What Happens When Every Developer Has 20 AI Agents

Demetrios: [00:00:00] If I have 5 or 10 or 20 agents that are working for me full-time- Mm-hmm ... and I'm one person, now you extrapolate that out- Enjoy ... by the millions of developers that are out there. Yep. Where are we bottlenecking? Uh, everywhere. Stephen O'Grady is the co-founder and principal analyst at RedMonk, where he spent over 20 years tracking the technology shaping software development.

Demetrios: His research spans open source, cloud computing, AI, and developer tooling, making him one of the industry's leading voices on where software is heading next. Could he tell us which trends are real, which are just hype, and what developers should actually be paying attention to? I get the feeling you don't subscribe to this overly simplistic idea of SaaS is dead.

Stephen O'Grady: I do not. Uh, I find probably the notion kind of laughable, in fact. There was a phrase for many years, it was, um, you don't adapt SAP to your business, you adapt your business to SAP. But what you figure out is that [00:01:00] customizing the software of your business comes with some pretty significant costs. I think we're still figuring out the paradigms here, right?

Demetrios: Okay, let's talk about some of the research you've been doing. What, what have you been up to?

Stephen O'Grady: Uh, we have been, I think like a lot of people, trying to sort of understand as best we can, uh, the implications of AI on developers, on other practitioners, on organizations, on open source. Uh, and, you know, I think the, the sort of clearest takeaway is not gonna come as a surprise to anybody who's tracked the field or used the tools, is that things are moving very, very quickly.

Demetrios: Yeah.

Stephen O'Grady: So that makes a lot of work for us because it's, you know, your, your, you know, conclusions you come to one day are maybe different, you know, sort of the next day. So a lot going on.

Demetrios: What are some key signals you've been seeing?

Stephen O'Grady: So key signals, uh, I [00:02:00] think the, a couple things I think to, to think about, um, so we're here at the MCP Dev Summit, right?

Stephen O'Grady: So I, I ran some numbers before I came down, uh, looking at, uh, downloads of MCP, um, related software, the SDK primarily. Um, the growth in that was pretty extraordinary. You know, it was a kind of a hockey stick going up and to the right. Um, you know, with the caveat that that's just downloads, right? That doesn't necessarily imply, uh- Production use

Stephen O'Grady: sort of actual usage- Yeah ... and deployments and, and so on. Um, the sort of flip side of that coin is, is that if you, you know, with technologies that sort of init- you know, experience an initial surge of interest, um, you tend to see sort of a lack of deployments in that the downloads, you know, sort of ultimately rise and then crater, uh, as people figure out like, "Oh, hey, there actually isn't that much use here for me."

Stephen O'Grady: Uh, that hasn't happened in the data that we've seen yet. Um, so we can't prove that it's sort of being in use everywhere, but is, uh, like I said, you know, at least the consumption [00:03:00] of, of the, um, uh, MCP related software is, is sort of up and to the right. Wait, what are some past experiences that have- Oh, uh-

Stephen O'Grady: done that? Probably one that comes to mind, we didn't sort of track it much because we were sort of convinced that there wasn't a there there, was, um, the Web3 stuff.

Demetrios: Uh-huh. Um,

Stephen O'Grady: so Web3 saw a ton of sort of initial interest, and then people figured out, "I don't actually have much use for this." And so the, uh, that initial surge of interest, uh, in many cases just sort of, um, petered back out- Mm-hmm

Stephen O'Grady: right after, uh, sort of a lack of, just a lack of sort of immediate, um, applicability, you know, or use cases. So, uh, so that's one thing that we've been looking at, um, uh, trying to measure, uh, you know, sort of the As best we can, you know, sort of via a variety of sources, you know, sort of the uptake of, of, um, uh, various other open or sort of AI sort of related packaging, um, sort of across a wide variety of, of software ecosystems.

Stephen O'Grady: [00:04:00] Um, again, it is just things are moving so quickly. There's tons of it. Um, you know, and, you know, obviously as, as, you know, folks who study what practitioners do, we use these tools ourselves, you know, quite a bit. Uh, so I was joking the other day that the, uh, uh, you know, your, your tokens are exhausted, um, you know, come back at 2:00 p.m.

Stephen O'Grady: is kind of the, you know, sort of the new, uh, wait, I have to wait for my compile to finish. Uh, so there's many cases where we're sort of waiting on, on, uh, our, our tokens to refresh, you know, sort of as it were, as we use these tools. Um, but yeah, I think that the biggest-- I think the biggest takeaways are things that we've known sort of in, in ways for a couple of years now, which is that, uh, the tooling has the ability to generate software at high velocity, sort of on an individual basis.

