#129: The challenges of working with AI in single-player mode | Andrew Somervell (Co-founder @ Hamster)

Amazing. We are live. Welcome Andrew. Welcome Ben. Welcome Mark. Thank you. I feel welcomed and ready to go. Also this is special. I feel like I haven't seen Ben in five weeks. We built a backlog for the podcast so this is our first time recording since end of July I guess. Also I'm super excited to spend time with Andrew, love what he's building at Hamster and also love the topic that we're going to discuss today. I think the reason why I was super excited to have this conversation with you Andrew is that you're very unique. You're building an AI product which I guess is at this day and age not that unique. You were early in the trend. You were early in that trend but what's really unique about you is that you, you know, when everyone was starting to build AI products for individual people, you know, like you see the codexes, the cursors of the world, cloud code, it started with multiplayer. I think multiplayer, which if you tell some people are like, oh man, this guy's crazy. He's starting with the hard thing. And so maybe I think that's probably a good place to start is, you know, what makes multiplayer unique or different from individual kind of player mode. And I'm also like, yeah, what are the key challenges that you're kind of running into as you build this product from scratch? Yeah, for sure. And you know, it came out of an insight from, we released Taskmaster early last year. We basically invented plan mode at the beginning of last year. We've had a 2 million downloads of that. And so the insight came from listening to customers saying, okay, great. So like you've solved the context problem on my local machine, but how do I take this to work? So I'm an engineer or I'm a PM and I'm using cloud code, but then what happens? And so that was really what like led us to think, huh, there's this, the bottleneck shifts, an individual is more productive now. How might we, you know, bring that back in together? And yeah, I think the, the first insight was that the new bottleneck. By the way, Andrew, one thing that I don't know if Ben has the context is, so when you started plan and it was Taskmaster, which was basically like you made plan mode before all these other products like Cursor and, you know, Codex, Cloud Code made it, which is really cool. You're really early in that trend. At that point, you're already kind of like exploring ideas with your co-founder, right? Like you're already kind of like in that mode or, and then that's what kind of, when Hapster became into play, that's more or less the context there. Yeah, so I was, I actually, AL, my co-founder was on X being like, you know, he could kind of solve the problem on the weekend being like, man, you know, AI sucks. This was GPT three or four days and it kind of did, and I had the same problem. So we, we actually met on X and, and we're sort of trying to solve the same problem locally and teamed up. I was in Auckland, New Zealand, he's in Montreal. So I was getting up at one o'clock in the morning to, you know, start this idea. And then yes, it was only a few months in, we realized that it was just from customer feedback and understanding the problem space that we thought it has to be like, this was a CLI on your local machine. I don't think product people want a CLI. I think they're just in there cause that's what they've been given. And I think that there's a collaborative space that's web-based at SAS that makes more sense for that kind of a surface, better product surface for collaboration. Yeah. So that was like the genesis going from one to the other. And then we had plenty of super interesting. A lot of people were using Taskmaster in the Bay Area and a lot of inbound interest. VCs that were using it themselves and wanted, you know, tech support on X. And we were, we were giving them support and fixing their problems for them. And so they were pretty keen to help get us going on a preset, which we did in July, August last year. And yeah, we've just been, we've been very grateful and lucky to have that happen, which let us like knuckle down and just focus on this vision and execute, you know, for now, yeah, what are we, 13 months, 14 months. Definitely thought that we were going to use AI to be super fast and we were just going to smash it out and have it all done and got our first product to market in October and probably needed, we did it early enough that we were embarrassed by it, which is I guess the old, the old adage. We got plenty of feedback, tail between our legs, went back to work. And it is interesting how much product, how much platform you still have to build and how much you still have to get right to keep pace with the frontier models or the frontier labs and their harnesses. But yeah, like really happy with where we're at now. And we're very much in the mode of that, just that customer feedback loop, you know, cranking product market fit, cranking retention. The classic product stuff is our day and our world. And we just happen to be doing it for product people. So our users get it pretty good with feedback. So it's fun. Can I make sure I'm following the story so far, just to kind of bring us to today? So Taskmaster launches as a collaboration between you and Eyal, who met on X, you guys built plan mode before plan mode was like a native feature in products like Cursor, Clod, Codex, etc. Some of your early power users who were vocal with feedback were based in the Bay Area. So you took that feedback, started iterating, and then you continued improving single player mode in Taskmaster while the problem or the kind of like seed of the idea for multiplayer or hamster was starting to kind of take root. Is that fair? That's right. Yeah. Through June, July to sort of October. I mean, I literally went to Cursor Cafe in October, hang out with the team, we know them. And they're like, we all love Taskmaster, we use it. But we're building plan mode. And I'm like, yeah, it's an open source product. We don't have a moat. We're on to the next thing. I'm not gonna tell you what the next thing is yet. But like, you know, welcome, you know, love, love all players. And it was, it was fun. Yeah, I actually don't, I'm like so stupid when it comes to like how to build an open source business. Like I've never built an open source business, but I'm guessing some people would have seen what you had kind of like the traction that you had, what seemed like product market fit with your open source project with Taskmaster and chosen to, I guess, like double or triple down on Taskmaster as like the business or the company and try to figure out how to like build an open source business model around that. It sounds like you chose to kind of keep it alive, keep it in, keep it running. But the real opportunity, like the big thing is this other thing and we're going to build a whole separate company that is going to be like a whole separate product called Hamster and that's going to be like the new the new thing. That's right. And you can play open source business models very many, many different ways. I mean, we toyed with the idea of being an AI gateway back then and adding margin to tokens and, you know, that was one way of doing it. We actually built it. But we felt like the opportunity that was a, that was a, the open source projects more like a give to the community and certainly gave us breakthrough when it comes to attention and, you know, being in the minds of individuals. And we had an early access wait list. So it was reasonably easy to tell the alignment story between the two. I actually smashed out Taskmaster 2.0 last week. Just like every, you know, what would you do this right now in the same context, if you're just trying to save the same problem, which, you know, more to come on that. But yeah, we just, we use Hamster to build lots of stuff. And so we build a lot of open source things and we spin them out just as like, Hey, you know, I built this thing with Hamster. This is useful and valuable. Would you like to use it? And my team has to use Hamster to build stuff. So we just go, Oh, interesting opportunity space, use Hamster, build the thing, release it to the public for free, you know, MIT license it and just keep that flywheel going. And we're happy to keep doing that. It works in terms of like, we like building stuff. It tunes us and helps us, you know, whittle what we're doing. And then if it's of any value to anyone else, then that's the community. You know, that's what we love to contribute to. Ben, I see you. You want to ask a question? Yeah, I do have a follow up. Just to clear, it's a clarifying question, then you can go, Mark. So just to make sure I'm clear. So when