More agents are useful for speed: they can solve many problems faster. They're less useful for boosting AI intelligence. Multi-agent gains in solving harder problems often diminish incredibly quickly and may even saturate.
The reason they believe robots haven't generalized like LLMs isn't that the models aren't smart enough, but that the data has been a fraction of a percent of what humans naturally generate every day, captured through interfaces that distort the very behavior they're trying to record.
Amazon, more than any other company, actually operates with decade-long timeframes, consistently making real-world investments at massive scale that (1) convert their marginal costs into capital costs and (2) gain leverage on those capital costs by selling them to other businesses.
It's opportunity costs, not marginal costs, that are the challenge facing hyperscalers. How much compute should go to customers, and which ones? How much should be reserved for internal workloads?
Kalanick spent nearly eight years in what might be the most extreme version of stealth mode any modern founder has pulled off: he's hired thousands of employees, in 30 countries, bought and developed hard real estate assets, built a full-stack food infrastructure business, and did it all like lasagna.
"We have the aspiration to enable the global flow of compute, and we think that's going to be a multi-trillion-dollar market," Moushey says. "And to do that, we have to be a little bit more clear that we're not this weird compute provider anymore. We're actually building the market infrastructure to solve this dislocation of supply and demand."
There is an argument that without the organizational skills and production strategy Forrest brought to the table, Fender may have not been able to make the jump from a comparatively small builder to the mammoth industry leader it became.
7mo ago
Underscored — save the words that stop you in your tracks.