Stephen O'Grady: Um, and the implications for that are vast, right? So the implications for that on an individual basis, right, if I can output, you know, some multiple of the software that I used to, uh, what does that mean for me as an [00:05:00] individual? Uh, what does it mean for the infrastructure that I rely on? Um, what does it mean for the businesses, you know, sort of that, uh, you know, developers work for?

Stephen O'Grady: Um, and what does it mean for sort of other ecosystems like open source, right? Um, so yeah, it's a, um, it's a, it's a brave new world with a lot going on.

Demetrios: I get the feeling you don't subscribe to this overly simplistic idea of SaaS is dead.

Stephen O'Grady: I do not. Uh, I find, I find the notion kind of laughable, in fact. Um, so before becoming an analyst, I used to be a systems integrator, and part of my job was to take various pieces of packaged software, um, CRM, portals, and sort of all these sorts of things and implement them.

Stephen O'Grady: And one of the things you figure out pretty quickly is that if you sit and ask a business to describe how they want an application to function, uh, it takes forever, 'cause everyone has an opinion, uh, everyone wants their sort of little, you know, sort of a [00:06:00] workflow sort of included. And so, you know, for many of the SaaS, um, offerings, you know, kind of the, SaaS is kind of like the I don't know.

Stephen O'Grady: Th-there was a phrase for many years, it was, um, "You don't adapt SAP to your business, you adapt your business to SAP." Yeah. Uh, and, you know, on the one hand, that seems, you know, kinda, um, backwards, right? You- Mm-hmm ... think you, "Hey, I wanna customize the software for my business," but what you figure out is, is that customizing the software for your business is-- it comes with some pretty significant costs, right?

Stephen O'Grady: Um-

Demetrios: And for what value?

Stephen O'Grady: For what value, right? Um, but it makes hiring more difficult.

Demetrios: Mm-hmm.

Stephen O'Grady: Right? So, in other words, I can go find somebody off the street that knows Workday or Salesforce or, um, SAP or whatever. Uh, I can't hire somebody off the street kn- who knows my bespoke ticket, CRM, ERP- Mm ... um, you know, HR system.

Stephen O'Grady: Uh, it also, uh, to me, places too much emphasis on the software as the only asset, [00:07:00] right? So, if you're Workday, as an example, and you are responsible for things like HR compliance and geographies all over the world, do you really wanna have to recreate that from scratch? Probably not. Um, so look, I think, you know, AI will have, uh, certainly an impact on, um, you know, various SaaS businesses, but the whole SaaS is dead, you know, to me is a...

Stephen O'Grady: To me, that suggests that, uh, the people saying that are people who think that the only thing that goes into these equations are the software, and that is, in my experience, not the case.

Demetrios: For me, it feels much more nuanced. It feels like you have certain chunks of a workflow that don't need to be part of a greater SaaS solution, and those you can claim as your own But that's not the whole story.

Stephen O'Grady: Well, and also I think a lot of it depends on, you know, look, if, if a SaaS can't give you what you want, right? If you are sort of a, [00:08:00] a outlier, right? For whatever reason, okay? You know, AI can be a, uh, a good way to, uh, you know, present alternatives. But in many cases, do you really need to reinvent the CRM wheel or the ERP wheel?

Stephen O'Grady: Probably not. Now, all that being said, and what I, I am certain you will see, and, you know, we're seeing it a little bit, uh, already, is that, you know, AI is used as the proverbial Amdahl mug, right? So the Amdahl mug was, uh, back in the day when, uh, IBM sold mainframes, right? Um, they used to be able to come in and dictate terms to their customers 'cause it's like, "We're IBM, we're the only place you can get these mainframes."

Stephen O'Grady: And Amdahl comes along and makes a, a, uh, a competitor product. And so even if they had no intention of switching, what they would do is when the IBM salespeople came in, they would prominently have an Amdahl mug sitting on their desk, right? You know, to remind them like, "Hey, I have leverage from a negotiating standpoint."

Stephen O'Grady: Yeah. So I'm certain it will be used that way, um, but [00:09:00] I don't buy the wholesale, you know, sort of replacement niches. The cost to me, you know, significantly outweigh any potential benefit.

Demetrios: Well, for me, there are a lot of use cases that I find fascinating when Folks can stand something up faster-

Stephen O'Grady: Mm-hmm ...

Demetrios: than they can get a tool through procurement.

Stephen O'Grady: Mm-hmm.

Demetrios: Yeah. So that's like the opposite end of the spectrum where you say, "Okay, these Workdays and Salesforces, those have been in my company since before- That's right ... the beginning of time."