you say your team is using Hamster to build these products, sounds like you have kind of like a portfolio of products that you've put out there, all built by the team using Hamster as their primary collaborative workspace. And the clarifying question is, does that mean that your team is not using something like a cursor or codecs or like one of these other products? It's like you've, you've kind of cannibalized for your team, these different workspaces into Hamster. And that's kind of like where all the work is happening. No. So we've also done a very first class plugin ecosystem with everything, I think. So we live, people say, what's your ICP, right? I'm like, it's a team. That team can't be an ICP. I'm like, well, it's worked for Atlassian for a long time. So let's move on. It's PMs and EMs, right? And so anyone in the EPD kind of space, but an engineering manager and engineers want to still have control and be able to bring things back on their local machine and say, hey, like, don't commit to PR and like, that's all done for me. I need to tweak this and I need to iterate it. So Hamster works beautifully as a plugin, MCP, CLI all in the background, but it's a plugin for all of the things, CLIs, IDEs, all of them. Where those partners of ours and where we play in the marketplace have not done a great job is in, I have a hunch and I have an idea. I need to ideate it with my team. We need to refine it together and then align with stakeholders. And then I go and I grab it. Now we have Hamster Cloud Agents, just like you get Cursor Cloud Agents and you can use Cursor or Cloud or whatever. You can go into Cloud and say, hey, let's ship this brief. And it just knows what to do. It syncs the brief to a markdown file on your computer and you just go ahead with it, deliver it. And it updates Hamster. It's like, okay, these tasks are done in the plan. So all works nicely together. That's something that like, if we were like an all in or nothing, I think there'd be plenty of engineers who'd be like, nah, I don't like it. From a PM space, I think that PMs are using cloud code because it's the only surface. Certainly cloud co-work is not doing it for them. And I don't think PMs want a CLI. I don't think PMs want to be managing a skill pack all the time. I think PMs want to be engaging with customers and understanding insights and setting goals and figuring briefing, like what needs to be done and then collaborating in that cross-functional team with the rest of the team. I still think that's powerful. So we felt that was the surface that we wanted to work in. You can say ship with Hamster Cloud and it spins up a container. It builds the thing, it creates a PR. But I would expect that these little nits that a PM can just be like, just fix that button. Like it doesn't need to go through a whole engineering loop. And increasingly you can ship a whole initiative or a whole goal. But I think we all know that with AI, someone who's experienced a senior wants to be shepherding it and just keeping it on rails, like through. So if you want to use that in Cursor or Cloud Code or Codex, you absolutely can. And that's how I use it. I mean, I will get alignment on a big, we actually now have engineers or individuals owning an initiative and they brief things together and get feedback on each component part, but then they're shipping. And I'll still pull, like right now, the delivery all can now happen with an open weight model on Hamster Cloud. But I might pull Astra in and be like, Hey, validate this for me. How do we go delivering these different parts? Do they all, are they all cohesive? And Astra these days is usually like, yeah, cool, did well. But I want that right now. What are we, September? I still want that feeling of control. I still want that feeling of review and like get an adversarial model and do some security testing and things. So we enable that, right? But we're not trying to replace your subscription. But I do want, I do think there's a product surface and a collaboration space that you work with as your team, which works for everybody. So that's the theory. Yeah, which, I mean, it's funny, I think Ben and I, you know, we like to say that we operate in an AI native way, right? And I think in a weird way, I do think that like brief collaboration is like the hardest one, like especially when we're kind of like coming up with like an idea or a plan, because basically what we do right now is like, you know, we'll have a meeting, we'll discuss a bunch of ideas. And then basically one of us will be like, Hey, I'm going to own this initiative, right? And then as the initiative owner, what we're going to do is we're going to spend more brain cells thinking. We're going to also use some of the tokens to help us think, you know, get our thinking to work better. And then once we're happy with the thinking, then we're like, okay, like now I want to present this information in a way so Ben can react to it. Like there's a shit ton of context that I don't want to force into Ben. So then I start thinking like, you know, what is the most information dense way to do this, but also that's succinct and clear and that I can get what I need from Ben, which is like, okay, this makes sense. I'm aligned or this doesn't make sense. Change this few things and then let's ship it. And right now the way we're doing that is we just built almost like a HTML brief, right? Where unfortunately, given how, you know, cloud code or codex work, like we need to basically like push it into GitHub. Then Ben needs to like pull that and then kind of react. Or sometimes they'll send it directly on Slack or something. Yeah. Or they'll say, yeah, I'll paste to Slack. Yeah, exactly. I'll download the HTML and put it into Slack. And then I have to open it in a browser and then I have to read it. And then I have to choose how I'm going to give feedback on it. Am I going to record a loom? Am I going to just type up some quick feedback in Slack? And then to your point, you then have to, as the recipient of the feedback, take it, probably feed it back into your system so your agent can make sense of it again. Whether it's like ship it, you ship it, or it's iterate, then we have to do the loop one more time. It doesn't feel like, it feels like the best we've come up with so far, but it also feels like we're probably going to look back not that long from now and say, like, that's crazy. Like that. You were doing it on hard mode. Yeah. Yeah. Like that was like the fax machine equivalent. Like those guys were using fax machines. That's what everyone's doing. Everyone's doing. Everyone's like, I need this. I need to be using AI. There's plenty of products out there, but everyone kind of gets shoveled into the like the one subscription. And then, yeah, it's like, so we kind of, one of our, you know, key things is alignment in minutes, not meetings. And it's, you just need that one shared workspace. So answer broadly is imagine multiplayer AI with the AI in there, Google Docs, Slack and Google Docs and, you know, cursor together, meaning that we, we, I, we've been using it ourselves a lot. So we ended up initially with briefs and then I was like, oh man, we're ending up with all this technical stuff in a brief and a brief should be something that you as a business team can just, you know, three paragraphs, like, yes, that's the thing we want to ship. That's we're aligned on that. So we were like, okay, needs a spec so that the engineering stuff can go in a spec. And then there's a plan, which means the engineering team can go through and go, okay, yes, that's how I want it to be laid out and it gets shipped. But then we're like, okay, so there's these three modes. Maybe yeah. To, to recap. So it sounds like what you're saying is that any project or initiative should basically have like three like atomic work units, almost like at the planning level and that kind of cascade from each other. Right. And the first one is the brief, right? Which is like, Hey, almost like, why are we doing this? What problem are we trying to solve? And here's maybe some context or user problems. Then the next thing is the spec, right? Of like, Hey, like here are the things that I think we should build to solve those problems. And here's how they connect to that initial brief. And then the next one is like a cool, like if the spec is correct, then here's actually how we're going to build this back from an implementation plan. Yeah. Is that, is that more or less how you, like the way you think, like those are just separate things basically. Yeah. So just, um, this week, probably this week, maybe next week, we're going to ship auto mode, which lets you YOLO through all of that. So it's not gates and it's not slow, but you want. If different