High Signal Podcasts Evidence ledger
Method
Browse
← Back to evidence

Evidence receipt / commitment

Published · transcript-backed

Speaker unverified: commitment

3 Apr 2025 Latent Space The Creators of Model Context Protocol

“I think we will not be prescriptive in that sense. I think there will be inherently, you know, there's a lot of power.”

— Speaker unverified

Source trail

Everything needed to verify it.

Speaker
Speaker unverified
Attribution
Not verified from this transcript
Claim type
commitment
Recorded
3 Apr 2025
Publisher
Latent Space

Transcript context

…Is there going to be like a MCP standard library, so to speak, of like, hey, these are like the canonical servers, do not build this. We're just going to take care of those. And those can be maybe the building blocks that people can compose. Or do you expect people to just rebuild their own MCP servers for like a lot of things? I think we will not be prescriptive in that sense. I think there will be inherently, you know, there's a lot of power. Well, let me rephrase it. Like, I have a long history in open source, and I feel the bizarre approach to this problem is somewhat useful, right? And I think so that the best and most interesting option wins. And I don't think we want to be very prescriptive. I will definitely foresee, and this already exists, that there will be like 25 GitHub servers and like 25, you know, Postgres servers and whatnot. And that's all cool. And that's good. And I think they all add in their own way. But effectively, eventually, over months or years, the ecosystem will converge to like a set of very widely used ones who basically, I don't know if you call it winning, but like that will be the most used ones. And I think that's completely fine. Because being prescriptive about this, I don't think it's any useful, any use. I do think, of course, that there will be like MCP servers, and you see them already that are driven by companies for their products. And, you know, they will inherently be probably the canonical implementation. Like if you want to work with Cloudflow workers and use an MCP server for that, you'll probably want to use the one developed by Cloudflare. Yeah. I think there's maybe a related thing here, too, just about like one big thing worth thinking about. We don't have any like solutions completely ready to go. It's this question of like trust or like, you know, vetting is maybe a better word. Like, how do you determine which MCP servers are like the kind of good and safe ones to use? Regardless of if there are any implementations of GitHub MCP servers, that could be totally fine. But you want to make sure that you're not using ones that are really like sus, right? And so trying to think about like how to kind of endow reputation or like, you know, if hypothetically. Anthropic is like, we've vetted this. It meets our criteria for secure coding or something. How can that be reflected in kind of this open model where everyone in the ecosystem can benefit? Don't really know the answer yet, but that's very much top of mind. But I think that's like a great design choice of MCPs, which is like language agnostic. Like already, and there's not, to my knowledge, an Anthropic official Ruby SDK, nor an OpenAI SDK. And Alex Roudal does a great job building those. But now with MCPs is like. You don't actually have to translate an SDK to all these languages. You just do one, one interface and kind of bless that interface as, as Anthropic. So yeah, that was, that was nice.…

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

Search evidence