episode 28 | November 17 2026
Manik Surtani: Open Source in the Age of AI, and Why It Needs a Neutral Home

Manik Surtani: Open Source in the Age of AI, and Why It Needs a Neutral Home
Beyond the Noise
About the episode
In this episode of Beyond the Noise, Matt Klein, bitdrift CTO and creator of Envoy, talks with Manik Surtani, co-founder and CTO of the Agentic AI Foundation (AAIF). Manik shares his career path, from messing with other kid's computers in the newly minted computer lab at his high school, to bootstrapping a consultancy that might've been huge if it weren't for the dot com crash.
They go on to discuss open source at length, from Manik’s time at Red Hat creating Infinispan, an open source answer to expensive proprietary data grids, to his 12.5 years at Square and Block, where he worked on payments infrastructure, took Cash App to the cloud, and co-launched gRPC with Google. He shares how Block’s open source coding agent Goose, and its co-development of MCP with Anthropic, led to the creation of AAIF. Companies were using Goose but didn’t trust a single vendor to own the roadmap and trademarks, so he and the MCP team decided to build their own neutral foundation.
They wrap up by talking through the biggest challenges facing open source right now, its place in the world of AI agents, and some of the hardest questions in tech right now: how can we ensure the benefits of AI reach everyone, not just a few big companies?
Episode #28 - Manik Surtani
===
[00:00:00]
Matt Klein: All right, folks. Welcome to another episode of Beyond the Noise: Signals, Stories, and Spicy Takes, the show where we dig into the stories of the people shaping the future of app-based computing. I'm your host, Matt Klein, co-founder and CTO of bitdrift, as well as the founder of Envoy Proxy. Each episode, we'll talk with engineers, founders, and technical leaders who've transformed the way their companies build and understand what's happening inside their systems.
We'll dig into the challenges, the breakthroughs, the lessons learned, and wrap it all up with their hottest takes, so let's dive in. Today, I am thrilled to have with us Manik Surthani, who is the co-founder and CTO of the Agentic AI Foundation. He's working to make sure AI agents are built on open foundations with developers and organizations able to shape the technology together.
He conceived and launched AAIF while at [00:01:00] Block, bringing years of experience in open source and infrastructure to a new set of challenges. Manik joined Square as an early employee and spent more than a decade at the company as it grew and became Block. He worked as a staff engineer and held several senior engineering leadership roles, including VP of engineering for platform and cloud at Cash App.
Most recently, he led open source at Block, working across its business and with the wider developer community. Before Block, Manik was an engineer and open source leader at Red Hat, where he created Infinispan. At Block, he also played a foundational role in creating gRPC. His work has given him experience of open source from several angles, writing the code, leading the teams, and helping organizations collaborate beyond their own walls.
Welcome. This is gonna be a, a really fascinating conversation. We have so much in common and lots to talk about, so, um, I, I think we're gonna have a hard time fitting it in one hour, but thank you again for joining.
Manik Surtani: Thank you for that awesome introduction, Matt.
Matt Klein: Yeah, yeah.
Manik Surtani: Um, wow, I didn't realize it was, uh, you [00:02:00] know, w-when, when I-- when you asked me to tell you a bit about myself, I was like, "Shit, that's a lot that I wrote down."
Anyway, hope that wasn't too much of a-
Matt Klein: No, no, no, no, it's fine. I think it's great for, for people to get to know you. Um, before we get into all of the interesting stuff, how I typically like to start is I'd love to learn about you, how you got into computers and, uh, just tell us a bit about how you wound up where you are today.
Manik Surtani: Yeah, absolutely, absolutely. Like, open source has always been in my DNA. I've always been the kind of guy that just always enjoyed the ethos behind open source. I think I downloaded my first Slackware kernel in the late '90s off of an ADSL connection, tried to compile it. Lots of fun back then. Um, yeah, no, I mean, I've always been into computers.
Even as a kid, I'd been taking apart my parents' VCR and, uh, sometimes not always being able to put it back together again. And I'm sure everyone's had similar stories. Um, but yeah
Matt Klein: And I, I mean, did you, um, you, uh, did lots of programming when you were a kid? Or like, how, how did you get into the [00:03:00] industry?
Manik Surtani: Yeah, very much so. Very much so. I mean, like, I was, um, very lucky that the, the high school I went to-- I grew up in Sri Lanka, so, um, the high school in Sri Lanka at the time actually had just ordered a brand-new set of BBC microcomputers for their brand-new computer lab, which, you know, very few schools at the time had, in Sri Lanka at least.
So I had access to one of those, and that was a lot of fun. Um, yeah, on, you know, BBC Basic. Mostly, I don't know, I mean, trying, trying to emulate the actual command line interpreter itself so that I could hijack other people's commands or do stupid things with it. Um, mostly just trolling people. But, you know, it was a fun little thing I used to do as a kid, and, uh, that's what got me into thinking, "This is actually cool.
I can program this to do whatever I want. This is a lot
of fun."
Matt Klein: Yeah. I, I mean, I, I think the world was a bit different back then in terms of access control, and you could just do-
Manik Surtani: Yes
Matt Klein: ... do whatever you wanted back then.
Manik Surtani: Very much.
Matt Klein: That's, uh, not, not the world that we live in today. Um, so a- and then f-from there, I mean, you know, you, you said that you wound up at Red Hat.
So just t-tell us a bit about, [00:04:00] again, like, how you actually started doing professional open source development.
Manik Surtani: Yeah, that, that, that story's a good one as well. Um, like, I was in university in Singapore, and sort of partly to fund my way through university, I was doing a lot of programming jobs at the time, and mostly as a freelancer.
This was in the late '90s, where, um, online commerce was just starting to become a thing. Um, there was this thing called Amazon, and they started selling books online and what have you, and every other bookshop in the world was like, "Crap, we need to do the same thing to keep up."
Matt Klein: Right.
Yep.
Manik Surtani: And, um, they were finding ways to, to try and sell their books online, but state-of-the-art at the time was this thing called Intershop.
You download this software package. Cost a lot of money. I think it was like $100,000 per installation at that time. Um, and not everyone could afford that, of course. So a lot of folks would come to me saying, "Hey, you know how to code, right? Can you help me build a web shop?" And I was like, "Sure. How hard can it be," right?
So here I was cobbling together a bunch of PHP scripts, running in MySQL, um, [00:05:00] and, and selling books online, really. Not nearly the scale of Amazon, of course, but this was for the mom-and-pop shops that were trying to keep up. So-
Matt Klein: How
did, uh, h-how did people do payments back then? Like how did you-
Manik Surtani: Yeah
Matt Klein: ... how did you accept payments?
Manik Surtani: Yeah, it's really interesting. So again, early days, but you know, you had things like WorldPay, predates Stripe. Um, I was using things like WorldPay to actually process credit cards online.
Matt Klein: Right.
Manik Surtani: It worked. It was very clunky. The APIs were horrible, um, very brittle, and they changed all the time, which broke things.
So I ended up building all these abstraction layers and things. But in the process, I realized I had built such a lot of infrastructure to make my life easier that meant that I could scale this quite easily. So when I left university, I decided to start a company, a consulting company, to start to build more and more web shops for, for people, um, using the infrastructure that I already had built.
Um, and that became a really interesting business. So I ran that for about three years, but we ended up getting caught up in the dot-com crash where a lot of our clients, a [00:06:00] lot of the, our companies paying us money were all startups themselves, venture-funded startups, and one by one they all started disappearing
Matt Klein: Yeah.
