Raycast founder Thomas Paul Mann - quality, YC and AI

We also talk about AI, but honestly, it's all just cliche. It's genuinely very cool. Raycast is basically trying to be what Apple intelligence should be, but so far isn't even out yet as far as I know. It's a very fun episode. Enjoy. I actually don't think I've properly heard the story of how you started why you started Raycast.

And a few trends that we noticed, like, one, over our career, we felt like we had to use more and more tools outside of our IDE. And when you compare them to an IDE, they were not great, to say the least. They were kinda like slow. They didn't have keyboard navigation. They just felt like I go out of my IDE and things become not nice to use anymore from a developer standpoint.

And in the space of, like, IDEs, we had always sort of our command palette. Right? And and supply was there. Like, Versus Code made it even popular. But then it started going outside of IDEs as well, and apps like Superhuman or Linear kinda in 2020 picked up that pattern where you can press command k and search for stuff to do.

And so the keyboard shortcuts suddenly became the thing where power users wanted to have that. And and so we had those two things that we felt like really interested around. And then we also were always tinkering around, like building little scripts and little tools for our work and making that better. And I think every developer has a bunch of shell scripts, right, that they usually have Yeah. To make them more productive.

And we felt like, it's a shame that those things just like live in the terminal. They're not so accessible. The terminal is very limited. You can't render images really easily. And and they're very technical.

Right? And so we've combined all of that and said like, hey, look, wouldn't it be cool if you have one global interface, which is a command palette, that you can activate anywhere you are. It's connected to all the tools you have, super keyboard driven, super fast, ergonomic. And so that sort of was the initial idea of of RayCost, and it was meant to be early on already, like a drop in replacement for macOS Spotlight. Because we felt like the interface is kind of there, but it's not powerful enough for developers or other technical professionals.

And so we thought about, you're really gonna use just one of those interfaces. Mhmm. And so how can we be sort of a drop in replacement for the de facto default on Mac, but make it 10 times better? And so that was sort of the initial idea five years ago. And so a lot has happened since then, but that's sort of how it how it started back then. And that's also I mean, we went then into Y Combinator essentially with this idea and win the 2020 batch, which was also a fun story.

So you're going there for three months. You have weekly office hours and, like, trying to make progress every week and, like, pushing basically how much you can do in a week. And you're working towards this goal of, like, at demo day, going in front of investors and pitching your startup idea in roughly two minutes. So was super condensed. Yeah.

And try basically to say, like, why are you gonna be the next unicorn in a And what was very different for us is, like, we were basically working, working, working towards there. And then I think a week before demo day, COVID popped up out of nowhere. Right? And the initial reaction was like, okay. Cool. There is this virus. Let's maybe bring some hand sanitizer and, like, call it a day. Right? It would still happen. And then a day later, it was like, yeah, we changed our minds.

We can't do it in person. But what we're gonna do is a very professional video where you can pitch, and then all the investors gonna get this video. That didn't happen. A day later, I was like, look, we can't make this happen. Can you give us a slide, a single slide with five bullet points?

All investors get that. And then a day later was like, by the way, if you're an international founder, you should really get out of here because the border's going to shut down. And so we were hopping on a plane. Like, think this was literally booked a plane not even twenty four hours later. And then we were flying back to London where both of us lived.

And I think when we landed here, sort of demo day started, and then the world was very different from there on. Everything was in lockdown. This basically hasn't existed there. When we talked to investors, they didn't know Zoom. They usually would go with them drinking a coffee, meeting in person.

And so far as it took them a time, like getting our seed round together, which we ended up doing with Excel. And we haven't met our initial partner for, I think, almost a year because everything was under lockdown. And so from there on, things got very different. But, yeah, that was sort of our first three months of the of the company.

But, yeah, those were the first early ones. The thing that really made the change like, we were at some point is like, look, this is cool, but can this be a big company? Like, it's like all a bit utilitarian. And then so what we did starting to do, we talked to YC alumnis. And so these are basically former founders that joined other companies and lot of them are engineers.

And so we looked in the internal forum of YC, who are engineers working at, like, companies like Airbnb, Lyft, whatever. And the cool thing about YC is really, like, easy to get in touch with those people. Mhmm. You ping them an email, and they usually come back because they've been in the same situation. Yeah.

So you're kinda helping each other out. So we hopped on a call with them and asked like, hey, how do you work? What kind of stuff do you do there? What are things that we could improve? And sort of this one pattern that reoccurred was like, everybody was using Jira at their work.