stakeholders are like, I just want to understand what we're doing and why and where that came from. Then as an engineer, you might have architectural questions. So what Hamster does in the move to the spec is it'll have an engineering manager be like, Hey, do you want to use the service or do you want to park it over here? And that's really, that's almost all an engineering manager needs to do now. Yeah. Let's just make sure it's there and follows this pattern. So the spec lays out the engineering pattern. The plan you're usually pretty happy with then, you know, once that's done and we use, we use frontier models in the back of Hamster to do that planning and specs so that it's like the plans nailed. Um, yeah. So there are, there are artifacts that are correct for like the, the, the stakeholder or the party that wants to validate them. But I don't want to see them as like, you know, it goes chug, chug, chug over the course of weeks, you know, back to old school mode. It, it, it can pretty much go through it. We will. We love the fact that I can be like, Hey, this customer came out with this insight. Um, I'll brief it, uh, Max, my lead engineer, I'll come in and go cool. Yeah. Generate the spec, answer a couple of questions. He just fashions it. And then anyone on the team picks it up and delivers it. So it works quickly because it's multiplayer, right? Like it's the same, same product surface, the same thing. And then, yeah, so you guys have got it. I'm just like, uh, It reminds me, it reminds me Mark of like the days of, you know, when I, when I was pre AI, when you're working in a traditional kind of like software pod, you've got your PRD or your project brief, which is typically something that PM kind of like is responsible for generating, getting some input. Then once there's general alignment, your engineering manager and you, or maybe like your tech lead will come together with a high level, like architectural, like maybe like a general plan for how we're going to approach this from a system perspective. But that's not yet getting into breaking down the specific tasks and JIRA or whatever. Um, at least this is how my teams have gone. And then once that's agreed on and you've maybe even identified which engineers, like as ICs are going to be responsible for each part, then their job is to then kind of like break down the architectural plan into the specific tasks. And as the PM, by that point, I don't give a shit about those specific tasks, right? Like I don't care what the backend task is, like the 12 backend tasks or the five frontend tasks, you know? I would say I would care, like if all of a sudden there's like this small requirement that I was like, oh, it's more of like a nice to have. And then all of a sudden it's like, it's like 10 story points. And I'm like, what the fuck? No, like, yeah, get that to two points or let's cut that puppy because like, I mean, and I know like that's getting a little bit like more, you know, it's getting easier with AI, but like if something is like 10 points, that means like there's some hidden complexity or the requirements are not clear. That should generally get caught in the, in the design review when you're doing your initial system architecture review with like the engineers, if it's like, here's the thing, this looks great. What do you guys think? Like how long might it take if at that point they're like, this is probably like a week, week and a half of work. And then they come into like the breaking down of the tasks and it turns out to be four or five weeks of work. That's like shitty scoping on the, on the engineering side. Like they should have pushed, they should have done a better job up front thinking about it. Yeah. Yeah. Anything, anything bigger than five points needs to be broken down. No, I mean, we're, we're out of that mode now. Right. So I just, I just want to say like now. I sound like a dinosaur talking about this. It's like, bro, why are you, why are you going so far back? No, yeah, why, why are you going back to last year? So just on that, um, what's happening now? Can I ask a question? Like, cause before we go deep into this thing, cause like actually one, one thing that feels there's a gap, like, I think we've covered, I would agree with Ben that we've covered most of the things, but like, I think there's like a big kind of like hole in the whole thing that we haven't talked about. That's like design, right? It's like, I didn't think before, like, I even would go deep with the EM, we usually would do like a team review where we like go through the mocks that the designer had put together. So then like, it's kind of like you're, you're putting visuals and like interactions in, like, you're kind of almost like creating a connection between the spec, the brief and how the things look. And then that also kind of gives a much better sense for like, you know, how the plan, the scoping and like how difficult are going to be things. And yeah, I just feel like we haven't talked about design at all. Is that step dead? No. For a lot of teams? No. Well, I mean, I hope not because, so I love design. My thing is design. I love product design or behavioral design or visual design. And so, yeah, so we started with the brief and we ended up having to go two ways. One, one is downstream of the brief. I want the brief to be the thing we're doing that you stakeholders can align on and then we get technical. But I was also like, you know what we ended up doing? We ended up in Hampshire having a. So like, you know what we ended up doing? We ended up in Hampshire having a whole lot of draft briefs. Wait. So why, why were you guys having a whole lot of draft briefs? Like how, well, what led to that situation? And then why, why did that feel like a problem? Yeah. So we were just loading up on ideas and I, I had this, uh, we're going to shoot the PRD because the PRD was such a mechanical arm wrestle, you know, the whole, um, oh, the requirements are too big, ask engineering when it can be delivered Q3 next year, like why? Like it was just such a, I just hate always hitting a PRD. So I shot it. I was like, you just brief. Apple does design briefs. It works. Tesla and SpaceX do briefs. Let's just brief stuff. So I thought I was like dogmatic. Everything's a brief. And then I was like, oh, okay, Andrew, you created a problem. Now you've just got to hold a draft briefs on the cutting room floor and you have to sift through them to find the ones that, you know, so you build filters into the table on the product. And I was like, um, there's a, there's a mode here that we need to surface properly and specifically around design, which is, and design is not just visual, right? There's interface interaction, UX, but also technical and like cross-functional. I think in the world of AI teams are far more cross-functional now. So I feel, I felt like I needed a new thing. So it's like, it's a note. We're going to do notes. Notes feel ephemeral. They feel something I can start easily and just yarn to hamster and say, Hey, I got an idea, or I can bring a screenshot or a post-hoc error or a customer insight and just reason on it myself, you know, and by myself. So, so I was like, right, there's a new thing. It's called notes. They're basically briefs, but they're just way lighter. You could think of a, um, I like FigJam or, uh, you know, just, just wherever you want your scratch pad to be, which just enables that creativity a bit more. So we now have three, we have three meetings a week in at Hamster. We do direction setting. So broad strategic stuff. But the one I love the most has actually been discovery. And so everyone in the team is encouraged to, you know, everyone's got fun ideas and great insights as they're building and using the product. So we just come up with notes. And so in the discovery meeting, it's a full ideation session. Everyone can bring a hunch, you know, or a, or a proof of concept. But one of the interesting things now is I'm a whiteboard guy, right? So I actually quite like to go do it. Even if it's a 30 second, like whiteboard squiggle and bring that and chuck it in a note and bring it to discovery session. But people are spoken word, uh, they might, an engineer might just put a branch together and a proof of concept. Everyone should be able to bring their ideas to the table, however the fuck they want to bring them to the table. And that makes a great discovery session. And we have these awesome, you know, oh yeah, that's interesting. It's like a show and tell mixed with like a design crit, kind of. Yeah. Yeah. And, and design crit still comes at the end of delivery. Right. But, but yes, basically how do you get everyone in that mode of fashioning an idea and, and cross-functional design? Right. So