Manik Surtani: So yeah, it was, it was a fun three years, a lot of learning, um, a lot of open source work, 'cause everything we built was on open source, which meant a lot of contributing back to those open source communities.
Matt Klein: Yeah.
Manik Surtani: Everything from PHP to MySQL to, to Linux itself, and that's sort of where I cut my teeth on all this stuff, saying, "This is actually an amazing community with some really cool things here," and the fact that I can leverage all this work for my business was really interesting.
Matt Klein: Do you ever, do you ever think back to that time and think that if you had powered through the, the dot-com burst, you would've been Shopify?
Manik Surtani: It would've been very different, absolutely. Um, I mean, this is one-- my, my co-founder at the time refused to take any venture capital money.
Matt Klein: Right.
Manik Surtani: So we, this was-- we were basically bootstrapping ourselves, and he had this really, really strong perspective that he wanted-- he didn't wanna lose control of the company.
Matt Klein: Right.
Manik Surtani: And I was young and naive, and I kind of went along with it. But in another world, if we had [00:07:00] probably taken venture money, we would've been-- we might've survived the, the burst.
Matt Klein: Yeah. I mean, it, it's, um, I think it's so fascinating where I started a company when I was very young, which failed. Um, and I, I don't know, you know, you hear all these stories of, of people that start companies that are super successful when they're young, and I actually just talked about this on a, on a podcast that I just recorded.
But I think it's, I think it's interesting how, you know, you can gain a lot of perspective obviously when you're older-
Manik Surtani: Yeah
Matt Klein: ... but you do a lot of things when you're young. So I think that's super interesting.
Manik Surtani: Absolutely.
Matt Klein: Um, so, so that went bust, and then I'm-
Manik Surtani: Yeah
Matt Klein: ... guessing from there you, you wound up at Red Hat at that point or, or- Still a couple of
Manik Surtani: steps in between.
Matt Klein: Yeah.
Manik Surtani: Still a couple of steps in between. Um, one of-- I mean, a-as we wound down the company and, um, let go of all the people we had hired and everything else, we still had one last outstanding contract with a firm in the UK in London. Um, and I just had to go and wrap up all that work myself. So I flew out to London for all that, um, finished up all the work [00:08:00] there, and I was like, "Well, the company's gone now, so I'm just gonna go try and find a job out here."
And I did. I landed a, a consulting gig in, in London, again, based on a lot of the, the tech stack that I was working on anyway. So that gave me the, the chops, the qualifications to land a job in London, mostly again around Apache and PHP and Java-
Matt Klein: Yep
Manik Surtani: ... and what have you. So like great, cool, I know this stuff like the back of my hand.
This is great. Um, landed a consulting gig in London, um, stayed there for that. But again, as a part of that gig, I was contributing back to a lot of these projects. One of those projects happened to be JBoss, um, the Java app server from-
Matt Klein: Yep
Manik Surtani: ... the early 2000s. Um, and they realized that who's this guy, like, contributing all these patches all the time and what have you?
We've gotta go talk to this guy. So they reached out to me and said, "Do you want a job? Why don't you come and work for us full time?" I was like, "I could do that? You're gonna pay me to work on this open source thing full time?"
Matt Klein: Yeah.
Manik Surtani: "Really?" I was like, "Yep, absolutely." I was like Blew my mind. This is madness.
Matt Klein: Yeah.
Manik Surtani: I, I- Sure, I'll do that.
Matt Klein: Yeah, I mean, uh, as, as someone that has obviously worked on a very large [00:09:00] project myself, I think one of the most fulfilling things that happened to me o- over the, you know, 10-plus years of doing Envoy is watching other people get jobs, like, through-
Manik Surtani: Yes
Matt Klein: ... working on that. I th- I think it's such an amazing-
Manik Surtani: It's incredible
Matt Klein: thing that happens. Uh-
Manik Surtani: Yeah, yeah ...
Matt Klein: anyway, sorry, keep
going.
Manik Surtani: That's really building-
Matt Klein: Yeah
Manik Surtani: ... an ecosystem-
Matt Klein: Yeah
Manik Surtani: ... and building a community as a result.
Matt Klein: Yeah.
Manik Surtani: Um, but yeah, so I, I was an early, uh, engineer at JBoss and, um, running-- Like, I wrote a lot of the clustering code for the app server myself, so both the HTTP clustering and EJB clustering and things like that.
Um, very quickly after that, we got acquired by Red Hat, and while at Red Hat, it was clear that people were using all the clustering code that I had written for the app server, but they didn't want the app server. They were pulling the clustering co- code out of it and using it to share arbitrary data-
Matt Klein: Yeah
Manik Surtani: um, instead of sessions.
Matt Klein: Yeah. A-a- and when you had, uh, worked on that clustering code, uh, ha- had you worked on that type of distributed systems [00:10:00] stuff prior to that?
Manik Surtani: No, that was my first time.
Matt Klein: Or, or, or was that, like, a learning on the job-
Manik Surtani: Yeah
Matt Klein: ... situation?
Manik Surtani: That was learning on the job.
Matt Klein: Yeah.
Manik Surtani: That was my first time getting involved in all of that, which was great.
It was really, really interesting.
Matt Klein: Yeah.
Manik Surtani: Um-
Matt Klein: I, I, I just like-- To me, it's, it's so interesting to look back at that time because for most people who are, who are listening, I mean, many of the listeners, you know, like, they might not be aware of all the technology, where it came from, and all of the underpinnings of, uh, of, you know, of everything that we have today.
But so much less was known back then about how to-
Manik Surtani: Right
Matt Klein: ... how, how to build, how to build distributed systems. Like, everyone was just figuring it out. I mean, there was nothing back then.
Manik Surtani: Yeah.
Matt Klein: So,
like-
Manik Surtani: There
was
no CAP theorem in back... There
was no CAP theorem.
Matt Klein: No, no. Yeah, yeah. So, so, I mean, like, tell us a bit about how, how did you do what you did back then with the distributed systems-
Manik Surtani: Yeah.
So,
I mean, it was a lot of fun ...
Matt Klein: kind
of figuring out
Manik Surtani: There, there was a project called JGroups that JBoss relied on quite a lot, which was a, um, a UDP broadcast library in Java.
Matt Klein: Yeah.
Manik Surtani: Um, and the guy who created [00:11:00] JGroups had joined JBoss as well. He's the guy who actually recruited me and said-
Matt Klein: Right
Manik Surtani: ... "Hire this guy," you know.
And I learned a ton from him, so he-- it was great because, like, he was an academic prior to that. He'd spent a lot of his time prior to that researching network broadcasts and things like very low-level network protocols-
Matt Klein: Yeah
Manik Surtani: ... um, in his academic career, and his role at JBoss was to implement all of that in Java for the app server.
Matt Klein: Right.
Manik Surtani: So that was great. I had someone who was very, very steeped in, in, in all of that. Um, so-
Matt Klein: Yeah
Manik Surtani: ... I was able to work with him very closely. That was basically my, my on-ramp into that. And-
Matt Klein: Amazing to
have that experience.
Manik Surtani: Yeah, it was fantastic. And then once, like I said, once Red Hat acquired us, I found that a lot of people were pulling out these libraries that we had written, these components from the app server and using it for other things.
So I kind of spotted, spotted some demand there, and that's where Infinispan was born from. It's like, cool, we need an arbitrary distributed data stored in-memory data grid. This is what people want. I can see why they're using it. I can see how they are using it. Um, and that's an unser-underserved [00:12:00] market.
Um, Tangosol Coherence was a product at the time, uh, which, which was very popular in that space, but very expensive. It was proprietary, and people were looking for an open source alternative.
