Discussing how open source is changing in the LLM era, Kerr speculates that model makers could one day sell preference for particular tools in the architecture advice their models give developers.
…It's a space where LLMs have a preference for certain types of technology and patterns because the corpus of training data says you should use bun for this, you should use whatever it is for this. So I don't know, Adam. What do you think? I mean, you talk to so many people. How do you feel about
…your code out there in open source and you've got cybersecurity concerns now. Well, now because an LLM is so skilled, you can find the loopholes and the issues in. I mean, this is an absolute bloodbath of sorts in the NPM ecosystem, for example. I mean, like, it's just everywhere. So I'm not sure
…about when it comes to, say, capitalism, how long is it before someone who's running a project, commercial project that has an open source angle, potentially starts saying to the Frontier Labs, how much would it cost me for your LLM to prefer my solution over others when people are making architectural decisions? You know, and I'm not going to name any companies' names here because I get it. It would be a business model in the same way that Google will rank certain destinations higher when you do a search. That's how they monetize that. The LLMs have a lot of influence now into making architectural decisions, some of them very consequential architectural decisions. If you're working with a bank and some of your advice is coming from a model, that's potentially going to lead to a very consequential long-term relationship and decision.
Yeah. You may choose to build something versus buy it when you could have just bought it or bought the open source version of it. That's what I mean by build versus buy. Like use somebody else's thing. Literally buy it or get it for free via open source. You may find yourself, if you listen to only…
Yeah. And it is tested to absurd degrees now.
Absolutely.