design is, Apple has a material science person and a software person and something else in the same design section, which makes sense. And it makes sense in a world where we're all builders and the roles have kind of melted away and we're just a team that's cross-functional, that's just working on like what the best thing is. So that was my, the hamster now has notes and you use that in discovery and then you refine it together. And then you just, you know, hit a button and be like, cool, brief it. Or, I mean, it was, um, Evan Spiegel on David Senner's podcast was like, you know, we have, he does, he spends like seven hours with a very small design team every week and they have a hundred ideas and maybe shit one of them, you know, and maybe it's just like, you just got to have a lot of ideas to have great ideas and then you also- It reminds me of the, uh, the writing room for like SNL or something. It's like they just like, it's just like you throw away the vast majority of the ideas. Yeah. And, and we're, we're, we treat ideas like pets and everyone comes with this idea in an AI world, like someone goes and ships this huge PR and brings it to table, like tonight we eat and you're like, where the fuck did that come from? Like, so in a world where AI, you can build anything, what should you build? And the team should be aligned, like working in a team as a great team that like feeds off each other is a better outcome than like there's solopreneurs that are going to build it, you know, one off one man billion dollar business. But for the rest of us who like quite enjoy riffing on the idea, I wanted this AI assisted space where you could still bring stuff together and, you know, and have that ideation, like we're a few weeks in, um, and I really liked that time. Like I really enjoy that meeting and we let it run, you know, and it, but it's not when we're setting strategic direction, it's not when we're knuckling down and saying, Hey, what are we committing to this week? That's our delivery meeting. But it's an safe space, ideation space, notes, you know, misc and other, bring them to the team and it's working really well. I just think people are feeling empowered to come up with dumb shit or great shit. And yeah, it's working. Yeah. Maybe to stick to the step back, cause I think it just might be helpful to like get the lay of the land of how everything fits together. Yeah. And I think, so you have the direction setting meeting, right? Then you have the discovery and then you have delivery, right? Those are the three, three meetings a week, three meetings a week. I think what'd be really helpful is who, who attends to each meeting, right? How long is the meeting? What is the objective of that meeting? And then maybe what is the homework that people have to do before that meeting? And then like, what are the things that get decided on that meeting? And then maybe like how they're all connected. And I think maybe it makes more sense to like cover them in sequence. So like maybe like starting with the direction setting, then discovery, then delivery, just like, I think it just kind of helps to just kind of like get the full picture and then kind of like, then we can start to see, okay, how does that map to like, you know, quote unquote, the old way of doing things back when, when Ben and Mark were, were doing product work. Well, it's also like, what is the right way to do things now in a team? So there's artifacts that weren't wrong. They were just slow. Right. I think like there, there was pain in some of the things and I've been through that pain a lot in many different product roles and felt like there's stuff that was valuable and like, like group collaboration and there was stuff that was painful and politics and arm wrestling. And I was like, we don't need that shit anymore. And certainly in an AI world, you don't have much patience for it. So with direction setting, uh, Hamster just has goals and you can pick a goal framework, OGSM, Northstar, OKR, set it up at the beginning. And I wanted, or actually Thierry our designer wanted a, you know, strategy on a page, you can look at this one goals page and understand where we're heading and why. Um, goals, we use OGSM. So you have pretty much one high level objective, a bunch of goals, strategies to achieve them, and then metrics. And we build into like post-hoc and others so that you can, the metrics are live so we can see our, you know, two week retention climbing and it has, it's like quadrupled while we've paid attention to it. Um, so everyone on the team, it was interesting, in our direction setting, sorry, in our discovery meeting, um, this week, one of the team was like, Oh, I, I don't know if we want, you know, engineers to see the, the direction, like the goals page. And I was like, what? Like, why not? You know, that's, everyone on the team should know exactly why we're doing something and it should be there for everyone all times. And then every initiative ties to a goal. So strategy on a page, classic, um, pick, pick a framework that you're used to. You don't have to go too hard on it, but Hamster is really good now at, you just talk to him like, Hey, I want to do a new thing. And it's like, great. What framework do you want to use? You're like, I don't care. Pick one for me. And it's like, cool. We use OGSM. And then you have a conversation and it builds it all out for you and ties the metrics to it. And you're like, Oh, look at me. I'm like the world's best PM. Um, so. Wait, how many people are at Hamster? It's probably a good time for me to ask this. Just so I get a sense of like the total size of the team. Yeah. We're 10. Uh, and, and all over, uh, different countries and stuff, different time zones. Mainly New Zealand in the U S um, one in Bangkok. We had one in Paris, Ales in Montreal. Uh, so we're, we're doing the distributed life. Um, I would have loved to be able to hire a bunch of gun, uh, people all in one office in SF, but, uh, we know what that's like at the moment. So, um, yeah. When you get a fat, fat series A and relocate a bunch of people. Yeah. Fat seed. $1 billion seed round. Yeah. Like my head of engineering is like, when am I coming to SF? I'm like, just got to get that seed done. And I'm like, you're here. That's good. Wait, so who's attending the direction setting meeting? Is it the 10 people? Is it just you and your co-founder? And like, is it, is it mainly like an alignment meeting of like, you know, like me and Andrew, like, you know, went to Joshua, you know, me and my co-founder went to Joshua Tree and we saw the vision and this is what we're doing type of thing. And here's how it like changes the direction or is it like everyone can come in and contribute to that direction? Everyone can contribute to the direction and we certainly reference it a lot. Um, at some point I think co-founders and actually, um, product and lead engineer just get together and say, right. Given the fire hose of information on X at the moment, does anything change? And we're in a good space where we, we only do like 45 minutes to an hour a week, um, four of us and we're tweaking now, we're not, there's not some massive, you know, huge shift in strategy every week, but you're, you've got to be tweaking every week, right? Like it's no longer quarterly QBR planning and like, you know, years in advance, if you can get six to eight weeks out in advance, like good luck. But we're trying, we're facing into some chunky stuff at the moment, which is all very just customer and PMF focused. And it just kind of delivers to one big audacious goal. Um, but the strategies and we're shipping so quickly that you're iterating like what's next and what's the next most important thing and what metric have we turned? So we'll hit a metric that we intended to hit and go, right. What's next? You know, are we re-upping that from 75 to 80%? What are we going to do to get there? Or are we doing a massive shift? Um, So, so are you, um, just like, just in the spirit of continuing to kind of understand what's different about teams that are working with like, um, multiplayer, you know, kind of AI native experiences versus single player. I think everything we just talked, like a team that's working in a bunch of single player mode could do that meeting, could do that meeting too. Um, so what do you think is different? Is it that there's maybe, uh, yeah, I just love, love to get your take on like what's different about teams that maybe use Hamster to do these kinds of meetings versus teams that are, that don't have kind of like a multiplayer solution. I think you've got to keep everyone out. Or the prep or. Totally. Uh, so, uh, so in direction setting, we're just trying to set a North star and make sure everyone's