Matt Klein: Yeah.
Manik Surtani: Um, and, and we became that. That's what Infinispan became.
Matt Klein: Fantastic. And then did you work on that for a while?
Uh, like-
Manik Surtani: Yes
Matt Klein: ... what, what, what, what was the experience of growing that community
Manik Surtani: like? That was an absolute rocket ship. Um, it-- I, I, I clearly remember the day it started. We were actually, um-- So Red Hat was a distributed company. Everyone worked remotely. Everyone worked from home. But we, every quarter, we'd all try and meet in person for design sessions, for hacking sessions, things like that, and the company paid for that.
They sponsored that. So this was one of those very early design sessions, uh, where a group of us had met up, I, I, I believe it was in, in the Czech Republic, in Brno, where Red Hat had a physical office. So they agreed we'd have physical office space and all of that to use. Um, there was a group of us out there.
We were talking about the next version of the app server and stuff we wanted to design in there and stuff, you [00:13:00] know, features that people were asking for.
Matt Klein: Yeah.
Manik Surtani: And I remember that I stayed on for an additional week after that with a couple of folks from that team, saying, "You know, there's this thing I wanna try.
People are using all this clustering code for other things. Let's go design a thing around that." And we spent that first week just basically hacking out the first version of Infinispan, and it came to being in, in the end of a week really of like this pressure cooker environment, uh, just, just like four or five of us hacking on it.
Um, went back and we launched it, and, and it, it just took off from there.
Matt Klein: Amazing. Yeah. And then is that technology still used today? Is it-
Manik Surtani: It is ... still out there? I'm surprised actually it still is.
Matt Klein: Yeah.
Manik Surtani: I haven't, I haven't heard of or touched it in a very long time.
Matt Klein: Right.
Manik Surtani: But of course, it's, it's a, it's an important product line for Red Hat.
They productize it. They've got a commercial version of it.
Matt Klein: Right.
Manik Surtani: They sell that to a lot of the big banks. Um, and the open source code still exists as well. I mean, it's, it's progressed-
Matt Klein: Yeah
Manik Surtani: ... many, many versions since I last touched it. Um, it's got a large community around it now and everything else, which is really nice to see.
Matt Klein: Amazing,
yeah.
Manik Surtani: When I left Red Hat, there [00:14:00] must've been, gosh, I wanna say eight or nine full-time engineers at Red Hat on it. Um-
Matt Klein: Yeah
Manik Surtani: ... I think there must be twice that now, and of course, lots of other folks in the community, which has grown a lot since, since I last touched the project.
Matt Klein: How do you feel about starting something like that, that has become so successful, and then walking away from it, you know?
Because, like, something that I think about myself a lot is-
Manik Surtani: Yeah ...
Matt Klein: start- starting Envoy, I, you know, to be honest, I'm not really day-to-day involved in it anymore. I just don't have the time.
Manik Surtani: Right.
Matt Klein: And, um, I don't know. I just, I find that topic so interesting because I think you see two camps. You know, you see these people that start these projects, and they do it for life, and it, at least for me, I'm not that kind of person.
I can't do something-
Manik Surtani: Yeah
Matt Klein: ... for life. It's just not me. Um, but I don't know. Do you ever think about that, you
know-
Manik Surtani: No, I'm very much the same. I, I can't do one thing for life either.
Matt Klein: Yeah.
Manik Surtani: Maybe, maybe it's my ADHD showing or something. I don't know, but, you know, there's, there's... I'm always looking for that next challenge-
Matt Klein: Yeah
Manik Surtani: for the next [00:15:00] interesting thing.
Matt Klein: Yeah.
Manik Surtani: And, um, I mean, like Infinispan got to the point where I was writing less and less of the code. I was now more of a, a, a g- gardener, for want of a better word.
Matt Klein: Yes. Right.
Manik Surtani: Just fielding in pull requests. I mean, you probably feel the same.
Matt Klein: Sure.
Manik Surtani: I mean, I'm spending more time looking at pull requests and feature requests and saying no to them because, I mean, you're just getting swamped with it, and I've got no time to actually do the things I like doing, which is actually-
Matt Klein: Yeah
Manik Surtani: building stuff.
Matt Klein: Well, and that was back then before now, which we'll get to, you know-
Manik Surtani: Yeah
...
open source was getting
Matt Klein: swap- swamped times 10, but anyway, um,
so-
Manik Surtani: But that was a clear signal for me that I needed something else.
Matt Klein: Right.
Manik Surtani: That this wasn't doing it for me anymore.
Matt Klein: All right.
Manik Surtani: And it was, yeah. And then that's when I spotted an opportunity at, at, uh, at Square.
Um, an old friend of mine from Red Hat had moved to Square back then, and, um, um, and I knew Bob Lee, who was then CTO of Square as well, um, through the Java circuit, and we used to speak at JavaOne together-
Matt Klein: Right
Manik Surtani: ... and do that. And, uh, um, I happened to be in San Francisco. He, um, went out for a drink. He was like, "You know, [00:16:00] why don't you come and join?
The company's going great. There's lots of cool stuff here, really interesting challenges," so on and so forth. And I said, "You know what? Okay, fine. Let, let, let's see what this is all about."
Matt Klein: Yeah. I, I, I, I mean, from, from, from talking to people in that early era at Square, it seems like their strategy for getting people was to just pull in really smart open source people and let them do, do good things, and I just-
Manik Surtani: Yeah
Matt Klein: you know, I th- I think that's an interesting recruiting strategy, and I don't know if that's your impression of what-
Manik Surtani: Yeah
Matt Klein: ... it was like, but-
Manik Surtani: I think it's a really powerful recruiting strategy because, I mean, open source, I, I mean, you're clearly talking to someone who's, who's very convinced that open source is the right way forward here.
Right. But, uh, even back then, even in terms of recruiting, it meant that you got people who cared deeply about what they were doing.
Matt Klein: Sure. Of course.
Manik Surtani: It wasn't just for the paycheck. Right. So suddenly you've got stickiness. Um, people aren't gonna get distracted by someone else offering 20 grand more or something, and they'll go and quit their job over that.
That didn't happen within the open source community as much.
Matt Klein: Right.
Manik Surtani: There's far more [00:17:00] stickiness to the project and to the work. And also what that, what that meant was open source begets more open source people, right? So you're in the community already. You're hiring people you already know.
Matt Klein: Yeah.
Manik Surtani: Your con- the confidence level of the people you're hiring in is so much higher, right?
'Cause, you know, you may have only spent, um, um, whatever, half a day interviewing that person, but you've seen that person's work-
Matt Klein: Absolutely
Manik Surtani: ... for many, many months.
Matt Klein: Absolutely.
Manik Surtani: You've worked with that person probably in the open source community many times over, like talking over a pull request, going back and forth over a feature or a design.
So, you know, the level of confidence when you hire people through the community I think is so much higher, and we saw that in the early days of
Square.
Matt Klein: Yeah. I, I, I mean, at the risk of jumping ahead, I, I have so many questions that I wanna ask you- ... and that I ask myself daily now about how much has changed in the last two years.
You know, h- how-
Manik Surtani: Yeah
Matt Klein: ... the agentic software lev-- the agentic software revolution has changed open source. And I think people don't, don't know yet, but [00:18:00] I know from my own experience, just either just seeing not only the volume of pull requests, but even for myself, my willingness to fork and just say goodbye is infinitely higher than it-
Manik Surtani: Yeah
Matt Klein: than it, than it was previously. And I, I just think we are undergoing a, you know, a sea change in software development that we haven't seen in decades. Um, and I, and I, and I'm so curious to get your thoughts on it. But before we get there, um, would, would love to briefly just talk about your time at Block.
