Lex Fridman: belief

Transcript context

…I think if you talk to people at the labs, they use these in their training and production code. Claude Code is built with Claude Code, and they all use these things extensively. And Dario talks about how much of Claude’s code… It’s like these people are slightly ahead in terms of the capabilities— —they have, and they probably spend on inference. They could spend 10 to 100 times as much as we’re spending, like we’re on a lowly 100 or $200 a month plan. They truly let it rip. And I think that with the pace of progress that we have, a year ago we didn’t have Claude Code and we didn’t really have reasoning models. The difference between sitting here today and what we can do with these models—it seems like there’s a lot of low-hanging fruit to improve them. The failure modes are pretty dumb. It’s like, “Claude, you tried to use the CLI command I don’t have installed 14 times, and then I sent you the command to run.” From a modeling perspective, that thing is pretty fixable. So, I don’t know. I agree with you. I’ve been becoming more and more bullish in general. Speaking to what you’re articulating, I think it is a human skill issue. So Anthropic and other companies are leading the way in understanding how to best use the models for programming; therefore, they’re effectively using them. I think there’s a lot of programmers on the outskirts who don’t… I mean, there’s not a really good guide on how to use them. People are trying to figure it out exactly, but— It might be very expensive. It might be that the entry point for that is $2,000 a month, which is only for tech companies and rich people. That could be it.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.