Stephen O'Grady: Yep.

Demetrios: We're not going to really worry about that, but for a new tool-

Stephen O'Grady: Mm-hmm ...

Demetrios: you really have to sh- have a strong value prop as to why I'm gonna use that and go through all the headache of taking six months to get it through my procurement team.

Stephen O'Grady: Yeah. Yeah, and as the person at our company who deals with procurement, I am sympathetic to that stance. I think the flip side of that is that, um, you know, when I've had these conversations with folks, they're like, "Hey, what if I could take my licensing costs, you know, to sort of make [00:10:00] them zero? And what if I could do X, Y, and Z?"

Stephen O'Grady: And it's like, right, but then you have to operate it.

Demetrios: Mm-hmm.

Stephen O'Grady: Right? And so there's a whole lot of, um, uh, a, a number of sort of very large banks, as one example, uh, you know, sort of a decade or more ago figured, "Okay, you know what? I don't wanna pay, uh, you know, sort of a, a platform tax, so I'm gonna write my own, um, essentially developer portal software."

Stephen O'Grady: And w- what they found then is, is that now you, you own that, right? You are responsible for keeping it up to date, you're responsible for securing it, you're responsible for operating it. Uh, and a number of these banks have subsequently come back and said, "You know what? This was not the best use of our time, 'cause this is not differentiating for us," right?

Stephen O'Grady: "Us being able to host applications doesn't give us a competitive advantage over other organizations." And so, you know, I think you'll see, you know, sort of in some cases, you know, companies that, that sort of, uh, get frustrated with procurement and are like, "Okay, you know what? I'm just gonna go build this."

Stephen O'Grady: [00:11:00] Then realize 6, 12, 24 months down the line, like, I don't really wanna be in the business of operating this, and it's not really giving me sort of anything, um, uh, you know, I couldn't get sort of easily and more, uh, more easily and, and cheaper elsewhere. What

Demetrios: I saw, which was very eye-opening for me, is a company that I feel is very advanced in the way that they're doing agents.

Stephen O'Grady: Yep.

Demetrios: They have had the talk to my data use case since basically when ChatGPT first launched. Yeah. That was one of their first use cases that they set up. They've been tuning it for the past three years. Mm-hmm. And it is very well done. They just hit me up the other day and said, "Can you tell me all of the talk to your data vendors, because I think we're ready to outsource it?"

Stephen O'Grady: Yeah. And I, I think you'll see a fair, a fair amount of that because it's... Look, it, it's easy to, um, you know, I, I tell this story all the time now, right? So, uh, you know, up [00:12:00] until November, uh, when, you know, the first of the new models arrived, right? In, um, uh, Opus 4.5, and obviously, uh, uh- 4.6,

Demetrios: right?

Stephen O'Grady: 4.6 now. Yeah.

Stephen O'Grady: 4.5, I think, was the one that was released in, in, um-

Demetrios: In

Stephen O'Grady: November ... uh, late November.

Demetrios: Yeah.

Stephen O'Grady: So we had this inflection point, right? There's a before and there's an after, and the tools were very sort of useful before. They would do interesting things. In the wake of that, oh my God, it's just a night and day difference, right?

Stephen O'Grady: So the canonical example I point to there is a LA Dodgers fan who happens to do the, uh, website design for CNN. So he built this, uh, essentially data analytics portal for the Dodgers and released it as open source software. So I've always thought like, "Oh, hey, I should port this over for the Red Sox," and so on.

Stephen O'Grady: Um, and just never found the time to do it, and I would always throw, you know, all of the tools like Hero and, um, Cursor, Windsurf, uh, Antigravity, all these things at it, and they'd, they'd get most of the way there, part of the way there, and then choke and sort of stall out. And [00:13:00] then, uh, over the holiday, like a lot of folks did when Anthropic was giving people sort of more credits- Mm

Stephen O'Grady: uh, I threw Claude at it and it worked, and it did all the things it needed to do, which was great. The flip side is, is that after realizing this, it's like, oh, okay, you know what? I want a dashboard for this, and, uh, you know, I want an application that's gonna go do that. So I ended up building, I don't know, a dozen, you know, different things, like little small toy apps.

Stephen O'Grady: And you quickly... This is just on an individual level. Mm-hmm. Right? You get to the point, it's like, right, wait, which one was that, and where did I put it, and what is that actually running on? What...

Demetrios: Yeah.

Stephen O'Grady: I don't remember where it's hosted. What

Demetrios: conversation was I having- Yeah ... with Claude?

