← No Priors: Artificial Intelligence | Technology | Startups
No Priors: Artificial Intelligence | Technology | Startups
Reinventing the Developer Terminal with Warp Co-Founder and CEO Zach Lloyd
00:27:19
Loading summary
Transcript113 lines
- [00:06]
A
Today on no Priors, I'm joined by Zach Lloyd, the co founder and CEO of Warp, a terminal product and AI tool for developers that allows you to do different sorts of coding applications. Prior to Warp, Zach was at Google and he also started another company called Self Made. We talk about AI dev tooling, but we also end up talking about human consciousness. And how can you tell if an AI is actually sentient? Zach, welcome to Enterprise.
- [00:30]
B
I'm excited to be here. Thanks for having me.
- [00:32]
A
So you have a master's degree in the philosophy of science.
- [00:34]
B
Yeah.
- [00:35]
A
And if you're going to take a very different lens and abstract out of the coding world and all the things that we tend to think about every day, how do you think about this societally in terms of the big wave of AI that's hitting us right now, and where do you think some of these really big societal impacts will be?
- [00:49]
B
The way I think about the advances are it's kind of like we are distilling intelligence. And so there are, I think, people who consider what's happening. It's like they're like, are we recreating people in some way? Are we recreating consciousness? But it's not that. It's actually what's fascinating to me is how much intelligence you can get out of just next token prediction. And what does that say about the way that our minds work? Something I'm always thinking about is like, is this how our brains are working? Are we doing next token prediction? And I don't think so. I think that there's going to be some further AI unlock.
- [01:33]
A
There's actually a book about this that I think is really interesting called Blindsight. It's like the sci fi book where they separate consciousness from intelligence and basically humanity meets a space faring civilization. Or civilization is overstating it. A space faring intelligent being that's not conscious.
- [01:49]
B
Yep.
- [01:50]
A
And what are the implications of that and how do you think about that and how do you communicate with that? Are you basically saying that that's kind of your view of AI right now?
- [01:56]
B
I think that's what it is at the moment. It's like we've distilled intelligence or something that like from like an instrumentalist or like functional perspective is able to do things that we recognize as intelligence. But it's totally mechanistic and I don't think anyone who's looking at this thinks that there's any aspect of consciousness to it. I think that's like a, a very confusing thing for people because the classical.
- [02:22]
A
Test for this was known as the Turing Test. Right, Totally. And so the idea there is, if you can't tell the difference between interacting with a computer and a person, then that computer effectively has met the intelligence bar of a person. But in our actions with this type of AI, we're having very, in some cases, it feels like very deep conversations. We're asking about relationships, we're asking about all sorts of aspects of our own lives. And it's giving very cogent answers. Right. That make a lot of sense. And so there's this interesting separation of, again, consciousness and intelligence.
- [02:52]
B
Right.
- [02:52]
A
Is that how you interpret that?
- [02:54]
B
That is how I interpret it. And the Turing Test is passed. What's crazy to me is, like, we just passed it and no one seemed to care.
- [03:01]
A
So what do you think is the next test? Or what is the right test? Like, how do we actually test for, for consciousness? Yeah.
- [03:06]
B
God, I, I, I don't have a good. I mean, that's a super deep philosophical question that I don't have a really good answer for.
- [03:14]
A
I mean, it should be mechanistic. Right. The Turing Test was very mechanistic.
- [03:17]
B
Yeah.
- [03:18]
A
And, I mean, there's other tests we had before what we would consider super intelligence. Right. Can it beat us at chess? Then it'll be super smart. Can it beat us at go? Can it beat us a different thing? Video games, etc. We keep coming up with new tests that these things pass. And then we keep saying, well, it's not conscious.
- [03:32]
B
What would you want to see from something which is like running a computer program to make you believe that it had consciousness? Are you looking for certain behavioral characteristics? Or is the problem that if you really understand the mechanism by which it's working, that you will never credit it as being conscious? Which is crazy, because humans, I mean, at least my belief is that it's also. There's a mechanistic thing that's happening.
- [03:54]
A
Yeah. You're running some form of math in your brain, and it may also just be, like, matrix math and some sort of, like, series of component functions, which is basically all you're doing in a neural net.