I mean, it sounds like you worked-
Manik Surtani: Yeah
Matt Klein: ... on a lot of interesting things. So t-tell us a bit about that era.
Manik Surtani: Yeah, Block was a lot of fun. I mean, it-- I was there for 12 and a half years. That's, that's, you know, a quarter of my life.
Matt Klein: Yeah.
Manik Surtani: That's a long time to be in one place.
Matt Klein: Yeah.
Manik Surtani: And, um, for, for, for someone who's always looking for the next thing and what have you, right, that, that sounds counterintuitive, right?
Matt Klein: Right.
Manik Surtani: Um, I think I must have had at least five different careers at Block in this 12 and a half years. [00:19:00] Um, I was a staff IC on the original payments stack. Um, I was doing a lot of decommissioning of legacy systems and migrating to modern systems while you couldn't switch them off. You know, we were processing, you know, like hundreds of payments a second sort of thing.
And then while doing that, had to shut down that payment system and migrate it to-
Matt Klein: Yeah
Manik Surtani: ... you know, something that was more distributed and more scalable that we'd already designed and launched. So really fun challenges doing that through to running Cash App and running infrastructure for Cash App, taking Cash App to the cloud out of our data centers.
Matt Klein: Yeah.
Manik Surtani: Um, and, and at one point, uh, along the way, I mean, you know, launching gRPC or co-launching it with, with Google, um, that's actually a really fun story. So, uh, predating gRPC, Google had an internal system called Stubby-
Matt Klein: Yeah
Manik Surtani: ... which was very simple, and GRP-- well, very similar. The gRPC-
Matt Klein: Yeah
Manik Surtani: ... was based on that.
But so did Square. Um, Square had a bunch of ex-Google people who also knew about Stubby and wanted something similar, and there was no open source thing to do.
Matt Klein: Right.
Manik Surtani: So we built our own [00:20:00] version of it called Sake. And we had Sake running, and Google had Stubby running, and I was, I was, um-- that team reported to me at one point, and we were looking at modernizing Sake.
And one of the things we were looking at was, well, you know, HTTP/2 is out, and we want streaming and all this stuff, so we should now change the underlying protocol for, for the RPC system. We wanted it to be open sourced, and we wanted to modernize it based on Proto3 and everything else that was coming out.
It's like, okay, cool. There's all this stuff, right? This is basically a brand-new code base. This is not Sake anymore with all these major changes we're doing. Okay. So we're gonna call it something else. We're gonna open source it anyway. And then I got wind of the fact that Google were gonna do the same thing with Stubby.
So, um- Through a few n- connections I had, I got introduced to those folks and said, "Okay, tell you what, if you're gonna do this, let's go do this together."
Matt Klein: Right.
Manik Surtani: Um, and they were very excited about that. And, um, I believe Netflix was involved as well in very early stages saying, "Oh, the three of us, we're really gonna open source this thing together and, uh, contribute to a thing."
And we used to host all the design [00:21:00] meetings at Square.
Matt Klein: This is-- Yeah, th- this, this is what, like 2014, '15?
Manik Surtani: Yeah,
yeah, '15.
Matt Klein: Right. Yeah, okay. Yeah. Yeah.
Manik Surtani: And we ended up... But Netflix in the end bailed on it. They, they realized that it was going in a direction they weren't happy with or weren't as interested in, so, so they bailed on it.
So it was just Google and Square. And we did it, we launched it, we took it out and, uh, and, and it started to grow. Google, of course, having a much bigger vested interest in it, poured in more and more resources.
Matt Klein: Yeah.
Manik Surtani: On the Square side, it was the opposite. It's like, "This is great. We have exactly what we need, and we don't need to put in as many resources anymore."
So we started dialing back that rather large team that I had that was maintaining Sake, and today we have gRPC for free, and we don't, you know, have, need a team to maintain it.
Matt Klein: I, I mean, the, the, uh- Parallels between that and the beginnings of Envoy is actually fairly large. Um-
Manik Surtani: Oh, yeah?
Matt Klein: ... you know, because we-- Yeah, because we were, we were doing Envoy at Lyft in 2015.
Manik Surtani: Mm-hmm.
Matt Klein: And Google was doing something similar around gRPC. Um, and they caught wind of the codebase [00:22:00] and then decided to start using it. And I, I mean, you know, o-obviously, there's a lot of history there, but, um, in, in the same way, I was one person with one or two people helping me.
Manik Surtani: Right.
Matt Klein: And then suddenly, within six months, you have, like, eight people at Google who are working on Envoy and all of those things.
And then-
Manik Surtani: Yeah, that's right
Matt Klein: ... you don't, you don't need that many resources anymore. Like it is-
Manik Surtani: That's right
Matt Klein: ... it is, it, it is quite nice when that happens. Um-
Manik Surtani: It's quite nice also because Google, being Google's scale and size, can easily afford to put those resources on.
Matt Klein: Yeah.
Manik Surtani: Whereas for Block, at the time, it was like, you know, every engineer we save is, is really valuable-
Matt Klein: Yeah
Manik Surtani: to the rest of the company.
Matt Klein: Yeah.
Manik Surtani: So we started doing that.
Matt Klein: Yeah, I mean, we could have an entire episode, which we don't have time, uh, for, on all of the different framing formats and serialization libraries and all of those things. But I will say that I think Proto3 in particular, I, I think has been one of the most amazing things that has been given to the industry in a long time.
I mean, I, I think it's a, it's an amazing piece of technology, and I think people have wrapped [00:23:00] it in gRPC and other framing formats and Connect and whatever. You know, you can have-
Manik Surtani: Yeah
Matt Klein: ... all the framing around it, but I think it is a, a pr-pretty amazing piece of technology. Um-
Manik Surtani: No, absolutely. If you look at, I mean, how much of the world runs on Proto3 today, it's incredible.
Matt Klein: Absolutely. Um, so, so take us forward. I mean, you were obviously working on a lot of really amazing things at your time at Block. Um, what, what made you ultimately decide to leave and start the foundation? Tell us that journey.
Manik Surtani: That was almost by accident again as well. So, so I was running open source for Block at this point.
Um, um, Jack Dorsey was very keen to reboot Block as an open source first company. This was in 2023-ish.
Matt Klein: Yep.
Manik Surtani: And, um, it's like, yep, we did a lot of open source anyway. It was deeply a part of our DNA, but actually not everywhere, mostly in engineering, and that's it.
Matt Klein: Right.
Manik Surtani: But he really wanted the entire company to embrace open source, like open source values.
And, um, [00:24:00] that was like, "Okay, that's an interesting challenge," and he wanted me to spearhead that. It's like, okay, that could be a lot of fun. So I started doing that, but along the way, that just coincided and overlapped with the, um, generative AI era-
Matt Klein: Yeah
Manik Surtani: ... which was also starting to take off. So we had a group internally that started building Goose, which is an open source coding agent-
Matt Klein: Yes
Manik Surtani: which predates Cloud Code and Codex and everything else. And that was a small thing. A few people were using it, it's like, okay, that's gaining some traction. It's kind of interesting. Um, more and more people internally started using it. I could see that growth curve going up. It's like, okay, well, we should open source this.
It's pluggable by design.
Matt Klein: Yeah.