swimming in the right direction, you know, and it's an agreed direction and everyone has the context. But if I sort of say like the Kiwi in me says, if, you know, someone's fucking off in the wrong direction fast, they can go real fast and with AI now. So, you know, if, if someone's swimming against the tide, you know, like speed that way is not velocity that way in alignment with the team. So I think that's more important now. And I think, um, when you go to review and we'll get to review and deliver, you've got to tie it back to like, what was the thesis? What were we trying to try and improve? You smash it out in a day, but it doesn't mean you just all go off in different directions. I think the team should still have that North star and that, and that common goal. Um, so direction setting and hamster is, is in service of aligning velocity towards one shared direction and in any given two week period month, you know, just so that like as a team, you're feeling the lift, you're not all just off doing completely separate things. I don't know if there's, if you just stacked ranked a different customer problem and just let people be, um, you'd get stuff done. I don't think it would be a cohesive product with taste and craft. Um, we've suffered from that ourselves. If I've been like, okay, everyone's going to pick up a different thing. And then you look at the UX. And again, the AI designer and I will be like, ah, we got some rework to do to bring that back in line. So I just still think cohesion on one product surface is key. Uh, and, and not every idea deserves the same, you know, weighted kind of outcome. Right. So you do have to be making sure you're, you're spending your budget wisely. Uh, cause you've got to, you've got to have a cohesive product that the customer understands. And so it's got to feel like it came together and people understand your opinion and your position and like why you do what you do. And I think that in a team, if you're just all off in different directions fast, that you lack that in a product. So people like, what is it? I don't get it. How does this make my life useful? And your engineering team can be like, but it does all this great stuff. You're like, no, no, I get that. And not, not to, not to be difficult, but everything you're saying also applies in a world where people are working in with like single player AI mode, as far as, as far as I know. So I'm just trying to like understand what, what is different ultimately for. Like you still have to make sure all the work aligns to like the goals and the priorities and stuff in a single player, like people using single player mode. So like, if you had to really kind of like boil it down to the crux of like, what feels better for you with your team working on with, with like a single, with a multiplayer AI solution, like what, what, what's, what do you think like the kernel of the differences or like, what's the crux of the thing that is not true about the teams that are not working in a multiplayer way? One word is alignment. And I think second to that is speed and cohesion. Like, I think that, I just think that product teams are figuring out how to keep things on the rails when, when things are speeding up and I think that's hard. And I think politics gets harder and I think copying and pasting and which Slack thread am I in and which Google doc and which notion and which linear ticket. You're just like my understanding and data that I've seen and then talking to lots of customers is you're just context switching so much, you're like burning your brain these days and you need to just calm things down. And like the whole reason we called it hamster is that hamsters run the wheel and you have more time to think. Like you just need, you get more time, calm time. And I, like, it sounds like the data and surveys that are coming out of saying like Lenny's survey is, PMs are just, NEMs are just expected to do 40 to 60% more, a hundred percent more in the same hours. And I guess we kind of, it hurts. And so we want people on the team to have that sense of calm, that sense of cohesion, the hence sense of alignment while things are going faster. I think hamster does a better job of that than copying and pasting between four platforms. I want to take a moment before we get back to the episode to tell you about Insider Loops. This is a company that Mark and I started because we just believe this should exist. Companies should be offering candidates the best possible preparation for their loops, yet what they give them is vague and often not very actionable. So we took the matter into our own hands. We've put together well-researched and continuously updated guides for PMs interviewing at companies like Anthropic, OpenAI, Google, Stripe, DoorDash, and more. If you're getting ready for PM interviews at those companies, go over to insiderloops.com and learn more. Now back to the episode. Yeah, maybe like I can see it the way I see it and Andrew, correct me if I'm wrong or not. But I think kind of AI has fundamentally changed the way we do product in a lot of ways. But I think in the way that I think hamster has thought about it is that two things. One is like the one year roadmap planning or the six month year roadmap planning doesn't really make sense anymore. Because one, it's like we're constantly changing how fast we can ship things. So like it's like we kind of have to underwrite how quickly we can write things. And the other thing too, like the landscape is changing so quickly, right? Like all of a sudden you launch a model. So I think teams that are AI native and want to keep up like have to be way more kind of, I don't want to say reactive, but they have to be more flexible and more nimble for the market changes and the kind of how the environment is changing. And the other thing that's happening in parallel is that you can also move a lot faster. But the one in 60 rule, are you guys familiar with the one in 60 rule? It's like for one degree that you get off target and you're flying like 60 and let's say every 60 miles that you travel, it is more like on aviation, on plane, like you get one mile off from the target, right? So like the longer you go and you kind of, the far degrees you get off target, like the more expensive the mistake is because you started getting like really off track. Right. So I think what Hamster is solving is like, how do you make sure that like everyone stays aligned to that kind of plan or goal and that how I'm making sure that that goal is tightly kind of connected to every plan, every feature, and then also making it very easy for everyone to kind of update that goal, update that strategy. So then, you know, if things change, you know, it's not like this crazy, like 20 meetings alignment, we don't know whose version is the latest one. So like, it's almost kind of like enabling teams to be like, you know, like, you know, enabling teams to be nimble and also reactive to what's happening in the market. That's kind of how I understand it, but yeah, I'm not sure what I got wrong or right there. No, so like literally, it was interesting because the definition of velocity is speed times angle, like velocity is directional, is a feature. And so, and velocity does need to be aligned. So you're right. And then there's the context switching. So Hamster has a context graph and a workflow graph underpinning it. So you can go to any artifact and be like, how'd we get here? It's got an activity tab. You can say, Hamster can join this call and just listen in. And you can say, you know, it shows a knowledge graph and it shows, you know, Mark and Ben had this conversation here. They aligned on this brief. They agreed on it. This person delivered it. And that's how we got what we got. And that's really interesting for people who leave and join the team because there's your, like, your context isn't leaving with people. AL has been on calls with like 70 plus product leaders, right. And asked what their core challenges were. And there were nine. Alignment, eating their week. Work about the work, getting in the way. Context leaving with teammates. AI having to be retaught every session. The specs are not the bottleneck, not building. So you can build anything, but like, what should you build? Small changes after the build are too expensive. So incremental iteration and like that, that design crit and the little, the little tweaks are become painful. Tools don't share context. AI is making better artifacts, but not better decisions. And like how to roll hamster out safely across a team that would, those were the, that's luckily AI is great at distilling 70 transcripts into like the key, the key principles. But those are all things that product leaders are facing right