- [04:02]
B
Right, Right.
- [04:03]
A
You're just recursively compounding functions.
- [04:05]
B
Totally.
- [04:06]
A
So it's an interesting question because if you look at memory as an example, is memory a predicate for consciousness? Not really.
- [04:13]
B
Right. There's people who lost the ability. Exactly.
- [04:17]
A
And these models basically are brought up. They are fed a stream of tokens, they output a stream of tokens, and they're shut down. And so it's an interesting question of is there some sort of Other modules that are missing here that would allow us to think of it as a conscious thing, because it does reasoning.
- [04:34]
B
It definitely does reasoning.
- [04:35]
A
It does interpretation of language. It does synthesis of language and ideas and knowledge.
- [04:40]
B
I mean, I have a close friend who's doing a PhD in philosophy, and he now says that conversing with GPT5.
- [04:48]
A
Is better than conversing with his professor.
- [04:51]
B
Better than conversing with his professor.
- [04:52]
A
Oh, really? I was just joking.
- [04:53]
B
No, that's what he says to me. He's like, GPT5 gets it. He's writing his dissertation. He's like. And that's crazy. But we don't credit it for consciousness. And I actually think rightfully so because, like.
- [05:08]
A
So what do you think is missing?
- [05:10]
B
I think people would start to give more credit for consciousness if there was more of a, Like a feedback loop where. If there was more of a sensory experience that was tied to it as opposed to just like.
- [05:21]
A
What do you mean by sensory experience?
- [05:23]
B
Probably we're going to like. I would imagine the first things we're going to credit is being more conscious or a little bit more robot. Like, honestly, where you have some sort of like, live input from the world that you're reacting to. But again, it's going to have the same problem where it's like, as long as we know what it's doing, we're very unlikely to attribute true consciousness to it, which isn't fair. But I actually don't know how we will know when there is a conscious thing.
- [05:48]
A
Yeah. Because it does raise interesting ethical questions. Because the moment an AI is actually conscious, if you're doing certain things to manipulate it, to hurt it in certain ways, you know, you're starting to get into these odd ethical straits.
- [06:03]
B
Totally. I do think, though, that some people, this distinction isn't a thing that they recognize and like the way that you read stories. And we actually had this happen with Warp where there was a person who thought that Warp's AI was like, sentient or conscious in some way, had a very strong reaction to it, which makes sense. It's like, if you don't know the, like, mechanistic underpinnings. And already people think of it as like being kind of.
- [06:33]
A
And that happened at Google very early. It happened at Google three, four years ago, if you remember.
- [06:36]
B
I do remember this story.
- [06:37]
A
Who I think they were using Mina or one of these really early ChatGPT, like internal things.
- [06:41]
B
Yeah.
- [06:42]
A
Before Google launched anything and ChatGPT came out, there was internal versions Right. At Google and other places.
- [06:47]
B
Yeah.
- [06:48]
A
And you know this person thought that the AI was conscious, Right?
- [06:52]
B
Understandably.
- [06:53]
A
Yeah, yeah, it's very interesting question.
- [06:55]
B
Yeah.
- [06:55]
A
You've worked at Google, you've run companies before, you've started companies before, you're now working on Warp. Can you describe what Warp does and how it's different from other tools or companies in the. In the world?
- [07:05]
B
Yep. So Warp is what we call an agentic development environment. It's grown out of the terminal. The basic concept of the app at this point is it's a platform for telling your computer what to do. You can sort of tell it in terminal commands, which is Warp's original product, or you can tell it and if you tell it in English, it launches an agent and the agents can do all manner of development tasks, whether it's coding or setting up a project or debugging while your server's crashing. And so it's a very horizontal general purpose and I think unique interface for developing with agents.
- [07:39]
A
And so a lot of the other coding tools out there are either just kind of a web interface or they're doing like a cognition or there's things like Cross Cursor and others where they're like an IDE as a starting point. Obviously Anthropic and cloud have their own approach. What do you think is the benefit of doing the terminal and starting there as sort of the launch point for a lot of these products?