Right? This is like sort of how you track work, But nobody was really happy with it. They were feeling like it's a bit slow. They wanna have it a bit more modern. And then we were like, okay, look, can we like build an extension for this?

Like, just a better experience for Jira essentially. So that was our very first extension, which we built ourselves. There was no API, nothing. We just basically started building this extension and really tried to make the experience better. So we wanted to allow you to create an issue really quick, and we wanted to allow you to see your own issues and modify them and maybe copy a link if you need to, or close them right away.

And so we thought about this, are the 80% really that you do every day in those tools? Yeah. And and we that sort of became the philosophy about Raycos. Like, can we do the 80% of the tools in one single tool, but much better, easier to reach, quicker? And then the other 20%, which you don't do that often, like, I don't know, creating a project in Jira or iterating on a Notion page, we leave those in the tools, the purpose built, they do this much better than we could ever do.

But for all the other stuff, we saw then like, oh, people like to have the Jira there. They're probably gonna like GitHub there as well, Zoom for like creating quickly a a call. And now we have over 2,000 of those extensions in the store.

I think you can accept and merge them properly. We don't allow you to do a refuel because, again, that's where you would go to GitHub and do like a proper refuel there. But I use it, for example, to see like what's assigned to me. And then I see, okay, this is the stuff that I need to take care of, and then I can quickly go there. Or I can we have a few different repositories, so I can quickly open a repository and go there and then check out what I need to do.

But it's, like, saves me a bunch of clicks and potential distractions that I would have otherwise. Yeah.

And it's just like constant learning curve. Even myself, like, I I still learn stuff you can do there. There's so many things out there. I recently picked up a few different new patterns of we released this feature, hyperkey, which I wasn't a big user of before. And hyperkey is like you can modi you can map one modifier key, like caps lock, to basically, yeah, replicate a few modifier keys in one.

So this allows you to have basically more keyboard shortcuts to put it short. But you kinda need to set this up and get into the habit of it. But then somebody posted some really cool use cases of it, and so I picked those up. And so now they became muscle memory. But, yeah, it's a it's a it's an endless learning curve for for everybody.

But I I do think and hopefully a lot of people listening have used Raycast and kind of can back me up that this is not just like kissing up to you. Genuinely, you know, I think Raycast, you open it, it opens really fast. It feels very nice when you use, for instance, the Notes app. It feels satisfying to use it. It feels satisfying to make a bullet point.

I don't know what it is. It's just Yeah. There's this it feels well made. And I kind wanted to ask you about like how how does that happen? Because I'm sure it it doesn't it's rare. So Yeah. It's you must be doing something

So it's not just that. For us, like quality, I think there are a few things. Like when I when I look around actually in the real world, when you've when you have something high quality built, right, usually when you touch that and you lift it up, it just feels a bit heavier. Yeah. Which, like, feels very good designed, very solid, and you feel like you can't really break it.

And so we try to take those feelings to software almost. And for us, it was very important. It's like, we wanna make it really fast. And so there's a lot of, like, kinda like tricks that we do in the app to make it really fast. Like, we have a very intelligent caching layer.

So it really, like, feels like everything is always available. And so that makes it like a really fast experience. We care a lot about design, which we also feel like ties into quality. Mhmm. Like, I always say this, hey, our users are like eight, ten, twelve hours in front of their computers.

The best we can do is like show them some nice pixels. Yeah. But you also need to design it in a way that is highly functional, because we are this weird app that you basically want to spend as little time as possible in. So our design is quite minimal, but even that is sometimes challenging to do in a good way. But really, think what it boils down to, like, we're building a product for ourselves.

So when we put out a feature, like, it goes into internal, and then everybody, like, downloads internal builds, and they get automatically updated for the whole team, every night at least, and sometimes during the day when somebody kicks it off. And so that means, like, everybody who is in Raycost using the features before they go out for, like, a couple of weeks. So when it goes out, the usual paper cuts that you would otherwise experience are all gone by that moment. Or, like, at least most of them. Because we have this feedback culture, like everything that pops up getting shared internally.

And this can be like little things, like literally visual glitches, which like sort of is a meme internally that we don't ship glitches and hold off releases even if, like, we see something that happened a few times. But more importantly, also, like, a lot of, like, UX issues that we identify. Because when you use it, like, as often as we use Raycast during the day ourselves, you, like, experience every little detail. Yep. And so then we erase this and try to fix that.

So a feature when it drops in internal is usually very bare bones, and it goes through this relatively quick loop of iterations to make it really rock solid. And so when we ship it, like, we ship something that we feel like is really good. It's not always perfect, right? There is, like, bucks and, like, little things in there that we would like to improve, and you also need to say at some point that's enough. We're also shipping every two weeks.