Stephen O'Grady: I have-

Demetrios: Where's the context

Stephen O'Grady: here?

Stephen O'Grady: I have no idea where we're at, and that's literally on an individual, like-

Demetrios: Yeah ...

Stephen O'Grady: you know, sort of, uh, uh, very amateur basis. And so- Yeah, I think when, uh, a lot of enterprises figure out like, "Oh, I can start churning these things out," I [00:14:00] think there'll be this burst of churning things out and then after a little while it's like, mm, okay, this one is real and, and sort of new and differentiated and I wanna, I wanna keep this.

Stephen O'Grady: But in many cases it, you know, they're gonna find, you know, the cost of, of sort of operating, securing, protecting, maintaining, you know, continuing to develop, um, you know, these one-offs is not gonna be worth it over

Demetrios: the long term. They're gonna start throwing around the sexiest word in the English language, governance.

Stephen O'Grady: Yeah, right.

Demetrios: Yeah.

Stephen O'Grady: Exactly. Yeah. So-

Demetrios: But I, I-- You know what I noticed when playing with it over the holidays is I had this theory of, oh wow, software is becoming local first-

Stephen O'Grady: Mm-hmm ...

Demetrios: as opposed to I'm going to create an app and then expose it and expect scale. Yep. It's like, no, I'm just gonna have my own things and help my own workflow.

Demetrios: But then it went from me going software is local first now to I've fully pivoted and now I'm like, but why do I even need an app if I can just build [00:15:00] a skill-

Stephen O'Grady: Mm-hmm ...

Demetrios: that can do it? So I don't need the overhead of an app or even a local first app-

Stephen O'Grady: Yep ...

Demetrios: is not It's too much to worry about. Yeah. Let's just do it with a skill 'cause it's the lightest- Mm-hmm

Demetrios: thing that I can think of.

Stephen O'Grady: Yeah. Yeah, and that's the thing. I think we're still figuring out the paradigms here, right? Right. Um, we're still figuring out, you know, to your point, like what do we need an app for? What does that app look like? Um, do we need an app at all? Do we need a skill? Do we need, you know, sort of some, you know, sort of stub or approximation like...

Stephen O'Grady: So, um, I mean, I, I think that the biggest thing is, is that the sort of as we talked about at the top, you know, when you take the cost of producing code and drive it, you know, maybe not quite to zero, but near zero, um, we're gonna have to rethink a lot of different things, right? Yeah. We're gonna have to rethink, you know, what we do individually.

Stephen O'Grady: We're gonna have to rethink the infrastructure we rely on. Businesses are gonna have to rethink what they build, um, you know, because it's, uh, I'm trying to remember the context in which I brought this up from a research standpoint, [00:16:00] but one of the things that we see, um, you know, we've been through this sort of, you know, multiple ways where we've taken something sort of relatively monolithic and broken it up into pieces.

Demetrios: Hmm.

Stephen O'Grady: Uh, so virtualization's probably the simplest example of that, right? You know, we had physical machine, you break that up into sort of some multiple of that virtual machines, and one of the things you figure out is, is that that comes with benefits, but it also comes with costs, right? And, you know, I talked to one of the first large banks that moved over to, to the VMware and said, "You know, hey, how's it going?"

Stephen O'Grady: And they were kind of lukewarm. And I thought, "Oh, is this, you know, a problem with the software? Is this an issue?" And he said, "No, no. Like look, we've hit all the marks that we want. Our utilization rate is way up, you know, the, the, the appointment density is way up." And I was like, "Well, what's the but?" Yeah. He said, "You know, look, you know, managing, you know, 500 physical machines is one thing.

Stephen O'Grady: Managing 5,000 virtual ones is a different thing," right? And so I think what we're gonna see for a lot of these organizations is, you know, if you now have 10X more applications to manage, that is going to be a special challenge of its own. [00:17:00] Yeah.

Demetrios: Or even I know folks are talking about that with agents.

Demetrios: Mm-hmm. I've now got 5,000 agents- Yep ... that our company uses or 30,000 agents.

Stephen O'Grady: Mm-hmm.

Demetrios: It sounds like they're testing that and they're running fast and they wanna do that. Yep. But then later they'll recognize maybe it's not worth it.

Stephen O'Grady: Yeah. Yeah, I, I think that's right, right? And it's, um, you know, we're in that initial burst of creativity where we're just creating things because we can and, and because it's easy and it's frictionless and so on.

Stephen O'Grady: And, you know, we're gonna get to a point where, you know, we, we have to reckon with that- Yeah ... volume, right? And if we're not there yet, um, you know, I think many individuals are. Certainly I am. I'm like, "Oh my God, where is..." As I said, "Where is all this stuff and where do I put it?" Um, organizations are gonna get there pretty quick.