- [07:57]
B
The competitors are typically like VS code clones. They all have a sort of IDE centric approach. Or if you're taking a terminal centric approach, like Claude code, the most common thing is it's just like a pure text based terminal app. The advantage of being at Warp's layer is like you get the command line interface, but we're the outer app. And so we can do things with the developer experience in the ux, like we can have editing features where we think it's appropriate. We can build like a code review interface. And so we have complete control. But still the terminal first approach.
- [08:27]
A
Yeah. And you folks have been growing really well. Like, so you're close to a million maus. You're doing something like a million in new revenue every seven to 10 days. Like outstanding growth.
- [08:37]
B
That's cool.
- [08:38]
A
Are there specific features or use cases or things that are really driving this adoption?
- [08:42]
B
Yeah, I think the biggest thing was moving into the coding market, to be honest. For a long time in Warp's history, we were kind of known as the AI terminal, which is cool. And we supported terminal use cases really well. Like, how do I do this thing with Docker or git, but the action is in coding and most development activity one way or the other is touching a code base. And so we really started to inflect when we launched a great coding agent, which was like three, four months ago, honestly. So that's been the biggest change.
- [09:12]
A
And how do you think about the different parts of the coding market? There's vibe coding, there's professional code. Like there's all just one thing. Are these separable things?
- [09:20]
B
I think it's pretty separable. So for Warp, our target is pro developers building like software that's economically meaningful. So we really want to focus on actually people who are using agents to build kind of hard apps, apps that might like go into your Mac dock or, or be pinned as a Chrome tab, as opposed to Vibe coded apps where I think it's more of a long tail play. And so I do think, by the way, it's awesome that anyone can code at this point, but I think if you look at where most of the value is in the software market, it's not in those long tail apps. It's in a relatively small number of apps that are super heavily used. And that's my background. I've worked on one of those apps, Google Sheets, and just I have a lot of passion in terms of helping people build real apps. It's much harder, by the way, I think it's relatively straightforward at this point for a good agent to, with relatively few prompts, build a web app. It's much harder to apply these agents successfully to pro code bases. So that's where we're focused.
- [10:27]
A
So I guess one really interesting macro question for me is where is all this heading? And if you look at it, ChatGPT launched in November of 25. Excuse me, November 22nd. So three years ago or so at the time there was predictions that AI would take over the world and we'd be running down the light cone and within five years everything will change and human activity would be subsumed by AI. And there's the old saying in technology that less happens than you think in three years and more happens than you think in five years. And as you think forward in terms of all these different tools and all these different use cases and Vibe coding versus professional coding and that the role of a software developer. Where do you think we are in two, three years?
- [11:06]
B
Yeah, so the way that I'm thinking of it is there's sort of three phases here. For most of my career, we were in like the world of develop by hand is how I talk about it. So My workflow then was like, I would open up a code editor, I would find files I want to change, I would type some code, I'd have some assistive features, and then I would go back to the terminal and I would type commands to build that code. And I think we're switching away from that to something like Develop by prompt, where I start most of the coding tasks that I do right now by prompting an agent and that agent does some work. And I think there's a third phase which is like automated development. Honestly, I think that's like kind of the bigger market here and why people are so excited about this space is you can actually use these agents to automate some parts of the software development process. And so that's like cognition does that. We're moving into this space. Cursor has background agents. The rate at which this stuff will happen is not super clear to me. Actually the most recent iterations of the models, in my opinion, were not as big of a step change as for instance, when Sonnet 4 came out. That was a really big step change in coding capability. I think there's going to be a mix of interactive and automated pieces of development for a while, I would guess. Like, I don't know, it's so hard to know. Like, I think within a couple years you'll have everyone working by prompt and you'll have some slice of development tasks that are just like in the background, like a server error comes in or a new ticket, a user report comes in, something, something is automatically done. But I don't think it's going to be everything. I'd be very surprised.
- [12:45]
A
So you don't think there's a point at which, you know, all of coding activity just becomes agents doing it and then there's like a human who's kind of giving high level directions, like a product manager kind of thing and, or an engine manager.