We wanna be extremely fast. So there's always a bit of a trade off. But with this strong dogfooding internally, I think we can catch most of the stuff and then have those delightful experiences that you pick up yourself in in notes and various other places inside of Raykos.

But I'm wondering if there's any, like, kind of, you know, non obvious or, like, obvious costs to it where it's like, we're willing to take that cost because we care so much about quality. Yeah. Or like, we're willing to sacrifice this because we care so much about quality or anything like that.

We we say we want to build something high quality, and therefore aim for this. What we often see working quite well is like, we start with something broad, which is usually every feature, because a feature or a project is just an idea at the beginning, right? So we start with that, and then we go around that and start working on it. We're usually relatively quick because we're programmers, we want to get stuff done, right, and basically build something. And then throughout building, we get a lot of answers.

And then when we feel good enough, we oftentimes set ourselves sort of a deadline. And so we have this biweekly release schedule. And so we try to aim, in every release, have what we call a highlighted feature. And so we can look like, okay, we're now here. In three weeks, there's another release.

We feel like we can make this. And so we say, let's aim to release in three weeks. And then I think you flip around the switch and you say, okay, we release in three weeks. What are the important things we need to get done? And then you work towards this release.

And then very typical, and we probably should map this out, is like, you see when it gets closer to the release, you see like a lot of these small paper cuts getting fixed. Whereas before you build the bigger feature and trying to figure out. And then you also make decisions like, hey, look, we don't need this in the first version, let's ship it. And oftentimes when we ship something, it's a relatively simple first implementation. And simple, oftentimes, is hard to get But we know like there are probably a bunch of stuff that people will ask for.

And so we pull it out and see really what people will ask for, and then we try to follow-up relatively quickly afterwards with a second iteration, if possible, to address basically sort of the initial feedback. And so it's this fine line you wanna, like, basically find where you you wanna build a good enough feature, but you also don't wanna overbuild it in a way. And rather have a smaller scope, but then in a in a good quality essentially. Mhmm. That's kinda how we

This is this is what we thought is going to be this perfect thing. It's not really necessary day to day. And then you get this, like, really good opinion about a feature and feel quite quite good about it when you have it ready. And then oftentimes also, like the features that go out there feel for us so normal at this point. When we ship it, it's like we sometimes surprise like, oh, wow.

People really love this feature. Yeah. But it's like we we got so accustomed to it, right? When you work on something for like weeks, like it's so normal to you, and this happened to me, like, a few times where I feel like, this is a much bigger release than I than I thought it will be. And then as I realized, hold on a second.

Yeah. I've been using it for, like, six weeks already. For me, it's already normal. Right? So I don't have this novelty effect that that people have when they experience it for the first time.

So if you type send feedback in RayCost, you get a little text box, and you can just send us anything you want, and we we will read it. And so this obviously got a lot more challenging nowadays, but sort of how it started was, look, how it really started, like, back in YC, we did onboarding calls with users, and then we wanted to see how they're using it. A ton of stuff didn't work. We were fixing it, gave them the new version next time. So we did that a few times and feel like, oh, this is super valuable, but it doesn't scale up.

And so what's the next best thing? Okay. We let them send us everything they have as easy as possible and answering everything we get. And so we did that. In the early days, often happens when, like, they send us a bug report. And because we were reading it, it's like we were fixing it, and thirty minutes later said, hey, it's fixed. Gonna be out tomorrow. Wow. And I'm really building this rapid loop. And I think it adds a lot of loyalty.

Because how often have you sent someone feedback and never heard back? Yeah. It happens all the time. And so we wanted to be better than that. And so the next evolution was when we had a big volume.

We're still replying to everything, but the sort of middle step we had probably two years in in the company, what we did is we still wanted that everybody can observe what happens, but, like, not everybody needs to reply. And so we had sort of this encore where a few people would reply. But every feedback from every day was like basically the next morning posted into Slack, so everybody can read through the feedback. And the good thing about that is like when when you're an engineer and you worked on one feature and you see constant feedback coming in about that feature, you see what's broken and what works and what needs to be And then we're picking those things up individually, where basically people say, hey, that's my feature. I own this.

I make snippets now better. People wanted to have this feature. And then they were hacking hacking it in, and then we were building it and shipping it again. And so we had that. And so now we have, like, a support team taking care of that, basically.

