, they would already be running businesses at scale. A fully automated successful business "would be a huge milestone and a legendary PR achievement, but the tech just isn't there yet." Another commenter turned the same logic on the labs' own pitch: if agents were largely autonomous and replaced workers, labor would not be the constraint.
The McKinsey comparison and its critics
A prominent argument started with an analogy: McKinsey knows how most Fortune 500 companies run and still has not taken them over, because "AI is brains. Companies need arms and legs and relationships." Critics took it apart. One commenter said the real reason is that "reality has a surprising amount of detail" and a gap exists between a slide and generating value in the real world. Another said McKinsey sells scapegoats and insurance, not strategy, so the analogy fails.
The shovel economics
A short exchange got at the incentive. One commenter said that in a gold rush the best strategy is to sell shovels, since the lab gets paid whether or not your idea succeeds. The founder of Bugsink replied that selling shovels at a loss would not be the best strategy unless there is some shovel lock-in effect. That is the unanswered point: token prices that do not cover costs only make sense if the labs expect lock-in.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
A Swedish AI host says the same request emits 8.5 times more CO2 in Texas than in Sweden
Berget AI counts about 0.4 grams of CO2e for a typical request in Sweden and 3.2 grams in Texas, on identical hardware. In Sweden, 89% of that is the chips' manufacturing.