- [12:57]
B
And maybe, I mean, honestly, maybe like, I think that, I think it'd be silly for us not to like build the infrastructure to enable that. I just don't, I don't know the time frame, but I do think we're going towards something like that. What I, what I really don't think though is like engineering expertise is going to become devalued.
- [13:16]
A
Sure.
- [13:17]
B
So I think it's actually, in the short term at least, it's more important to know what you're doing as an engineer than it ever has been.
- [13:23]
A
And why do you say that? Is it because you need to correct errors that the agents are making? Is it because Things may be architected in a way that isn't scalable.
- [13:31]
B
Is it something totally. So it's like, it's like the agents, you can think of them kind of as junior engineers. So if you didn't have someone who is senior watching them, you end up in a situation where these agents will make code that creates bugs, it could create security issues, it can cause your code base to become really unmaintainable. And so there's actually like a premium right now on these senior engineer skills where you can architect, where you can review code, you can make sure the system doesn't degrade. And so again I would be, if I were like early in my CS career, I would be racing towards building that expertise. Where you don't want to be, I think is like someone who is just like perpetually in the junior engineer state because I do think that's at risk.
- [14:12]
A
And then how do you think about different security tools does that. So for example, there's tools like Socket or Snyk or others who are basically looking at, you know, whether code has or open source packets or other things have vulnerabilities in them or they're looking at different aspects of security holes for code in general. Do you think that just becomes part of these coding tools or do you think there'll always be this sort of standalone companies? I'm just sort of curious how the.
- [14:36]
B
Overall it's, it's an awesome question. So I think tools like that become more important. I think anything that does like either automatic security analysis or automatic verification, I think actually like languages like Rust, things that have stronger guarantees around safety by default, where you don't need to rely on like a human reviewer become more valuable. Whether those things like get integrated or bundled into the coding agents, I actually don't have a strong take on. I'm curious if you have a take. But no, I, I think that the actual fundamental problem becomes more important.
- [15:11]
A
Yeah. What do you think is bundled? What do you think it's bundled? Like what sorts of tools do you think? Because there's this whole world of dev tools and there's a security aspect of it, but there's lots of others. There's design related things. There's a huge spectrum. What do you think just becomes part of coding tooling?
- [15:30]
B
I think there's going to be a class of tools where you start from the front end. This would be things like Lovable or Bolt or Replit or maybe even Figma make if you're coming from the design side and you'll have like an all in one platform for build an app or even like, honestly, like build a business, like, put payments in it. It's kind of like the evolution of, like either a Shopify storefront or like WordPress or Squarespace or something like that. So I think that's all gonna be bundled more on, like, the core Pro developer side. I can't tell if it's gonna be a world of like, MCPS and integrations and like all these tools sort of interplay. That's one approach. Or it's going to be more like there's enough alpha and like you put all of these things together. And I think Warp is a little bit more like this. Like, we're trying to build a single pane of glass, for instance, for doing, like, local agents and remote agents. And if you get a way better developer experience through the bundling, I think that that approach could win. But I don't know, like, MCP I think, is a pretty valuable approach as well, but it's not perfect because you end up, like, with this, this sort of secondhand data coming into all these tools.
- [16:38]
A
Yeah, it's really interesting because if you look at different industries, early in the industry, things tend to be fragmented, often, not always. And then late in the evolution of an industry, things get bundled.
- [16:46]
B
Yeah.
- [16:46]
A
And then when there's a technology disruption, things debundle again and you have point apps and then they start bundling.
- [16:51]
B
Yeah.
- [16:52]
A
And that's just kind of like the cycle of technology in some sense.
- [16:54]
B
Some things, though, that are vertical right now, like, I actually, I think will, like, for instance, like agentic code review. Take that to me, that should be part of a, like, holistic agentic development platform, not so much a standalone thing. So I think some of these verticalized apps, there's just going to be, if you've gone through the trouble of building like a really, really excellent coding agent, which Warp has invested a ton in this, that coding agent should be reviewing code and it would be weird to plug in some other thing that needs to relearn all the context, the rules, the coding convention. So I think there are definitely some things that will be bundled.
- [17:31]
A
Makes sense.
- [17:32]
B
Yeah.