Demetrios: Yeah. So then tho- that's a very clear signal. Man, what other stuff have you been seeing in the research?

Stephen O'Grady: So one of the things that, uh, is coming up in sort of various quarters, um, [00:18:00] is the strain on infrastructure, right? So, you know, when- You know, you think about individual developers, right? And you think about things like, uh, projects like Gas Town, right, from Steve Yegge, right?

Stephen O'Grady: Which is like, okay, you know, we have a team or swarm or pick your term of choice of, of individual agents, right? So if one developer suddenly becomes overnight a, a, um, essentially a swarm of dozens of developers, like what does that do to our infrastructure? Um, and is our infrastructure built for that?

Stephen O'Grady: And the answer is no. And so what you're seeing from, uh, you know, things like package sort of management repositories, uh, is that their traffic is through the roof. Mm. And their, uh, costs for simply maintaining things like bandwidth, right, and scale and so on are dramatically up, you know, because there's much more usage.

Stephen O'Grady: Um, and that's without getting into all the other things that go into our infrastructure, right? Like, so, um, you know, triaging, um, PRs, [00:19:00] right? How do I sift through these, right? Um, I was just talking to somebody earlier that apparently the, uh, the time to sort of resolve PRs is up significantly, um, you know, in part because there's just so many more of them.

Stephen O'Grady: Um, so yeah, the, like I said at the, I think at the top, the, the ripple effects from, you know, sort of taking this, this code and making it so much easier, uh, are profound, and we're seeing that, um, sort of manifested in just various strains and, and, um, uh, stresses on our infrastructure, you know, that we rely on.

Demetrios: Yeah. A- and which different parts of the infrastructure? Because I instantly think if I have 5 or 10 or 20 agents that are working for me full time-

Stephen O'Grady: Mm-hmm ...

Demetrios: and I'm one person, now you extrapolate that out- Exactly ... by the millions of developers that are out there.

Stephen O'Grady: Yep.

Demetrios: Where are we bottlenecking?

Stephen O'Grady: Uh, everywhere, right?

Stephen O'Grady: So everything from code review to build systems, uh, [00:20:00] production deployment, um, security review, right, I mean, pick it, right? A- any one of these things is getting swamped. Um, and you see this in open source, you know, where maintainers are just flooded, uh, with, you know, essentially AI slop- Slop code ... PRs. Yeah.

Stephen O'Grady: Um, and so there really isn't sort of an aspect of infrastructure that's not, um, you know, sort of being stressed, you know, by the volume, you know, that we're seeing. Uh, and- You know, sort of AI can help with some of this, right? So in other words, like, um, you know, there are companies doing, uh, AI security review, there are companies doing AI, um, uh, QA and software review and all these sorts of things.

Stephen O'Grady: But it's gonna be, you know, it's gonna be a bit, right, before, you know, those technologies are sufficiently mature to, to really help offset the stresses here. So honestly, at this point, you know, really the only thing that's gotten a lot easier is generating the code. Yeah. Everything that's downstream of that is, is more complicated.

Demetrios: Mm-hmm. Yeah, the code reviews, I, I have [00:21:00] not been sold on these PR code review type of agents because it feels like just a different prompt and a different sub-agent. Yeah.

Stephen O'Grady: Yeah. Well, and it's, you know, you see people trying to get around it. It's like, uh, I was talking to somebody last week who they just pit, um, the agents against each other, right?

Stephen O'Grady: Almost like a, a Ralph loop, you know, the Ralph Wiggum loop, um- That's right ... uh, came out, I guess it was over the holidays, I think. Um, but anyhow, they'll, they'll basically bounce, you know, sort of code between two different agents and, um, you know, from, uh, two different models from two different vendors to try to, you know, sort of get some, some, I don't know, outside objectiveness sort of as it were.

Stephen O'Grady: Um, but yeah, it's, it's, uh, it's, it's like I said, it's gonna be a bit before we, you know, digest how all this stuff is gonna work together.

Demetrios: Yeah. You did mention the package managers getting flooded.

Stephen O'Grady: Mm-hmm.

Demetrios: And I find it fascinating that We also saw UV-

Stephen O'Grady: Mm-hmm ...

Demetrios: get bought, and then Bun also [00:22:00] got bought. Yep. And do you have any thoughts on that?

Stephen O'Grady: Uh, yeah. So, um, you know, I think the short version is, is that, you know, what you're seeing in terms of the acquisitions are people, you know, who want to control, you know, sort of critical pieces of infrastructure, right? And-

Demetrios: Choke points.