And basically try to do the the same thing with just high volume. But, yeah, like, we really try to close the loop, and it comes down to how can you guys basically get, like, the signal to the person who builds it so they know what is broken, what needs to be improved, what people want, what they're missing. I think that's really where where feedback boils down to. How can you how can you make this loop as efficient as possible?

Yeah. And we also be honest, like, when we don't do something, we also respond like, hey, look, we're not doing this. Or like, hey, yeah, something we're working on should come out in a few weeks. Like Yeah. We also try to be very transparent when you ask us, like, if we work on that feature, we probably say some in some cheeky tone, yes or no. Okay.

And then oftentimes, because we talk from developer to developer, oftentimes with our user base, they're like, that makes sense. Thanks for explaining. So we we're taking time for that as well.

And I we probably are a good representation of the people we wanna reach. Mhmm. And so let's do something that we would genuinely like. And then also make a lot of those decisions. Like early on, we decided we're not requiring a lock in. And for a VC backed company like us, like, and our investors still probably doesn't like that. But like, we don't have a lock in. Right? And why are we doing this? Because we don't need it.

Like, if we don't need a lock in for the functionality of the app, why should we have one? And that sits very well with developers because they come in, they just want to play around with the tool, they don't want to create an account, they just want to get in there. Right? And so that was a deliberate decision. Analytics. Like, we have anonymized analytics, so I can't tell if you used the app. I only know somebody used the app today. Again, there's something which like I'm

Right? And so we always try to, like, come there with, like, our opinion and not like with, like, the best practices or the cargo hold or what's there. And so there was early decisions. And then further down, we like yeah. We basically grow through word-of-mouth.

We feel like we wanna build a product that speaks for itself, and then people will fall in love with and advocate for that. And we tried things like paid marketing at some point, and we feel like that doesn't really fit our brand. It doesn't really move the needle as well, to be honest. And then we're leaning more and more into this. And then now we're thinking about really how can you make this a lifestyle is sometimes what we take, Because it's oftentimes like, yeah, there is this software, sure, but then it's much more than that, right?

There's people behind building that. Our users really love the tool and they can't work on their Mac without it anymore. So we'd started doing in real life events, like meetups and hackathons to get people together. And there's always great dynamics. Like, it's a mix of like, literally, as we did earlier, I show you what we built very openly.

This is the kind of stuff we're tinkering around with. We're sitting there like an hour long doing Q and A and answering every question you ask us, and like, try to be really open. And I think this resonates with people because it's a bit different than others. And and we try to like basically building like a like a really momentum and brand around that. Brand is really what we're thinking about a lot.

It's like, what is the brand that we put there? How do people discover us? And basically making like software more than like what we have in our computers, but something really that you say like, hey, I'm a Raycost user. Basically, you're identifying yourself as a Raycost user almost.

And that's sort of inspirations. But then really, think like when I think about productivity, people think about it as something to measure. For me, it's much more like a feeling. Mhmm. Like, I can leave a day and feel super good because I got a ton of stuff done. And that makes, like, a great day. Not just a great working day. You go home and have fun with your family and friends, and you feel much better. And so that's really what we wanna get to. Like, how can you get this feeling?

And we all had it. Right? Like, we went into the weekend, felt like this was a great week, or like we wrapped up the week. Right? And how can we basically get people having that feeling more often?

That's kind of like where we're coming from. The closer that we've seen is like which also kind of sticks is like sort of the flow state. Like, that sort of when you think about how do you get to this feeling, it's oftentimes you're really doing concentrated work, challenging work, and you feel like you accomplished something. And so that's sort of kinda like a bit how we're trying to put it together. The tricky part is, I'm not gonna lie, is like with Raycast, you mentioned yourself, you can do a billion things.

And so how do you put that all under a good umbrella that people understand? Yeah. But we often am struggling still is, like, to describe what Raycost is in a simple way. Mhmm. Like, it's sometimes easier with, a tool that does one thing and you get an extreme value out of this one thing. Yeah. Reagas is the extreme opposite. You do a thousand things with little value, and then overall, it may be the same amount of value. Yeah. But you have the thousand things.

Right? And so that's a bit of a slippery slope. But, yeah, we're thinking a lot about how do you how do we get people feeling more often happy about their accomplishments?

Some people wanna be productive in their massive corporate environment. Some other just getting started with their startup, and they wanna get something out of this. So we see ourselves almost like as an enabler. Right? Like, not the end result. Like, we help you getting somewhere there. And so it's really like a tool at the end of the day. And with every tool, you need to learn how to use it. The more you learn, the more you get out of it. Right?