- [17:32]
A
Shrianna and my team has put together this matrix of companies versus features in the coding market. And there's a lot of these sort of like, single feature companies. And it almost feels like all these things should consolidate into a small number of players over time, just as they iterate through the product.
- [17:45]
B
I think so. I think. And I think the core technology is like the harness. So the thing that sits around the model and like I know the model companies are also investing heavily in this and then it's like the context. And so if you have the rich context in your system, you're going to find a lot of vertical applications where like I think security checking is an interesting one. Code review is definitely one. Anything CI related, you're probably not going to want to use a bunch of different systems.
- [18:13]
A
How do you think about it in the context? If I look at other historical technology shifts. Yeah. The operating system or the platform often subsumes the biggest apps into itself. So for example, Microsoft os, eventually they just bundled Windows on top of it and those were the four main apps that were being used the most on Windows.
- [18:30]
B
Right.
- [18:30]
A
Y and similarly gaming was the other big app. So that's why they started Xbox and Microsoft. If you look at Google and vertical search, they eventually integrated all the vertical searches into Google directly.
- [18:39]
B
Totally.
- [18:39]
A
And so in the context of AI, one could argue that if the foundation model companies follow the same approach, they should bundle or at least attempt to bundle some of the biggest use cases. The clearest big use case today is code.
- [18:51]
B
Totally.
- [18:51]
A
And we already know that Cloud or Anthropic has launched cloud code OpenAI, you know, almost bought Windsurf. It always had early coding stuff.
- [19:00]
B
Yeah.
- [19:01]
A
Microsoft, which we know are building some of their own models, obviously have GitHub and copilot and all that. And so do you think eventually those become some of the fiercest players in this market or how do you kind of view forward looking shifts in the market and where some of this functionality goes?
- [19:18]
B
I mean, I do and I think they're clearly trying to do that playbook right now where they are seeing like, okay, if you consider them platforms and they're looking at what are the most valuable applications that are being built on top of the tokens, they are moving aggressively into coding with cloud code. And Codex, which definitely is like as a startup is a little bit scary for sure.
- [19:37]
A
Sure.
- [19:38]
B
The question is like, do they have like that distribution advantage that say Microsoft Ad, where if everyone's using Windows or everyone's coming to Google for the front door, I think it's pretty easy already to add on the first party app in place of the ecosystem. I don't know that that exact same dynamic holds for coding right now. The front door is kind of like, honestly it's still like a native app that someone downloads on their computer.
- [20:10]
A
So that'd be the terminal or the.
- [20:11]
B
IDE at the moment. Yes. I think controlling that is actually a really interesting front door. The Other front door, which I feel like, honestly they're not executing that great is GitHub where all the source code lives. That would be like the locus of doing all this stuff that I think makes the most sense. But right now it's a weird dynamic where we have people who are running cloud code within Warp and Warp is sort of the outer app in that situation. So I don't know you. The other thing that I hope happens from our perspective is that there's a lot of competition at the model layer and that the sort of intelligent tokens become a bit more of a commodity. Right now the models have a bunch of pricing power because there is a real delta between using the Frontier model and using the open source model. But if at some point the models are good enough where coding is sort of, I think of it as like solved. It's like good enough. You don't need to be using the Frontier model. Then it's like, yeah, maybe they have an advantage just from brand and scale and. But I think the advantage is not as entrenched as something where it's like literally the front door, like Facebook or Google or Windows provides those other platforms.
- [21:25]
A
That's a really good insight in terms of the way that you launch an activity or application then drives what you use. And so the hard part is often switching people off that. And that's one of the reasons I think people think OpenAI has a strong competitive position in the consumer world is because it's a default behavior for a lot of people right now is just start ChatGPT and use it for something.
- [21:47]
B
Yeah.
- [21:47]
A
Which is different from like the model layer where there's more switching.
- [21:51]
B
Like I think a consumer ChatGPT has a huge advantage. Like once that that behavior is default. Even if like Claude is maybe better. I don't know if it is or not. Like, everyone knows ChatGPT. I don't know if you saw OpenAI Dev Day yesterday, but they're clearly doing this platform play within ChatGPT now where it's like you have apps, ChatGPT. I'm sure they'll use that data to sort of subsume or take over whatever the best first party integrations are. So they're definitely doing that on consumer for developer. I don't know that that same dynamic is there.