Stephen O'Grady: Yeah, right. Yeah. Um, and, you know, I think, you know, that makes sense.

Stephen O'Grady: You know, we'll, we'll see sort of how that plays out and, you know, whether that, you know, becomes a problem in terms of, you know, sort of how some of the projects are sort of developed, manifested, shared, et cetera. Um, but on the package management front, um, you know, really I think if you talk to any of the sort of major pack, um, you know, package repositories, you know, they're seeing significantly, um, you know, more traffic, right?

Stephen O'Grady: Um, and it, it's an inevitable consequence, right? If you, you know, as we, we were talking about a moment ago, if you take one developer and turn him, him or her into 20 developers virtually, right? Um, the systems are simply not equipped. They were already straining, you know, sort of with the [00:23:00] demand pre-AI.

Demetrios: Hmm.

Stephen O'Grady: And now you have, you know, sort of this additional multiple, you know, sort of applied. It's, yeah, like I said, it's inevitable in terms of how that plays out.

Demetrios: Mm-hmm. I have, uh, some questions from our mutual friend Alex that I wanted to pull up.

Stephen O'Grady: Okay.

Demetrios: And he was asking about-- We kinda answered this one. You know, RedMonk tracks developer adoption patterns better than almost anyone.

Stephen O'Grady: We

Demetrios: try. What signals are you seeing around MCP adoption? Is it following the trajectory of early integration standards like REST or GraphQL? Or is something different happening?

Stephen O'Grady: Yeah. So, uh, the way I'll answer that is that, um, I posted, this is back, looked it up, uh, before coming down here. I posted on Blue Sky in May of last year that, uh, in, in our experience at RedMonk, uh, we've been doing this for a very long time, um, MCP was the fastest project to become a de facto standard, uh, that we had seen.

Stephen O'Grady: Mm. [00:24:00] And, uh, Kelsey Hightower, uh, had replied, "No, no, I, I think it's Docker." And obviously a lot of it depends on sort of what metrics you're looking at and, and, and so on. But I went back, and for us, Docker had been our f- our sort of fastest growing, uh, de facto standard, and that took about 13 months. Um, MCT- MCP took about 13 weeks.

Stephen O'Grady: And so the sort of rise was meteoric. Um, and I have, you know, like I said, I've been doing this a while. I cannot remember another time when, you know, it was a period of, like, two to three weeks from our first hearing about it in, must've been in November, October, November, I guess, of '24. I think that's right.

Stephen O'Grady: And then, you know, within less than a month, talking to multiple vendors who are like, "Yep, this is the standard." I'm like, "I literally just heard about this last month." Like, that doesn't happen. Uh, and so my question- Why,

Demetrios: why do you think that is? Why do you think everybody jumped on board so quickly?

Stephen O'Grady: Well, I mean, I think for, for me, I think it's a simple, um, it's a fairly simple equation [00:25:00] in that the models are in- incredibly capable, but we didn't have good, uh, sort of good standardized ways to connect them to different data sources, right?

Stephen O'Grady: And they are that much more helpful, productive, useful, um, you know, capable, you know, when we can give them access to things that, uh, you know, sort of aren't public in sort of one way or another. Uh, and so, you know, I think-

Demetrios: So there was pent-up demand ...

Stephen O'Grady: pent-up demand. It was like, "Hey, this is great, but I'd like to have this over here."

Demetrios: Mm-hmm.

Stephen O'Grady: So as an example, if you go and look at, um, Google Trends data, all right, and you, you look at data centers as a trend, there's a noticeable spike, um, in December of, uh, 2022, I think it is. Um, and it occurs roughly a month after ChatGPT-

Demetrios: Mm ...

Stephen O'Grady: uh, debuted. And so, you know, who knows what it is, but certainly one potential explanation for that is that people looked at this and said, "Wow, this is incredible capability and technology.

Stephen O'Grady: What if I could have this on site and do it, you know, sort of feed it my data?" Uh, and, you know, I think that was [00:26:00] a lot of the sort of vision and promise for MCP up front is, okay, I can begin c- to connect this and sort of make it useful. And so, you know, my question coming in, uh, you know, to this event was, is that still true, right?

Stephen O'Grady: You know, we, we know the data was sort of where it was, you know, about a year ago. Um, but, you know, you see in, in, um, what was it? February, right? There's a piece that came out, it was like, "Oh, MCP is dead. Long live CLIOps." And so the question was, okay, you know, are we seeing anything in a- uh, in a way that would indicate, you know, sort of a drop-off or, um, as we talked about at the top, right?