Manik Surtani: I wanna see what the world does with it and where they take it. So we did it, we open sourced it. That started getting very popular very quickly. I think we had 30K stars in a month or something like that from launching. Um, and we were working very closely with Anthropic at the time, so we co-developed MCP with Anthropic for Goose originally, because Goose w- wanted more tools, more access to tools in a standardized way-
Matt Klein: Yeah
Manik Surtani: and things like that. So we became design partners with [00:25:00] Anthropic. Um, and then fast-forward a couple of months later, we were seeing a lot of adoption of Goose, but we were also seeing a bit of back pressure coming from the community where people didn't trust Block with it. Mm-hmm. They're like, "It's great that it's open source.
We love the Apache license. It's nice and permissive, but who owns the roadmap really? And who owns those trademarks?"
Matt Klein: Right. And,
yeah.
Manik Surtani: "Is that gonna be a rug pull at some point?" Right. "Are you just gonna start charging for it or something?" So, you know, I mean, like Red Hat were very keen to distribute Goose in Red Hat, but they didn't wanna do it until they had some guarantee that it was-
Matt Klein: Yeah
Manik Surtani: gonna be there for, for the long haul. Um, we had Stripe, we had Databricks, we had loads of other companies that were using it internally and adding to it and forking it and, you know, improving it-
Matt Klein: Right
Manik Surtani: ... but didn't wanna share their improvements because of this, because of the lack of neutral governance.
Matt Klein: Yeah.
Manik Surtani: So we went down that path of exploring what does neutral governance look like. Um, and I realized that at Anthropic, David, uh, who created MCP, was going through a very similar challenge as [00:26:00] well. And, you know, we sort of like shook hands. We talked about it like, you know, just like sharing each other's findings.
Like, you know, is the Apache Foundation the right place? No. Why not? For these reasons. Okay. What about the Eclipse Foundation? No, probably not for these reasons, et cetera, et cetera. And at one point, I just asked David, "Should we just go create our own damn foundation?" He's like, "You know what? That sounds like a good idea."
So we'll craft it, and we'll design the foundation to suit our needs, right? And there's nothing in there that's doing anything for agentic AI specifically right now anyway, so it makes sense. And that's how the foundation was born originally. Um, OpenAI got wind of it. They wanted to play as well, so they came in, and when we launched in December, we had 60 companies behind us.
I was like, "Okay. Wow, this is something." So as we wrapped up that, like December 2025 when we launched it, uh, so as we wrapped up the year, I was looking back at the year saying I had a lot of fun collaborating with Anthropic and OpenAI and launching this thing and what have you. And then looked back at the rest of my day job saying, "I don't think I wanna do that anymore."
Matt Klein: Yeah. Was it, um, w-was it [00:27:00] tough to, I guess, talk with the leadership at, at, at Block that this was the right thing to do or, or they were supportive? Like how, how did that process go?
Manik Surtani: Oh, they were highly supportive. I mean, as I mentioned, Jack Dorsey is always a big-- has always been a big supporter of open source in general.
Matt Klein: Yeah.
Manik Surtani: He was very supportive of launching the foundation and, um, basically what, what I told him was, "This is taking that very same idea and taking it to the world," right? Taking it outside of Block's doors and taking it to a bigger audience, and he was very supportive of that.
Matt Klein: Yeah. I, I mean, I, I'm gonna show my own bias here, whi-which is that having been fairly involved in the CNCF, the Cloud Native Computing Foundation in, in the mid-2010s, um- I, I, you know, these foundations can be so-- come so, so political.
Um, a-and I'm not asking for, for dirt about, like, particular companies.
Manik Surtani: Mm-hmm.
Matt Klein: But I, I, I mostly just wanna understand that how do you view [00:28:00] running these foundations in terms of balancing some of the business politics versus the technological progress, you know? Obviously, you can't, you can't completely set them aside, right?
Um-
Manik Surtani: Absolutely.
Matt Klein: But, but it's like how do you, how do you think about this in terms of that type of environment?
Manik Surtani: No, it's definitely very tricky and then, you know, challenges always exist, but I like to go back to values, purpose. Like, why did we create this foundation? Why does it exist, right? And I usually like to take people back to that first, and usually that's a really strong aligning point.
Um, people are only involved because they care about the community. Um, I mean, there-- y-you get a few companies that try and use it as a place to market their product and all of that, but, you know, we try and talk all of that down. That's not the purpose of this foundation. This doesn't exist to be a marketing platform for your-
Matt Klein: Right
Manik Surtani: startup. That's not the purpose. The purpose here is to grow the community and to provide value to the community. And that, again, goes back to take one step even further back, um, why is open [00:29:00] source important in the age of AI? And again, we align on those values first and then take people back to what's the purpose, this is why we're here.
And that usually works to diffuse a lot of the things that I
see
Matt Klein: Let's actually talk about that because that, that was what I was asking you before is from m-my perspective, um, just like living through the last two years and especially the last year and especially the last six months, things are moving so quick.
I think open source is changing right now. Um, and I, I would love to learn from you, you know, how you think open source is changing, how it's not changing. How do you view the role of open sourcing in the core-- or sorry, the role of open source within the core tooling itself?
Manik Surtani: Yeah. Well, it's all changing very fast.
I mean, we're all learning from each other. I mean, I, I'm no-- you know, I, I don't have any particular wisdoms that others may not already in this space.
Matt Klein: Yeah.
Manik Surtani: Um, I think there are multiple facets to it. There is open source from a values [00:30:00] perspective that isn't changing, and I think that's really important for AI, and I'll talk about that in a second.
But I think the day-to-day practice of open source is the bit that you're talking about that is going through this massive change. Um, there is, uh, you, you touched on this earlier, people are getting swamped with pull requests and with issues being filed and things like that. A lot of it's AI slop. A lot of people are just sitting and just pumping out stuff out there and, and that swamps maintainers.
That's really hard. Um, I see that not even in open source projects, but just in, in the foundation itself with people creating, um, new project proposals, and you can see that this is just AI slop really coming at you. It's like- What do we do with this? I mean, it's easy to say fight AI with AI and have your own AI reviewers and close things down and what have you-
Matt Klein: Right
Manik Surtani: but that doesn't-- uh, you miss a lot as well. So, so there is that. Um, I don't know. I don't, I don't think there is a clear good answer right now as to how you deal with that. [00:31:00] I, I see different projects doing different things. Like, I know of projects that have basically just stopped accepting pull requests, saying, "If you have an issue, file an issue.
We will generate the code," because code's being generated anyway.
Matt Klein: Yeah. I, I think what I am most personally curious about is when you were talking before about how you, how you got your start. You know, you were working on the LAMP stack and you were-
Manik Surtani: Mm-hmm
Matt Klein: ... contributing to Linux and MySQL and Apache and all of these things.
And to me back then, the value add of open source on the world in terms of its benefit economically, um, its ability to create better outcomes, just like the compounding effect of multiple people working on the same thing is so abundantly clear, right? I mean, it's like I think-
Manik Surtani: Mm-hmm
Matt Klein: ... very few people could dispute, you know, how much benefit there has been from, from, you know, that entire environment.
What I am really curious about right now, and I think what I would concretely ask you, [00:32:00] is do you think the value economics of open source itself have actually changed? Um, and I, I don't know the answer, but I'm not sure, and I think it's amazing for me to say that, which is that it's just not clear to me right now that open source-
Manik Surtani: Yeah
Matt Klein: has the value that it did back then when creating code of increasingly high quality is cheap now.
Manik Surtani: Right.
Matt Klein: So I, I just wanted to get your opinion on that thing specifically.
Manik Surtani: Yeah, that's an interesting one. I actually think, um, you are right. A lot of open source goes away because there's no need for it anymore.