now in a team setting from their words. And I think we do a pretty good job and we certainly focus on those things. Yeah. So like velocity, I think Ben you're right. A high-performing team certainly can keep that on track, particularly if they're in person or they're just great at Slack or they're great at communication. I'm not sure every team is a high-performing team or is keeping the things on the rails and just want a little bit of help. We're not trying to solve their end-to-end product process. We're not trying to build their product for them. We're trying to give them control and like a say in the product along the way. So just to make it, make it maybe concrete, coming back to the example of how Mark and I are working like cavemen, I guess these days where we share these like HTML files with each other and you know, that whole clunky workflow we talked about before. If we were working in a AI native multiplayer way with a tool like Hamster, what would be, I guess, like the modern way of doing what that, that workflow we were describing as kind of like a contrast to the current way we do it in single player mode. Yeah. So I think that like the two of you would very easily be able to set direction, be a simple, simple hangout and you can be multiplayer. Let's say we don't want to meet. Sure. So we like, we like to do a lot of stuff async. So like, yeah. And by the way, Hamster joins Slack. Yeah. Hamster joins Slack and Hamster can, you can upload a voice note and it'll take a transcript and just read it. So, so you can just say Hamster, like you could be in a Slack thread and just do what you do now, but then be like, Hamster brief this. And it will have the context of the Slack thread and everything you've added to it and brief it for you and come up with that artifact that you would align on. And so let's say Mark already came up with the artifact, right? So like Mark took the, took the time to put together some kind of like HTML that's like describing an idea. And then it's like, Hey, like, let me know your feedback on it. Is the idea that I then just drop the feedback directly in Slack and I tag Hamster. Then what, what, what then happens? Like, does Mark still have to do, like, does he get, it's like the workflow, because the workflow is new, it's hard for me to have, I don't have a mental model for like, what's actually different versus just like adding in the agent into our. different versus just like adding in the agent into our Slack co-workspace that's just like an agent, right? Yeah, so we don't use Slack for much anymore. We, I don't know if we should have, but we built Slack into, you know, into Hamster where it's a multiplayer thread, a chat where you're both on it with the AI. So you can chat on and iterate and ask Hamster to tweak a sentence or just tweak it yourself, just like a Google Doc. And so the workflow is that you take a hunch and you ideate it through Hamster. And if you've got the HTML, just drag it on and say, what does this do? Think about it on the note, if you want to refine it together, turn it into a brief. Now, like, yeah, this week you'd be like, YOLO deliver and Hamster can spec it, plan it, deliver it in a container, PR it, have it ready to go for end to end. But if you want more control on those artifacts along the way, you just jump in, there's a thread for each artifact. So you just go, hmm, don't like that, add this task, remember to test, and it adds them to the plan. And then it just continues. So it's, you're not, you're no longer copying and pasting. We're actually, this is something, again, this week, I was thinking, we've taken a lot from the success of Grokbot in the last few weeks because I've been jumping up and down that Hamster's a bit wordy, kind of, it's a bit like too descriptive and one of the things I love about Grokbot is like, cool, done, you know? And you're like, awesome. That's how I want to be talked to by my bot. I don't want all of its thinking. So we're dialing that in a little bit more, but what I like the idea of in that, so we were thinking more about what is that thread that carries the way through and that you don't have to restate and don't have to copy and paste and don't have to put in HTML and don't have to put in cloud code. And I was like, well, it's just, it's a thread. Like, by definition, a thread. So I think you'll see threads turn up as a menu item in Hamster, which is just, things are just threads and you start with an idea and you follow that thread through to delivery and all of the artifacts along the way are all connected on the same knowledge graph and have context for all the meetings that you've riffed on it, or you haven't done a meeting, you've done a voice note, you're async, or you're in Slack and Hamster talks in Slack as well. So it's keeping the thread together. Like if there's like one thing, it's that provenance of an idea. And then if you think about a team trying to juggle priorities and spin plates, how do you go back and get into that mode? Like, where are we at with that? Who's done what? Awesome, okay, I can see where that's at without having to have a meeting. You know, I don't have to have a point-perking meeting anymore. I don't have to do sprint review. I don't have to do, you know, all of the old school theater. I just know where things are at on a given thread. So it sounds like basically the way we change is like right now we have a lot of the context scattered in Slack, Granola, maybe WhatsApp, like kind of what Hamster could help us, like just like centralize everything. Like you can do everything that we do in all those platforms with Hamster. And then that way, we don't have to be the ones kind of like, kind of trying to remember, hey, where did that conversation happen? And this thing, this other thing. And instead, just like, we all have to be Do we still use Asana? So Hamster two-way syncs with Jira, Linear, Asana and something else, another one of them. So Hamster can run Asana for you. You can just turn something into a plan. And if you want that in Asana, because that's where you have all your other tasks, then it can just keep that up to date in Asana. So as it is, you can just run it. And if you want it in Asana, you can just keep that up to date in Asana. And if you want it in Asana, you can just keep that up to date. And it can just keep that up to date in Asana. So as it delivers, it just ticks things off. I didn't realize, I always thought that project managers who then call themselves product managers were the ones that imposed Jira on engineers. And that's why engineers use Jira. And then I was talking to engineers, sorry, product people. And they're like, no, hell no. It's engineering that tells us we've got to use Jira. We use Asana for all our planning. And I'm like, oh, then what do you do? And they're like, well, we do our planning in Asana. And then, well, yeah. If Mark didn't used to work at Asana, I think it's like almost 100% likely we'd be using Linear, if I had to guess right now, for our project management. Right, probably, yeah. Right, and you try and keep that always. Linear's doing a great job. I always have a lot of respect for Cary in particular. And they're getting it. They understand the context graph. They get it quite well. What Cary did really smart early was he did two-way sync with Jira so that one team could peel out and use Linear, and the whole rest of the team didn't have to yet. So I've done the same thing to Jira, Linear, and Asana. So you can use Asana in general and just come and try Hamster and connect them and be like, oh, I'm just going to try it on a couple of things and it'll keep them in sync. It has 1,200 connections. So we do an MCP connection with Granola. We have a first-class integration with Slack. We do two-way sync with Asana. So I would like to think that Hamster AI and the knowledge graph that we have about all your stuff means that Hamster, a thread in Hamster that carries through nicely integrated with everything is a dream compared to the copy and paste and what have we got. Now, in a smaller team, you're like, I don't have a context problem. I know what three things we're working on and we might be async, but we just know what's next. So I don't think Hamster has been a hit with the prosumer, the individual engineer. But I do think when an engineer is like, I would like to be handed work by the rest of my team with full context. Remember the PRD was this, we don't want the engineer talking to the customer. We'll do the product work. Engineer gets all the requirements. Engineer says, I can't deal with this. It doesn't have the non-functionals that I need. Like all that goes away. Everyone has the context for everything always. And I just like the idea that a brief