Sam Altman: "50¢ of compute of a SWE Agent can yield "$500 or $5k of work."

Then perhaps they should have this SWE Agent make some non-trivial contributions to the Linux kernel or other large, impactful open source projects? It would be a great way to demonstrate that they're not full of shit and would help projects that will still be useful in the near future.

Unless, of course, there's some reason why their SWE Agent can't accomplish this.

This is a fantastic idea.

Perhaps then this should be the next test for these SWE Agents, in the same manner as the 'Will Smith Eats Spaghetti" video tests.

If this were remotely true they would NOT be offering a subscription. They would start a consulting company that does custom software development and dominate the market

“Can” but how often will it? And how costly are the mistakes it makes, especially if they’re not detected by an experienced (expensive) dev?

https://www.theatlantic.com/technology/archive/2024/10/sam-a...