You can just generate it yourself, custom to what you need. You don't have to grab some random JavaScript library that does, I don't know, something tiny where you can just generate that code and align it yourself, right? Um, so there is certainly that aspect of it. But on the other end of the spect-spectrum, I don't expect people to be generating their own operating system kernels every time-
Matt Klein: Right
Manik Surtani: they need to boot up a computer, right? Uh, so, so [00:33:00] there's a middle ground somewhere. There's a tipping point somewhere where on one side of it, open source is still really important.
Matt Klein: Right.
Manik Surtani: And on the other, all of the stuff that sat on that side isn't, and that just goes away.
Matt Klein: Yeah. I, I think it's just interesting to figure out where is that line, right?
Manik Surtani: Right.
Matt Klein: You know? Be-be-
Manik Surtani: Yeah
Matt Klein: ... because at least, again, just speaking from my own personal experience, there are things that I have forked recently that I probably would have put more effort into upstreaming, you know? And-
Manik Surtani: Yeah
Matt Klein: ... and I, and I just think the calculations have changed, and the world is trying to figure out how to actually deal with that, which I think is really interesting.
Manik Surtani: Yeah, so let me put it this way, right? Like, how many of those things that you forked were the Linux kernel?
Matt Klein: No, no, no.
Manik Surtani: Not
many, right?
Matt Klein: Yeah,
yeah, yeah, yeah. Sure,
sure.
Manik Surtani: So clearly something on one side of a particular line
you wouldn't-
Matt Klein: Of course ...
Manik Surtani: right?
Matt Klein: Of course. Yeah. I, I just think that as an industry, we are figuring it out right now, and I think-
Manik Surtani: Yeah
Matt Klein: no one knows the answer i-in terms of where it's gonna land, y- especially as, um... I'll be honest, I mean, the quality of the tools and [00:34:00] the output has increased so rapidly, um, that e-even, you know, now versus six months ago, what can be produced to high quality-
Manik Surtani: Oh, God. Yeah
Matt Klein: ... it has, has changed a lot. And even, um, I tend to agree with you that, you know, slop in project planning and communication and writing is still rampant, but the code output of these tools is pretty damn good at this point.
And like, I think, I think slop is more... probably you would agree, it's more on people dumping features that shouldn't be there. Like they're-
Manik Surtani: Not the maintainers
Matt Klein: ... they're more like high-level design decisions, if that makes sense.
Manik Surtani: 100%.
Matt Klein: You know? Yeah.
Manik Surtani: It,
it's not the quality of the code that I have a problem with.
Matt Klein: Right. Yeah.
Manik Surtani: It's the fact that, uh, this product doesn't need this-
Matt Klein: Yes,
exactly.
Right
Manik Surtani: ... this screen or whatever. Yeah.
Matt Klein: Well, right. And, and that even with the AI tools, it doesn't mean that maintenance is free. It's like you still have larger-
Manik Surtani: That's right
Matt Klein: ... context and larger burden and all of these things of, like, things that can break.
So, so tell us a little more about the foundation [00:35:00] itself in terms of what are the goals of the foundation? Like, what, what do you view your stewardship of moving forward? I mean, is it-
Manik Surtani: Yeah
Matt Klein: ... is it, is it gonna be, like, a, a project hosting organization only, like a neutral holding ground? Do you view it going into other areas?
Tell us a bit about that.
Manik Surtani: So initially when we started it, it really was pretty simple. We just needed a home for Goose and MCP.
Matt Klein: Right.
Manik Surtani: Um, the, the very, very early genesis of it, right? And then as we progressed, um, as other companies got involved, it's like clearly there's demand for, um, um, an open home for a set of protocols that the agentic world runs on.
Okay, does that mean it's more like the W3C then, you know, as a basket of internet-
Matt Klein: Right
Manik Surtani: ... of web protocols? Is that what it is? Well, not really, 'cause there's also a bunch of infrastructure and a bunch of, you know, tools and things like that. So okay, does that make it look like more like the CNCF for agents maybe?
But, you know. So right now, in terms of what type of projects we have, um, [00:36:00] it, it's not, it's not one clear thing or the other. And that's not a bad thing. It's because everything's changing so fast anyway. I think it's all right to say, "Yep, we host infrastructure, we've got tools, we've got protocols, we've got specs and standards as well."
Matt Klein: Yeah.
Manik Surtani: Right? Um, and you know, whatever it takes in between to make all that work. Another way of looking at it, which I prefer, which is clearer in my head, is if we look at all of what it takes to run an AI system, um, end to end- There are multiple layers in there. You've got the hardware layer at the bottom, of course, and that's, um, notoriously not open, but anyway, we'll g- we'll get back to that later.
Matt Klein: Yeah.
Manik Surtani: There's a hardware layer. You've got a layer of-- a training, um, layer, the tools and infrastructure needed to train models, the models themselves, serving models and model inference, and then the layer on top, which is, which is the agentic layer, where you're using everything underneath to, to interact with the world and provide real-world, uh, value, right?
And that is sort of the layer that I'm sort of focused on. If you look at it that way, it's all the stuff you need at that point above the model to, to interact with the world. [00:37:00] And that is, um, um, protocols like MCP or A2A, but also protocols like, uh, commerce protocols, UCP, uh, and, and MPP and things like that, and the infrastructure to make all that work, the gateways, the routers, uh, the frameworks to build those agents, the, the, um, um, the stan- the standards and the specs around, uh, identity and authentication to make all of that work.
Matt Klein: Right. A- and I, I mean, this might be a stupid question, but why, why wouldn't a foundation like this eventually also host open weight models? Like, why wouldn't it eventually host open hardware or open GPUs? You
know, it's like-
Manik Surtani: So I think it absolutely might-
Matt Klein: Yeah
Manik Surtani: ... at some point, right? Like, whether it's this foundation or a different foundation, I don't know.
I think one needs to exist.
Matt Klein: Right.
Manik Surtani: Um, purely as a scoping exercise, though, I wouldn't want it to be this foundation, at least not right now. There's, there's enough work on our plate j- just to make the agentic space work well. Um-
Matt Klein: Yeah
Manik Surtani: ... and everyone's very focused [00:38:00] on that. I think once we start to get into the models themselves, that is a whole lot of extra work, and I'd rather that happen somewhere else.
Um, probably another Linux Foundation, uh, sponsored foundation as well could, could be that.
Matt Klein: Right.
Manik Surtani: Similar to how, similar to the PyTorch foundation that focuses on the, the tooling for training.
Matt Klein: Yeah. I-- You know, a- and, um- It, it is such a fascinating space to me, uh, uh, I mean, obviously for many different reasons, but one of them is just because of how fast it's moving.
You know? Like things-
Manik Surtani: Yeah
Matt Klein: ... are constantly changing, constantly moving. And, and I think my next question for you is, I know that you're, you know, that you're very early, but how are you thinking about the maturity of the different tools and things that you host within the foundation? Again, just coming from the CNCF-
Manik Surtani: Yeah
Matt Klein: you know, they have their sandbox and their incubation and their graduation, and I have seen the drama and the ridiculousness of that over time, you know, but I understand the need of it. And I, you know, I, [00:39:00] I would just love to learn about these, even these core protocols like MCP, MCP v2, Agent Control Protocol, all of these things are changing constantly.
Manik Surtani: Yeah.
Matt Klein: So how do you, how do you signal to people what the relative maturity level is of these things, you know, so that they can start making their decisions about whether to build on it yet or wait?
Manik Surtani: Yeah. So basically, we started out pretty much copying, um, the CNCF structure around maturity, and very quickly realized that that doesn't apply.