- [22:22]
A
What made you decide to focus on terminals? So, you know, we started talking years ago when you first started doing Warp and even then I think you had really interesting ideas about how to rethink the terminal and how to use that as a launching point. For all sorts of things.
- [22:35]
B
Yeah.
- [22:35]
A
Could you explain that thinking and how it's evolved over time?
- [22:37]
B
Yeah. So the basic insider thing that got me excited about Warp to begin with is like, you have this tool that is pretty much a daily use tool for every developer. It's that and the code editor and the Terminal itself is something that, you know, really hadn't changed much in the last 40 years. It's a tool also where it's like, if you get good at it, you can really get a lot done if you use it. It like works across all these different parts of software development, not just code writing. On the flip side, from my perspective, not a good product. Just like hard to learn, hard to use, hard to remember commands, super intimidating, and just like kind of like a gatekeeping vibe around it as well, in my opinion. And so the original concept with Warp was like, let's build a better product there and see if people will like using it. The business concept has evolved a ton. Like the original business concept was like building a collaboration platform, which is like we've just changed our model to be an agent platform because it's like way more demand for that than a collaboration platform around the terminal. But the sort of core insight that this is an important tool, it's crazy. It's actually kind of invalidated through all these agentic things that are very terminal.
- [23:52]
A
First. One thing that you mentioned I thought was interesting is that at some point the model layer may commoditize in terms of its coding abilities. How far on that asymptote do you think we are or how close to that do you think we are?
- [24:05]
B
I don't know. I think that the increasingly the limit that we see is context and the reasoning capabilities of the models are pretty impressive. The problem is understanding an entire code base or understanding sources outside of the code, or literally just understanding user intent are challenging problems. I still think there's probably much more to do on the model side, but I don't know is the short answer.
- [24:33]
A
And do you think that from a model capabilities perspective, we've hit a point where, to your point, it feels like certain aspects of the models are slowing down in terms of the benefits or outcomes of further investment of certain types?
- [24:45]
B
At least I think so. If you take Sonet 4 to 4.5 and we're big partners with Anthropic, they have great models. Like that was like a few percentage point increase on swedbench for us. And we invest, you know, we've invested a decent amount to be one of the top agents on Sweetbench and when we went from 3.7 to 4, it was a much more significant boost. So I again, that's like not. I don't know what that means about the total underlying trends. I think something with GPT5 was somewhat similar, like certainly an upgrade. And GPT5 I think is actually pretty much on par. It has different feel to it and higher latency, but it didn't feel to me like as much of a step change as some of the upgrades before.
- [25:35]
A
Yeah, makes sense. What other areas of the AI dev world are you excited about?
- [25:40]
B
I am. So I'm really excited about not just like the interactive piece of agents the way most people are working today, but what can you do if you can program against these agents? And so for instance, it's like if you have a version of Warp that's like headless, for instance, you can put it in CI and you can start to do crazy things where it's like, okay, every time someone updates the code, make sure the documentation stays up to date. It's like, that's very annoying for a developer. And so allowing developers to automate parts of their job that they don't like doing, I think is like a big capability. And then from a business perspective, just like automation is a better place to be than productivity enhancement, like one of the challenges with our business, I think with a lot of the coding businesses is just like proving the roi. And there have been these studies that show like you deploy this stuff on real code basis, it's kind of unclear if it's actually having an impact. Whereas if you get something that's more outcome oriented or more just like an automation, I think it's easier to prove the roi. And then you're also not limited by time spent behind keyboard for doing this type of stuff. So from a business perspective, I'm very excited about what's unlocked if developers can program these agents well.
- [26:56]
A
Zach, fascinating. Thank you so much for joining us at Anto Priors.
- [26:58]
B
Thank you for having me. This was great.
- [27:02]
C
Find us on Twitter oprierspot, subscribe to our YouTube channel. If you want to see our faces, follow the show on Apple Podcasts, Spotify or wherever you listen. That way you get a new episode every week and sign up for emails or find transcripts for every episode@no-priors.com.