Making tech literacy irrelevant | Infactory’s Ken Kocienda

**Andrew Zigler** (0:06)
Welcome to Dev Interrupted. I'm your host, Andrew Zigler.

**Ben Lloyd Pearson** (0:10)
And I'm your host, Ben Lloyd Pearson.

**Andrew Zigler** (0:12)
So Ben, do you realize how hard I had to resist fangirling about this week's guest on the podcast? You know, you actually probably-

**Ben Lloyd Pearson** (0:21)
I did, because you've been talking about him nonstop, so yes.

**Andrew Zigler** (0:25)
It's really amazing to meet your heroes. It's one of the coolest people that I've met all year, and someone that I really related with. But I was most excited about the fact that when I met Ken Kocienda at the most recent ELC annual conference and got to sit down with him, he also brought a copy of his Creative Selection book that was signed that he gave me as part of the interview. So that's why I have not been able to stop fangirling, but I'm really excited to sit down with Ken Kocienda later in this episode. He's the co-founder of Infactory. He was a principal engineer at Apple, and he invented AutoCorrect. Yes, that ducking AutoCorrect.
Talk about impact on the amount of products and exposure that you've made. We're digging in on how Ken has been coding with AI. It's a really fun chat, so be sure to stick around. But first, there's some news that we want to dive into for the week, as well as some insights we learned on site from the folks at ELC Annual. So you want to dig into those, Ben?

**Ben Lloyd Pearson** (1:22)
Yeah, well, first of all, it's really great when you meet someone that you're a fan of and they live up to their expectations. So yeah, really looking forward to hearing today's interview.

**Andrew Zigler** (1:29)
Indeed.

**Ben Lloyd Pearson** (1:30)
Super awesome guess. But yeah, let's dive into the news. And first, I think it's a story we picked up from LinkedIn about MCP. So what do we have here, Andrew?

**Andrew Zigler** (1:38)
Yeah. So we found this fun post, MCP. In this post, I dove into the fact that MCP is maybe the first protocol of tech history where more people are building with it than there are actual end users. This is from Eduardo Ordax. And this is really great piece of insight that I have been rolling around since I read this post, which is why I wanted to share it here.
It's an idea that it's a protocol for builders, isn't to say it's a builder's problem. Are we treating everything like a nail and AI is the hammer, right? But I think it also draws attention to some of the usage statistics behind MCP that I found really interesting. Like, there's over 6,000 live servers, supposedly, on a lot of MCP server directories. But the top 10 servers hoard almost all of that attention. We're talking like 90% of the stars, most of the downloads, and it just comes down to a few core tools that allow your AI to interact with your computer, or the browser, or Git. Very predictable and repeatable things. But what I found most insightful about this is it shows how MCP is so critically an ingredient or a part of a solution or a problem, because these most popular MCP servers, they are tools that you use in combination with other things to build safer or faster. So that was an interesting tidbit. Ben, what did you think about this story?

**Ben Lloyd Pearson** (2:56)
In my opinion, what is probably at the center of this is that MCP is fundamentally a different type of product. Think about Perplexity, I think, is a great example of a company that launched in the AI era. When they launched, their website was literally just like a chat interface. If you wanted to find out what the heck this website was that you were on, you actually just asked the AI, like, what are you? It would basically do all of the marketing, all of the sales, like everything itself, because it had all of the context and the knowledge and everything that it needed to sort of interact with the user directly, rather than having to have like human workflows behind all of that stuff. And if you apply that then to all the things that we're seeing MCPs built for, I think it's just inherently like an extremely difficult product to just bring to the market.
You know, we helped Linear B launch our first MCP server. And so we've been having these discussions internally ourselves. And it really is just like, you know, it's a very powerful tool. Like when you do get it in a user's hands, like we've started to think of it as like an artifact factory. It's a way of generating things from data that just using simple prompts, you know. And it's kind of one of those things that gives you immense power. But with immense power, it can sometimes be hard to even know where to start. So I think it's just, it's a really challenge to package up in a way that is easy to consume. And I think until we get to that point, we probably are just going to see a lot of people building and not a lot of users. But this could change quickly. You know, I mean, we saw how quickly ChatGPT turns GPT technology from an obscure technology to something that is now a household term that everyone uses. So I could see MCP becoming something similar, you know?