Matt Klein: Yeah.
Manik Surtani: It doesn't work here because, you know, like, um, what was the second project in CNCF? Was Prometheus or something-
Matt Klein: I don't know. Yeah
Manik Surtani: ... which is already super mature. Right.
Matt Klein: Right.
Manik Surtani: It's like-
Matt Klein: Exactly. Right
Manik Surtani: ... I, you know, we're not gonna wait for years for something to mature-
Matt Klein: Right. Right
Manik Surtani: ... in this space and then bring it in.
Matt Klein: Right.
Manik Surtani: Everything is so young. So we had to change tack about how we think about things. Um, and the sandbox is really the space that makes the most sense for us, where we can bring in things at a very early stage. It doesn't matter if you don't have, you know, [00:40:00] multiple companies contributing to it and, you know, multiple production deployments and things like that.
Instead, what we look for is, is community signal. Is there community momentum? Are developers excited about this? Are they getting behind it? Are they really interacting with it meaningfully?
Matt Klein: Right.
Manik Surtani: And if that's so, well, okay, that's a good signal. Let's bring it in as a sandbox project and let's, let's help you grow, right?
There, there clearly is something here.
Matt Klein: Yeah.
Manik Surtani: Because guess what? Everything is early stage in agentic AI.
Matt Klein: Well, right. Of course. I mean, and, and it's, um- I think that, I don't know. It's just that right, right now there, even within each individual tool category, there's so much competition of, of people who are, who are jumping in, um, that, yeah, I mean, it, it is an absolutely, uh, whirlwind crazy space right now.
Um-
Manik Surtani: Yes,
it's a Cambrian explosion at the moment.
Matt Klein: Yeah.
Manik Surtani: And I think that's kinda cool from where I'm sitting because I get to see all these experiments taking place everywhere, and seeing what falls out of those experiments, right? Where do things coalesce? What [00:41:00] dies? What survives? What gets better? Uh, what merges together?
Things like that. And there's a lot of that happening.
Matt Klein: Yeah. What, what has been the biggest surprise to you so far?
Manik Surtani: The amount of energy people are putting into gateways and routers-
Matt Klein: I, I, I- And- I know. There, there, there's like 10 of them. I can't even keep track.
Manik Surtani: 10. You're- Yeah ... you're lucky if they're 10. I mean, 10-
Matt Klein: Yeah, right
Manik Surtani: maybe 10 open source ones.
Matt Klein: Right.
Manik Surtani: There's so many of the proprietary ones, and then custom-made ones that people run-
Matt Klein: Right
Manik Surtani: ... in-house. Yeah.
Matt Klein: Yeah. I, I, I mean, do you, do you think that over time these things are going to consolidate e-either in, um, either in features or just in the number of different solutions?
A-again, this comes back to what I was asking before, which is in the new world where it's so easy to customize things and kind of-
Manik Surtani: Yeah
Matt Klein: ... do your proprietary fork, um, it-it's, it's just so unclear to me where we land, you know, in terms of what are the pieces that we actually share.
Manik Surtani: I, I [00:42:00] think, I think routers and gateways are particularly interesting because it's not all just about software in those cases.
A lot of it is about who's actually hosting it as well. And there's, um, um, that, that's-- that suddenly makes things really interesting. Suddenly there is a, is business viability to doing that. Like we saw what OpenRouter sold for, for example.
Matt Klein: Right.
Manik Surtani: Um, all of a sudden, being the, being the piece of infrastructure in between your model provider and, and your agent is a valuable space.
Matt Klein: Yeah. A-and then I don't know how much you're able to talk about this, but, um, from the foundation perspective, is this a worldwide effort? Do you have participation from the organizations within China as well, or is it mostly a US, Europe-based effort right now?
Manik Surtani: So it started out as mostly a US effort, not even Europe.
Matt Klein: Yep.
Manik Surtani: Um, and, um, um, well, that, that has changed a lot since. So we do have now a lot of European companies involved as well. Um, we've started to see uptick in [00:43:00] China and Japan-
Matt Klein: Yeah
Manik Surtani: ... and Korea as well, um, which has all been really positive, positive signals. Um, we fielded a few project rec- project submission requests coming out of China, which I'm very excited about.
Matt Klein: Great.
Yeah. Um,
Manik Surtani: nothing I can talk about right now-
Matt Klein: Yep
Manik Surtani: ... but, you know, we would announce them soon. Um, yeah, I mean, they also are very active in our working groups. They're active helping figure out-
Matt Klein: Right
Manik Surtani: ... what the, where the gaps are and how do we solve for those gaps, which, which I'm very excited about.
Matt Klein: Yeah, I, I, I mean, again, like I'm gonna show my bias and, uh, you know, again, I don't know what you're willing to say, but I would imagine from your background, you, you probably agree with me that it just, it seems like this area is, to me, is ripe for open source takeover.
I mean, it, it just seems to me that so much of the infrastructure, whether it be the models, the tooling on top, all of these things, it's, it's just gonna evolve towards being open, um, because to me, that, that just makes a lot more sense. But-
Manik Surtani: Yeah
Matt Klein: ... I, I think like any other huge [00:44:00] category, you know, obviously there's a lot of competing concerns right now between, you know, a lot of proprietary venture-backed companies and obviously venture-backed open source and all of these things.
So it's just, it's such a, such a fast-moving space right now. So again, like I, I don't know what you're willing to share, um- My question, if I were to ask you a question, is how do you view the space right now in terms of, you know, both the amount of venture investment that's coming in, the closed source companies, the open source companies?
Like how, how do you, how do you view the ecosystem right now?
Manik Surtani: Yeah, it's a really interesting one. Um, right now I don't see a lot of friction in the sense that, uh, not in the agentic space at least, because a lot of companies still sort of look at value capture elsewhere, whether it's in building or hosting models or, uh, or the inference layer and things like that.
Matt Klein: Yeah.
Manik Surtani: Which means that all the stuff on the agentic layer, they're willing to make that [00:45:00] community first. In fact, it acts as a top of funnel for them, for their businesses. And so it's, it's a lead generator, things like that, right? So great. It suddenly means that I'm actually on a fairly easy pitch right now-
Matt Klein: Yeah
Manik Surtani: on the agentic AI space. The harder challenges are going to be the models themselves or the hardware layer, which I mentioned earlier, touched on briefly.
Matt Klein: Right.
Manik Surtani: And which is again, where I think they should be separate foundations because right now we've got lots of tailwind on the agentic layer. If we started to touch models as well, those tailwinds go away.
Matt Klein: Yeah. Um, I'm gonna ask you a very tough question now, which is that do you, um, as you're doing the foundation, do you think about the, the impacts of this technology on society and the world, or, or do you focus mostly on technology? I, I, I would just, again, just like love to learn how do you think about it, and again, you're welcome to answer that Either for your personal self or from the foundation perspective.
I, I, I just am curious, like, how you tease [00:46:00] apart that complicated topic-
Manik Surtani: It's really hard.
Matt Klein: ... from, from the, from the work of the foundation itself.
Manik Surtani: Yeah. There's a lot of this, this constant tension going on in my head between the multiple facets of, of that, of that statement. There is all of the wonderful benefits that we can see at, that, that, um, agentic AI, that AI will bring to the world, um, and that's incredible.
How do we make sure those benefits accrue to everybody-
Matt Klein: Right
Manik Surtani: ... and not just to a couple of big companies, right? Well, that's really hard, and that's sort of where I think open source has a really important role to play because it gives people freedom of choice, it gives people the ability to switch and change.
That's where open protocols help, right? Like, you can swap pieces around.