There are probably people that can operate a hammer better than me. They're getting more out

Yeah. And spoiler alert, you can. And it's just you can it's almost like chat it's like chat GBT kinda experience, but like it has full access to your computer. So it can do pretty much anything for you. Yep. And it actually feels like a really exciting experience where it's like you could just say what you want to happen on your computer and there's a good chance that Raycast can do it for you.

It, like, looks a bit more modern, gets a bit vibrancy and blur here, but it's still like the same thing. Right? I click icons, I do this, I do that. And so really, when you think about this, like, in your head, the steps you go through, it's like you wanna accomplish like, a thing, and then you translate it into, like, how do I do that on my computer? And you're going a level deeper, you're translating it into mouse clicks and and and key presses, right, to really, like, go that level down.

But why not saying to the computer what you want to do and let the computer do the work for you? Right? That's the kind of point of a computer. And so we kind of like getting there, think, slowly but surely. And the pleasure that we have, we feel like operating systems are so stagnant, and we probably the closest of an application that is to an operating system because it sits there and has access to everything.

And so could we be that layer, this productivity, this AI meta layer that sits on your operating system and enables you to do those kind of stuff? And so with the extensions that we have and with the access to the file system and other things on your computer, like, we're kinda having that now. It's still like, we have it in beta right now. It's like when you when we did the stuff earlier with the wallpaper, it's like, does this work? It does work.

It's sometimes a hit or miss. I'm not gonna lie. But when it works, it's like truly magical. Like, you do this and feel like, oh my god. That's that is the future. Yeah. And then over time, it will work more often. And at some point, it's just the norm how you use a computer, I think. It's not for everything. I was gonna say that.

Because there is some stuff which you wanna have, which you do and need the accuracy, and it's a different kind of work. But there's a lot of this kind of stuff that, oh, I wanna do some research. I take it off, and then it maybe saves the research directly into my Notion page. Yeah. Or like this more free and open ended things I wanna do.

And it's really different when you chat to something, and then also when you bring in potentially voice. Really changes how you operate suddenly with something. And I think those new modalities and user experiences gonna really change how we're gonna use our desktop and mobile computers.

Yeah. They have professional tools like Figma. They like, using things like Notion and Linear. So how do you connect those things intelligently with each other together? And that's sort of where we wanna take Raycost. They become smarter and, like, maybe becomes proactive and does things for you at some point. Yeah. And you wake up and it's like, oh, there is my there is my work. Or like you get creative. Like, hey, look, those were the pull requests that you need to review. Here's like

So Yeah. They have to use ChatGPT to like change it Yeah. To be like, okay, actually like low pitch on this one, high pitch on the thing. And I was just imagining like this world where Raycast, you can be like, okay every folder, every file in this folder, you know, anonymize it. And like Raycast goes and like has f of vpeg

Has the same problem. Right? So how can you share all this kind of stuff? And kinda like going back full circle, as we started with like, traditionally, there was probably one engineer who wrote like a bash script for that. And then they shared this Bash script.

But then the next person wanted to edit it and didn't know Bash or like, maybe they did and improved it. But now you can do those things with natural language, and more people can do this. But I think like really sharing those things with each other is really what's cool about it. Like, when you have those these are all mini tools. Right?

But that's oftentimes what we need for our job. Not everything is like a massive thing. You kinda need these little things to make you more productive, to automate a bunch of stuff, to improve your workflows. And so how can you do those little isolated things? Super easy, shareable, make it like instantaneously.

So it's it's very easy to do. And I think that's like pretty exciting when you have all those little things doing maybe things automatically for you or on your fingertips. You can just say, oh, yeah. I wanna blur out this video and also make it silent or make it a high pitched voice.

I think generally, when when you build an API that people can build their own things with, like, you have certain ideas what you want to enable, and then you're very much surprised what people were built with it. Just like I would probably not have imagined that we would have that many extensions by now. It's just like, yeah, it's crazy how many people build this. And everybody gets different value out of it. But, yeah, I think it's it's always surprising.

Yeah. I think that's that's one of the the ones that I always find good.

And so we have like a coffee channel in Slack where people nerd out about the best brews and whatever. And so then at some point we said, okay, let's make a coffee. And then Petro got in touch with with David from Terminal because David was in one of our meetups. And then one thing led to another, and basically did a coffee. And so now you can, in Raykos, actually fully order your coffee.

Like, this recently sent my parents a package of coffee. Yeah. It's really cool, actually. So you don't need to go anywhere. You just open Raykos and and put all all the stuff in there and then it goes out. It's That's amazing.