OpenClaw, a constitution for AI, breaking dark flow, and open source as a moat?

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

**Ben Lloyd Pearson** (0:11)
And I'm your host, Ben Lloyd Pearson.

**Andrew Zigler** (0:13)
So Ben, how's your week been? You know, I'm looking at the stories on our desk for today, and there's a lot of stuff going on. Doesn't it feel that way to you?

**Ben Lloyd Pearson** (0:21)
Yeah, the AI era is not slowing down anytime soon, it would appear, because here's some of the stuff that we're covering this week. So recalculating SaaS motes in the AI era, breaking AI's dark flow spell, reading Claude's public rule book, and AI drained CUDA's GPU motes. Andrew, where do you want to start?

**Andrew Zigler** (0:40)
Well, all those things sound cool, Ben, but there is something else that we have to talk about first. Then it's the elephant in the room, or maybe I should say the lobster in the room in this case. There's been a whole phenomenon shaking the internet, introducing people to AI assistance this week. I think it's called Claudebot or something. No, actually no, isn't it called Moldbot? Have you seen this?

**Ben Lloyd Pearson** (1:01)
No, you haven't heard the latest news now. As of this morning, I guess, it's called OpenClaude. I mean, we're in the AI era, you got to change the name as quickly as you can prompt it.

**Andrew Zigler** (1:11)
I love that. It's like the idea of the software being so useful that no matter how much it changes its name, we're all racing to it and we still keep talking about it. Exactly. This phenomenon, this AI assistant, it turns Claude into a super powered user of your whole life. People are setting this up on their machines and giving it access to really all of the kinds of software applications that they use in a day-to-day basis, including confidential information and is effectively Claude code wired into your life with a lot of unique hooks and abilities to talk with you, interact over messaging platforms.
The idea being that you can host this virtual assistant somewhere on your own machine or in the Cloud and talk to it from anywhere. That's an amazing phenomenon, I think. I've been experimenting with my own formats of this, but I can't say that I have connected something like OpenClaude in my life. What do you think, Ben?

**Ben Lloyd Pearson** (2:05)
Yeah. Goodness, no. I'm not quite ready to make that step. Yeah. I mean, I love the idea, and it really shows to me how we're sort of on the verge, I think, of the cat coming out of the bag, so to speak, on what agentic AI is going to look like when it actually starts to hit our real life. So first of all, I would not be connecting any of my sensitive stuff to it. I don't really even want it running on my personal laptop. You haven't heard, some people maybe are live streaming it, too, and that just sounds like a recipe for disaster to me. But it would actually be really awesome to have some sort of separate device within my home or even hosted somewhere where I have this thing running for me. So it's almost like an agent that I can just call in whenever I need its help. But yeah, I mean, I think this is starting to open up the door to what software engineers have started to already discover in terms of how to work.
Now, it's sort of broadening to more knowledge work capabilities. So I think it's just a really good representative of like the zeitgeist of what we're all starting to experience with AI.

**Andrew Zigler** (3:11)
Yeah, I strongly agree. The two points you made there about one, not necessarily having the appetite to put it on your own machine, but recognizing the power of having it somewhere accessible in the cloud that you can work with, one. And then two, the widespread mainstreamness of this coming to people. I think that the recipe of those two things coming together, we are going to see a really unique evolution in software that we have all been alluding to in the industry, the explosion of people creating their own microapps and services and departments within companies, vibe coding solutions instead of renewing their favorite vendor for the year. And that brings us to the first article that we want to talk about today, which is the latest from Steve Yegge, which is about Software Survival 3.0. And in this essay, Steve lays out the idea of a survival ratio for software going forward based on his own reflections of having worked with a type of engineering process as Gastown. You know, if you're not familiar with Gastown, check out our most recent episodes. We've done lots of coverage about this topic and about how it's transforming engineering in an AI way. But in this essay, Steve talks about how tokens, energy and money are now three parts of our life that are eternally constrained by each other. And in the software world, it needs to solve for one of those three things in order to survive. And he lays out some really unique positions that some software are in, whether SaaS or things that run on your local computer, to survive on an AI world. One that stood out to me is like GREP or RIP GREP. You're talking about an incredibly efficient text searching process with your CPU that the GPU is never gonna be better at. And to build something that's more effective than GREP would cost too many tokens than it would realistically save you. So an application like GREP survives. He breaks this down into further layers. I thought it was a really interesting dive into the kinds of software he thinks will survive based upon his own workings with this tool now. It's coming from like four decades of engineering experience. So there's a lot to unpack here. Ben, what stood out to you?