Matt Klein: Yeah.
Manik Surtani: Um, like things like, uh, routers and gateways, again, right? Really important piece in that overall, uh, stack to let you switch from one model to the next, to give you that freedom of choice. That's what holds companies accountable at the end of the day.
Um, but I mean, that's one part of it. Is it actually good for human beings and for society? Well, [00:47:00] there's a lot of back and forth on, you know, what is, what is actually good, right? Like, so yeah, um, um, definitely conflicted and often conflicted. Um, I think there's a lot more that as a society we need to do-
Matt Klein: Yeah
Manik Surtani: particularly around income distribution. What does that start to look like? Getting ahead of thinking about that. There are people far more qualified than me and far smarter than me who have been thinking about such things in academic circles, but how close is that to, to operationalizing or even experimenting with in small cases?
Matt Klein: Yeah. I, I, I, I mean, I wouldn't sell yourself short. You're obviously a very intelligent person, and it, it's, um... I just think it's a fascinating topic because, a- again, I'm just speaking for myself, is that I, I think, you know, I think the technology is real. I mean, I, I, I think it is going to change a lot of things-
Manik Surtani: It will, yeah
Matt Klein: about how, how we do a lot of things. I am fearful about its impact on society, you know, especially around-
Manik Surtani: Yeah
Matt Klein: ... job loss and job displacement, [00:48:00] and, like, a lot of those things. And, um- It just s- strikes me that y- you are at an influential position right now, you know, in terms of navigating a lot of this technology, and I, you know, I, it just, it just seems like something that would potentially come up.
Manik Surtani: Yeah, I mean, like I've got, I've got a hand in how the technology evolves, but not how-
Matt Klein: Yeah
Manik Surtani: ... um, the value accrues to people. Right. And that's the bit I think that as a society, we need to spend a lot of time thinking about.
Matt Klein: Yeah.
Manik Surtani: But I think the role of government changes completely in, in-
Matt Klein: Right
Manik Surtani: ... uh, in, in an AI first world, because a lot of what government does today will probably not be necessary.
And the only one thing that it does today which will be necessary is redistribution of wealth.
Matt Klein: Yeah.
Manik Surtani: And that needs to be, you know, the most important thing that, that governments think about moving forward.
Matt Klein: Yeah. I, I, I mean, it's, it's, it's, it's hard to wrap my head around, like all of the complexity of-
Manik Surtani: Yeah
Matt Klein: of this issue, for sure.
Manik Surtani: That's right. And more than the complexity of what the [00:49:00] answer might be, I think the hardest part's the stepping stone, the, the transition of going from where we are today to that future world of, of AI abundance and redistributing-
Matt Klein: Right
Manik Surtani: ... all of that AI abundance to everybody.
Matt Klein: Sure.
Manik Surtani: And that transition is gonna be bumpy, I, I'm afraid.
Matt Klein: Well, and the other thing that I'm personally worried about is that especially in this intermediate time of us learning how to use this technology, and again, I don't know about your own experience, but my experience, at least on the engineering side, is that when you hand these tools to very experienced engineers, they are a superpower.
Manik Surtani: Yes.
Matt Klein: Um, less clear to me what happens when you hand them to people that don't have a lot of experience. And my, my big fear right now is we, you, I, some other people, we're gonna retire not that long from now, right?
Manik Surtani: Mm-hmm.
Matt Klein: And then what happens? I mean, it, it-
Manik Surtani: Right
Matt Klein: ... it's not, again, it's not clear to me what, what ends up happening next.
I don't
know.
Manik Surtani: Yeah, yeah. Absolutely, absolutely. [00:50:00] Um, there is that side of things. But I'll tell you what my biggest fear is, well, not necessarily my biggest, but, you know, uh, an interesting one I like to talk about is assuming that we get through the first set of hurdles, right? Like it doesn't destroy us all.
Matt Klein: Sure, yeah.
Manik Surtani: We found a way to redistribute wealth. Right. We're in this utopian place after that, right? Um, like how the heck do we motivate ourselves to do anything?
Matt Klein: I, I, uh, I agree, and I'm gonna say something right now that I've been thinking about recently, is that I s- I still do a lot of engineering as a part of my job, and the tools have gotten so good in terms of their code output that they mostly do what I want.
I guide them, I do a little code review. I'm gonna be honest that like the coding part of the job for me is kind of boring at this point.
Manik Surtani: Right.
Matt Klein: Like, it's not really interesting anymore.
Manik Surtani: Yeah.
Matt Klein: And part of my brain is atrophying. And yes, I understand that people will say that you're supposed to step into higher level thinking, and of course-
Manik Surtani: Right, right
Matt Klein: I'm doing that too, [00:51:00] but I'm, I'm just agreeing with you-
Manik Surtani: Yeah
Matt Klein: ... that I, that I think that we have the potential of losing some of the things that make us human and actually make us think. I mean, we're now having like a very-
Manik Surtani: Well, that's the stakes,
right? I
mean-
Matt Klein: ... deep, deep conversation, but yeah.
Manik Surtani: I mean, I mean, we are problem solve- we're a problem-solving species-
Matt Klein: Right
Manik Surtani: by nature.
Matt Klein: Yeah.
Manik Surtani: And if there aren't any problems to solve, what are we then?
Matt Klein: Yeah. Okay. Well, um, on, on that topic, since we're almost out of time, um, I mean, what, what is your outlook over the next five to 10 years? Do you, do you think that we are going to solve these problems? I mean, I, I don't, I don't necessarily wanna end the show on like a-
Manik Surtani: Yeah, yeah
Matt Klein: like a, like a doomsday-
Manik Surtani: On a downer
Matt Klein: ... prediction. Um, but you know, I would like to hear your honest opinion- Okay ... given that, given that you're in this-
Manik Surtani: 10 years is a
very
long way out.
Matt Klein: Sure, yeah. Well, five.
Manik Surtani: 10
years is a very long way out.
Matt Klein: Five. Three.
Sure. Yeah.
Manik Surtani: But I, I think what, what, what's, what, what I think will happen is we're For, for all of our faults as a species, we're also a very resilient species.
Um, we tend to address things too late, but then once we [00:52:00] do realize something is that we do tend to come back quickly and, and, um, you know, we're, we are resilient. So I think what'll happen is it'll, it'll be bumpy. I think it's gonna it's, it's, it's gonna be a painful transition, but we are gonna come out of it in the other side in a, in a techno-driven utopia of sorts of abundance.
Um, I think we'll be on the, on, well on our track to curing, you know, uh, human diseases. We're well on our track to solving for climate change and, uh, the energy crisis and things like that, right? Um, and I think what... I mean, the bit that really fascinates me is, is what are we gonna do? Like, you know, and what does that mean?
Do we, do we become these creative beings of leisure?
Matt Klein: Right.
Manik Surtani: And what does that look like? How do you, you know? I mean, that, that's a fascinating thought experiment. What does that look like?
Matt Klein: Yeah.
Yeah. I, I mean, w- we should probably end on that up note. Yeah. Um, anyway, thank you. Um, this has been a fantastic conversation.
I think people are gonna find it very interesting.
Manik Surtani: Awesome.
Matt Klein: [00:53:00] Um, that's a wrap for this episode of "Beyond the Noise: Signals, Stories, and Spicy Takes." Huge thanks to Manik for joining and sharing his story. You can find this episode and all past ones on the bitdrift YouTube channel. If you had fun, drop us a review, tell your friends, or yell your favorite hot take into the void and just make sure to tag us.
I'm Matt Klein, and we'll see you next time. Thank you so much.