carries a thread through from end to end so that the person delivering at each step or aligning on it has all the context and it's not lost in a note or a Slack thread or in Granola. It's all there. And that's kind of what platforms is, what you'll see. Everyone will lean in on multiplayer over time. Atlassian's doing it Rovio and saying, we're one of the context graphs. Everyone will have four or five context graphs and we're one of them. So I think they're coming around to our way of thinking as well. Yeah. So I guess, I think like context engineering, context is like new currency, right? I think all these other harnesses, tools, like desktops, apps, et cetera, like also going to connect with everything, right? Like they're all going to want to connect. Like you want to make them more useful. You want to connect to Granola, Slack, email, spreadsheets, Confluence, whatever information lives. Like they're going to end up finding a way to connect to it, right? And I guess what you're saying here is like, we're definitely going to do that. Like that's a given, right? That's like kind of almost like the price to play and to be useful. But in addition to that, like what we're doing is we're almost encoding an opinionated stance on how teams should work together. Like what are the atomic units of work and that kind of to get from point A to point B, right? And, you know, that might be a brief, then a spec, a plan, and kind of like making sure that like, you know, all that context is contained within projects so everyone has clarity. Is that more or less kind of how you're thinking about it? Yeah. And the more I've been like, oh, now we need a note. Now we need a plan. I'm like, it would be nice if you didn't have to have all these individual things. But then I think the worst JIRA setup was always the one where a PM 10 years ago went and created all these gates and artifacts and it was like became this behemoth that only one person understood how it worked. So you do have to be opinionated in a way, but even like Anthropa came out last week with their SDLC course. And it was, you should do a intent in a markdown file and then you should do a design in a markdown file. You want to define SDLC? Software Delivery Lifecycle. Software, yeah. I always thought development. Delivery, development. Software Development Lifecycle. Yeah, Software Development Lifecycle. The end-to-end, you know, of everything. So everyone, like to your point about, you know, when context engineering came out, I was like, no one wants to be a context engineer. Why do I have to, and all of these things, if we're ever making you do work, like that is not the point of Hamster. The point of Hamster is it does it for you. You just keep it aligned to what you want to do and it runs the wheel. That's why Hamster works. But then you have to get a lot more explicit about the context. Like you have to get more explicit about goals. Like for example, Mark and I don't really have a clear, I mean, maybe that's a separate problem, but like we don't have, you know, like. We don't have problems, man. Yeah, yeah. Like, it's like if you were to add, if all of our context was perfectly available to an LLM right now, and I started hacking on a new note, let's say like a new idea, and I asked it to be like, how does this tie to our goals? I think it would probably generally understand like businesses are trying to grow. So like this is how this could help grow. It would probably be able to fill some of those. But if we were not like very explicit about here's the goals that matter and here's the goals that don't matter, here's what matter, you know, like that kind of thing, it would probably just be spit balling, right? And answer to this, it's so easy to set up. Honestly, if you go into Hamster and be like, we're trying to build a community and we were product people and now we're going to be also for design people. We need to go from here to here. It's very good at setting it up for you so you don't have to. Like I know you're like, so now I've got to get set goals. I've got to manage them. And like that feels more work, but Hamster is very good at just doing it for you. So you're like, oh, we're heading this direction. But isn't that kind of like delegating thinking? Like setting goals feels like very strategic work. Yeah, it is if you let it run with it, but you don't have to. Like, so yeah, this is, what was I, I think it was Patrick Collinson was, he did a YC thing a couple of weeks ago and he was like, you know, I still don't let, I don't like the way AI writes. I still have to do the writing. The RL that they've done on writing is terrible. I still need to do the writing and the thinking and I 100% agree. So if you're just like Hamster put a strategy and a plan and initiatives and stuff together, you will get an AI outcome. But if you're like, here is the specific thing we need to do in the next three months to get to feel like we're validated that designers want to be part of super, you know, whatever, whatever it is on the pod, then that's good context when it's helping you set up. By the way, all of our collaborations happening with our business insider loops. So just if that helps, yeah. Yeah, I mean, I think like, you know, there's a bunch of apples and oranges ideas that we toss out that I think even we have to discuss like, wait, like how does this move the metric? Like, how does this contribute to our goals? And a lot of it feels like art, but maybe like to some degree everyone feels special until they kind of see AI do a good job of maybe like giving them a good first draft of something that they didn't think of themselves and then they could kind of like workshop it. Got it. So, all right, this is one of the things when we were building Taskmaster, we're like, that helps an individual on the local machine and probably scales to like two or three people. Beyond that, when you have a team that's cross-functional that you want to, you know, start losing alignment. Like if you don't have an alignment problem and you don't have a direction problem because you're async and you're pretty aligned on what it is you're trying to achieve, then, so wait, like you probably don't even have to. That's right, that's right. I think that's kind of maybe what, one of my takeaways from this conversation is like single, so I came in this conversation feeling like single player mode is kind of like problematic the moment you start working with more than just one person. It's almost like, or there's like real problems with it. But I think one of the things I'm taking away from this conversation is if the atomic unit of your team is like two people, it's like co-founders, let's say, like Mark and I, and you're talking like all day long and there's no room for, like Mark, there's basically very little risk that Mark ever goes off in a direction that's misaligned with what I care about and vice versa, then there's not necessarily an alignment problem across the entire, all levels of altitude from like vision to strategy to like day-to-day execution. So therefore, it's almost like single player mode is fine because there's no alignment issue. But if I were to like, if we were to add two people to the team or something, and now we want to start to make sure that there's alignment about what everyone on the team is working on, if we just do what Mark and I are doing right now, that feels like it would break. And so that's kind of like one of my takeaways is like a two-person co-founding team is in some sense more aligned, as aligned as a single founder. And if it's like really well aligned, but then the moment you add more people to the picture, it's like extremely hard to get four or five people to be as aligned as like a two-person co-founding team. It's like, it's very, very, very challenging. So you need some more systems around you to keep you in that alignment in a way that allows you to still work fast in an AI native way. Yeah, and I think maybe we're a bad example because I think we're also like, there's not that many boundaries between us. Like I think we're like very like, we'll share very much what we think, what we feel, like we just kind of know each other very well to the point that like it's very hard to separate the two. So like, I know not only just what, where we're going, but also like how Ben as a person will react to things. So like, it's like, there's all these like almost like context that we have on each other, plus the frequency of communication, plus the simplicity of our setup that, and maybe we're extremely talented at getting aligned too. Like, I think we deserve some credit. And that, you know, like, that I think, yeah, we're, you know, we're in a unique situation, but I agree that the