Stephen O'Grady: We're not. Like the, the charts, at least at this point, are still sort of up and to the right. Um, you know, they're growing, tons of, tons of downloads and so on. So Yeah, I think the short version is, is that it is incredibly, incredibly, uh, meteoric in terms of its rise, uh, and that's largely driven by, uh, the fact that it is, um, yeah, just giving, giving models additional data and context and sources they didn't have.

Demetrios: I, I, [00:27:00] I want, I wanted to, like, meander down one of these streets, but I think I'm gonna, I'm gonna just leave it because I think that is-- you said everything that needed to be said. Okay. And it's like beating a dead horse a little bit. Works for

Stephen O'Grady: me.

Demetrios: So, um, I'll go on to the next question. The next one that I had from Alex was asking about, in your research, have you seen what folks are actually building with MCP?

Stephen O'Grady: Um, yes. Uh, so I don't-- I'd have to look at the, I'd have to have the graphic in front of me. So folks that are listening to this, if you go to LinkedIn, uh, and look up Stephen O'Grady, my, my LinkedIn page, I posted the graph that I'm talking about. Uh, I don't know, it's, what is it today? Today's April 2nd. It would've been April 1st or March, the last day of March.

Stephen O'Grady: I can't remember which day I did it. Anyhow, um, in there, you know, sort of there's a, a sort of rough breakdown in terms of categories that we're seeing. Um, you know, so most of it's, uh, predictably just sort of generic AI. There's CRM, there's, uh, commerce, there's all kinds of different [00:28:00] things in there. Um, but that's also, you know, sort of a rough assessment of categories from, you know, sort of base software descriptors, right?

Stephen O'Grady: So what we don't have insight into, um, and what's hardest to get always is what are people actually doing in deployment, right? Because not surprisingly, most, most folks don't let us come in and see their running inf- production infrastructure. So we have to sort of, you know, work backwards from, all right, what are we seeing in terms of, uh, patterns of, you know, downloaded software and packages and, and things of that nature?

Stephen O'Grady: Um, but, you know, the, the, I guess the gist of it from an MCP perspective is that there are lots of categories, right? It's not a, "Oh, hey, here are two or three or four or five." I mean, I think we graphed, I'd say probably north of 20- Mm. ... you know, sort of different sort of rough high-level categories of, of, uh, of usage.

Demetrios: Well, it begs the question about the ecosystem around MCP also and what you've seen in that regard.

Stephen O'Grady: Yeah. So the ecosystem, I think, uh, to me the, [00:29:00] the critical step was, um, and this'll sound self-serving 'cause we're at a foundation event, but, um, you know, the biggest issue to me was initially was, okay, it sort of took off as this de facto standard, um, but it's very difficult for, for large vendors to contribute to projects that are wholly owned by another vendor, right?

Stephen O'Grady: Because they are, uh, appropriately concerned about, like, "Hey, am I, why am I sort of underwriting your software development costs, like, you know, when you can on a whim do whatever you want with this project?"

Demetrios: Yeah.

Stephen O'Grady: Uh, so the, the key for me, and I, I frankly took longer than I thought it would, um- Uh, but the key for me was, like, seeing it sort of housed within a foundation.

Stephen O'Grady: Uh, because once you have a foundation, the ecosystem becomes much easier to form, um, because the large vendors who have the ability to throw people at the project are comfortable then working on a, uh, on a project in a, in a, uh, vendor-neutral space. Hmm. Right? And like I said, if that, uh, project stays with Anthropic, you know, um, not that Anthropic can't do a good job of developing.

Stephen O'Grady: [00:30:00] They iterate very, very quickly. They're very good at what they do. Um, but if you want industry support, you know, at some point, um, it has to become a sort of neutral project, uh, because you just can't attract the same type of commercial community, at least, and in many cases, the non-commercial community as well.

Stephen O'Grady: So that was the biggest step, I think, in terms of, um, allowing a, a large ecosystem to form around it.

Demetrios: And how about all of the different So you have companies that are supporting MCP- Mm-hmm ... and the s- the protocol.

Stephen O'Grady: Yep.

Demetrios: But then you have companies that are being built on top of it.

Stephen O'Grady: Mm-hmm.

Demetrios: Have you looked at that type of ecosystem also?

Stephen O'Grady: Not as much. Um, you know, we're, we're sort of more focused on the lens at this point in terms of, all right, you know, sort of what's the base level infrastructure. Um, so in other words, like, you know, just on stage this morning, right, you have the Duolingos and you have just all these companies who are like, "Here's what we're building on, here's how we're using it," and so on.

Stephen O'Grady: Um, those are great and those are [00:31:00] interesting, but in other words, like our focus initially has been on, all right, what is this, um, you know, sort of what's been the growth of this project and what does this look like? Because without a certain level of critical mass, a certain sort of traction, um, you know, the kind of the, the end user use cases are, um, academic because they're not, you're not gonna get there without that kind of, that, that traction.

Stephen O'Grady: So, you know, w- we see some of the sort of common ones, um, you know, that you'd see in, in, um, you know, things like the keynotes this morning, but don't have any, uh, quantitative insight into that.

Demetrios: One thing that the last guest who was here, Jeremiah, just said is the majority of the use cases he's been seeing are internal use cases of MCP.

Demetrios: I wonder if you have any insights into that.

Stephen O'Grady: Um, I'd say that's very likely. Um, you know, because like, you know, particularly in the early stage, I mean, you know, when, when, uh, MCP was really, like there was no auth initially, right? Now it's, you know, there's auth two, 2.1 I think. Um, and you [00:32:00] know, it's the, you know, you heard multiple times this morning, you know, the, the trifecta, right, of external access and, um, you know, sort of internal data, and all these sort of different sort of factors that can contribute.

Stephen O'Grady: So in many cases I think what we'll see, you know, the, the build out will begin internally, right? It'll say, okay, I want to build a system that, um, is not exposing things to the, uh, outside world, um, and I wanna limit the blast radius of any sort of learning, um, uh, experiences that I might have. So everything is gonna be focused initially inwardly, um, and then I'll get to the external piece, um, you know, eventually.

Stephen O'Grady: Um, you know, that being said, you know, I just had a conversation upstairs with You know who, somebody was telling me they have, uh, three external facing MCP servers at their house, right? So, yeah.

Demetrios: To turn on the lights

Stephen O'Grady: or what? I have no idea. I didn't, I didn't get into what they were for. But, um, but yeah, I think, you know, the, the sort of default pattern I would expect is for people to start internally 'cause it's the safer thing to do.

Demetrios: Yeah. [00:33:00] In theory, but then when you delete some databases or you drop tables, then-

Stephen O'Grady: Oh, safer.

Demetrios: Sa- yeah, exactly. Not, not safe.

Stephen O'Grady: There's a difference.

Demetrios: At least... Yeah, you're not ending up on the front page of the internet-

Stephen O'Grady: That's right ...

Demetrios: for something that- Yep ... wasn't the best. Yeah. Yep.

Stephen O'Grady: No, not like, uh, oh God, I can't remember who the, uh, you know, the meta researcher who was, uh, installed OpenClaw.

Stephen O'Grady: Yeah. It,

Demetrios: it was a me- yeah, it was like a meta safety person.

Stephen O'Grady: Yeah, it was one of their, one of their- It was

Demetrios: so ironic

Stephen O'Grady: So one of their safety researchers- Yeah ... was talking about having to run over and unplug their mini. You know, it was like defusing a bomb 'cause it starts- Just deleting everything. Like, "Don't delete my email."

Stephen O'Grady: It's like, "Okay, I'll delete your email," so.

Demetrios: Yeah.

Stephen O'Grady: So yes, you will... It, it's safer- Right ... not safe.

Demetrios: Have you done any research into the OpenClaw ecosystem?

Stephen O'Grady: Yeah. Yeah, I had it stood up for a little bit, and then it's like, yeah, you know what? I'm not, I'm not ready to dedicate the resources that I need to- ... to, uh, to keep this safe and secure.

Stephen O'Grady: Make sure it's safe. Right now I was like, yeah, you know. So I, I was in the fortunate camp. I already had a Mac Mini, so-

Demetrios: Right ...

Stephen O'Grady: I didn't need to, uh, I didn't need to wait on this train. Pay the premium. [00:34:00] Yeah, it was one of these like, "Oh, hey, this is really cool." Then you dig into it a little bit and you see the, uh, uh, the breaches, and I'm like, you know what?

Stephen O'Grady: Like, uh, there, there are people who are YOLOing it, and, you know, more power to you, but I don't have the time to, to worry about that right now. So I'll, I'll wait and, um, I like the vision, I like the promise, and, you know, my, my hope is that ultimately I'll get to something that helps, you know, sort of manage my day-to-day better, you know, much as OpenCloud does.

Stephen O'Grady: But, uh, it's a little, uh, it's a little early in that development process for me.

Demetrios: Yeah. Well, this has been great, man. Yeah, pleasure. I appreciate you coming on here.

Stephen O'Grady: Not at all. My pleasure.