moment you start scaling, but I think probably Andrew is not going after us yet. We would be a huge pain in the ass of a customer because I think like our problem, I honestly think it's like more functional than dysfunctional. And I suspect that when you introduce a team that is having a lot of, like the kind of problems you were talking about, Andrew, that when you did the survey, we don't have like, none of those problems really felt to me like things that we're personally experiencing. And that's indicative to me that I think your ICP is very different than us based on the kinds of pain points that they're experiencing. That's why, that's why just having a Claude code subscription as an IC is epic, right? Like you don't, and again, so the Anthropic SDLC thing that they've put out teaches Claude to run you through setting your intention, figuring out your plan, what's your delivery going to be, hand it off to sub-agents, all that kind of stuff, right? That's easy to do as an individual. And it's easy to do not only in a coding sense, but in a business sense. And so you're running based on your feedback loop and you can pretty much, you know, IC the thing. That's what Taskmaster did. And actually, again, I built a new version of it last week, completely clean sheet of paper, what would it look like now? And the only thing I was frustrated with as an IC was that I burned through Astra in a day, I burned through Grok Unlimited in a day, I have two Claude subscriptions, I burned through those. And every time I hit the cap, I was up for API pricing. I figured out that you could burn $136,000 a month doing a regular working month with Astra if you went to API pricing and just kept going. I was burning $14 a minute. It's really hard to justify using the latest models for like your everyday work. Yeah. Well, I was doing a blank sheet, like YOLO, yeah, Taskmaster. But I wanted to tell something that I had all these subscriptions and that I prefer Astra to do some thinking, but that I want Claude to do the planning and then I want GLM 5.3 fast to do the delivery and that I want to go back to Sol and have it do an adversarial review. And I was like, well, that doesn't really exist. Why does that not exist? And what's happening is these people are, every walled garden is giving you frontier level stuff and getting you baked into their harness. And then they're unable, you're unable to get out. Like the minute you hit this wall and your usage is gone and you feel trapped and it's actually quite a lot of pain to go to, like just if you had Claude and Codex. So to go from Claude and like, I've hit my cat for the day, I'll go to Codex. You've got to like retrain it. So I was like, I just want a workflow engine that keeps my context and I want to use Grokbot. I just want to use Grokbot to talk to it. And then I want Grokbot to go tell Claude and Codex what to do. So I built that last week and we've open sourced it. And I would like to think it probably does better than Taskmaster did. And it was just like tier point, the IC workflow, which doesn't, it's kind of how we do hamster under the hood for a team. And we were like, hmm, you know, how would we open source like current hamster? And I was like, let's just, again, clean sheet of paper. It's something that's like a one line install. Grokbot talks to it. It manages Claude and Codex and sets up your PRs. And now I don't have to worry so much. And it knows like where I'm at on my subscriptions and it just uses them appropriately. So that was an example of me. And this is why we build all these projects all the time. Deeply understanding, like I'm not in hamster world only, but in terms, you know, the G stack or, you know, office hours, or you can do grill me or all the classic skill packs. You can do all that in hamster. And so I just find hamster way easier to take an idea, even if it's just a project with me, just me as an individual and get it all right. And then it's actually easier for me. Cause again, I can go into Cloud Code and say, ship this brief. And it goes to MCP, figures out all the context, delivers it, tells it it's done. So I find it very useful. God, this thing's got to stop falling down. I find it very useful in my IC world, but I can understand how you're like, I don't need another thing. Like I'm good. I'm happy. I'm happy in my role. I've got my subscription. We just Slack and like do voice notes. So yeah, like the ICP is teams. And that was very clear from us. Like Taskmaster solve for the individual and that, that local context. And then hamster was very, if you're taking this to work and you've got more than yet two people and you're, and it gets unwieldy, that's where hamster is for you. So that was the problem we were trying to solve with it for sure. And very acutely aware of the fact that, like there's still people downloading Taskmaster and using it particularly in an open source world, which is an incredible growth in the last few, a few months. So yeah. Yeah. Thanks for, for bearing with, with all of my dumb questions along the way. It's, it's exactly the context that like, you know, someone comes to hamster and kicks the tires and goes, this is for me. And like, you've got to be very opinionated on like who you're for and who you're not for. But turns out we have a solution for you too. So love it. Hamster is the team mode. Amazing. I know we're coming on time and yeah. Two final questions before our gratitude corner. And, but yeah, Andrew, this has been awesome. And one is where can people learn more about you, what you're up to? And then two, how can our audience or listeners are kind of tuned in right now can be helpful to, to you. Yeah. So we're try hamster.com. Go try it out. I give you 10 free briefs a month and you can pull your team in and give it a go. And then being useful, we have a community Slack and, you know, all feedback is a gift. And that's basically the mode we're in as a company now is just completely listening to, you know, what people would want it to do next. And, and we can solve for that in, you know, then that afternoon or the next day, basically we, we pick customers coming to go, you know, it would be great if it did this, because it's pissing me off and we just ship it. So, yeah, there's a great feedback loop there and it's the, it's the old product loop, which luckily I really appreciate and love doing. So that's why we're, that's why we're doing it. Yeah. Amazing. That's an ambitious SLA same afternoon. I like it. We do. Yeah. It's fun. It's better. Cause when I was building banks, you wouldn't get it in customer's hands for 18 months. You can imagine what the impetus to get a PR out the afternoon was. It was like, it doesn't matter. Incredible. Uh, you want me to do a gratitude corner? Yeah. Okay. So, uh, our, our, our last question for you, uh, last but not least is our very special gratitude corner. So this is a chance for you to give a shout out, say thank you to one or. out, say thank you to one or multiple people who you feel have played, played a important role in your journey to get to this point and just give them a, give them some love. Yeah, I have a few. Uh, I gotta, gotta stick to one. Um, No, you can do, you can do, you can do them. You can do a few. Yeah. It's been a, it's a wild ride in the AI world when you start a startup, right? And I think, um, my co-founder and I both, uh, have the ability to have like robust conversations and, uh, you know, be pretty frank with each other and, and, uh, just move on and like, get to the next thing. I mentioned you guys have got the same thing where you strong opinions held loosely, um, and that's been great. Right. You need that when you're being, when you sort of have existential moments. So yeah, shout out to AL for being my co-founder and, uh, and I think we're still smashing it. Um, and then, uh, actually my super team be the first, um, first group that I've been in and, uh, I really appreciated the feedback from the whole team. Um, and, uh, actually, you know, Ben, I hope we get to hang out a bit more about the conversations with Mark. I think, um, I specifically put my head up to try and spend some more time with product leaders in SF, uh, and it's been a great, um, a great way to do that. Um, I've been pretty head down working, so I don't have a lot of like personal life to like pull from at the moment. So the, the work relationships that I've built and the feedback I've had from other product people on this have been really special, so really very grateful for all the feedback I've been getting from that. Amazing. Thank you so much. That, that means a lot. And with that, that's a wrap. Uh, thank you both. And I excited for people to check this out. Talk soon. Bye. Thanks for coming on Andrew.