Group Chats With AI Don't Work — He Built What Comes Next

Awesome. Okay. Hello everyone. Welcome back to TwoSet AI. Today I'm with Steve Ruiz, founder of TL Draw. So he's the guy who built the infinite canvas that powers whiteboards for companies like ClickUp, Google, and Autodesk. Steve, thanks for joining. Tell us a little bit about yourself. Yeah.

Thanks. Yeah, my name is Steve Ruiz. I'm going to position it right here. I am based in London, which is where the majority of the team is. We're kind of like an office startup. And yeah, I'm the founder of Tealdraw. So Tealdraw is about a four-year-old startup. started it. It kind of grew out of a kind of an open source project that got really popular. I'd kind of given myself a couple of months before going to take a job at Adobe and was like, hey, let me finish this open source thing that I'd been kind of hacking on before that ⁓ and kind of get it out of my system, so to speak. But then it kind of got popular and then it kept getting popular. And then now I'm still there and we're a team of like 15 people who all work on this ⁓ just like a canvas, like an engine for building. Canvas apps. So whether that's a whiteboard or a design tool or all sorts of things that would kind of leverage this Miro-like or Figma-like Canvas, ⁓ you can do that on top of the engine that we've built, which is the Tealdraw SDK. And yeah, that's me. Or at least that's the most recent couple of years of me.

The latest version of you. ⁓ Can you, can you, for people who don't know about whiteboards or building with whiteboards or canvases, can you explain what TL Draw does? Like a simpler level and maybe more for one more level for more technical audience. So for instance, one level to elementary schoolers, like our kids, our seven year olds. And how do you explain to your daughter and how would you explain it to someone technical, maybe someone who's building AI agents?

Yeah, so to kind of explain it like M5 version of the SDK or just the canvas in general ⁓ is it's a place where you can draw, where you can write text, and where you can kind of just move things around. And the way that you move things around and resize things and so on, ⁓ it just kind of feels the way it would feel in real life, right?

Like on a paper? Like on a piece of paper?

If you want something over here, hmm? Yeah, like on a piece of paper, like on a chalkboard ⁓ or a whiteboard if your child has spent a lot of time in an office setting. But if something's over here and you want it to be over here, you just kind of grab it and you kind of just move it across. Or if you want something on top, you just draw it on top. I it should just work the way that it seems like it should work. ⁓ More technically, Teal Draw is

you

is just kind of a normal website. ⁓ It is built out of ⁓ HTML and CSS, just like any other website. It just feels like a whiteboard, ⁓ which is very unusual. Most of the types of apps that act this way, like a Figma or like a Miro or a Miro or Whimsical or an Excalibur, or take your pick, ⁓ they kind of, ⁓ what you're seeing is not like normal website stuff, it's basically just an image that is being repainted with the contents of the canvas. And there's a million reasons why you would do that. It's the same thing that a video game would do for performance and for speed and things like that. my bet with Tealdraw was like, no, I could probably just make normal website stuff just as fast and ⁓ just as performance. And if I could do that, If I could just make HTML and CSS and these building blocks of normal websites work like a canvas works, then I would be able to put anything onto that canvas. Video, GIF, interactive stuff like forms, entire other chat, or code editors, or text editors, or even instances of tealdraw itself on the canvas inside of itself, on its own canvas. And that's what we did, and it turned out to work really well. And it turned out that not everyone was like other companies. Very few of them who were interested in Canvas were building design tools or graphical editing tools where you really do want a Figma-style Canvas which is squeezing every little bit out of every frame. ⁓ even with gigantic documents, Teotra is very fast. It handles things very, very well. ⁓ But more importantly, it's fast enough for the type of content that people actually wanted to do, which was, hey, I want to put my AWS dashboards on a canvas so that I can zoom out and see them all. Or I want a place where I can bring in some of the notes that I have in my document and put them on the canvas and be able to compare them. Or I want to take every page of my website and put that on the canvas. And then as I'm editing the code, I want to see those changes ripple out across all the different websites. And the use case is just keep going and going and going. The fundamental one, the common set of elements, is basically like a whiteboard set, diagramming, drawing, things like that. But ⁓ because we've built with these common building blocks of the web, there's really no limit.

Mmm.

to what you can do on the canvas. And we've seen some pretty crazy stuff from our community in terms of what they've done with the canvas. And we've done some crazy stuff as well, actually. ⁓ that's,

Yeah, I just saw one of your latest video on your channel about having ⁓ a voice agent to control the canvas. This is very interesting as well. I'm definitely not a pro user for canvas. The first use case that would come to my mind would be, as you said, diagramming, drawing, things like that. But you're saying it's basically a piece of paper on your browser that you can do anything with, right?

Hmm. yeah. Like, yes. Yeah, I mean, like...

Whereas a website doesn't do it. A website, you can't really do anything. Like it's fixed, right?

I think the, yeah, there's kind of two things, right? Like one is the website ⁓ is a document. It's like one dimensional, right? You read it top to bottom ⁓ and I guess you can zoom out, it's like a, you're kind of like in the document. You're kind of viewing the document. There's no like space around the document if you wanted to like leave notes or add sticky notes or something like that. Number two, it's like a

Yep. Mmm, good.

Normal website is a private experience. It's not a collaborative experience. ⁓ Tealdraw is a very collaborative experience, and we have an extremely... If it was only an app, I would say it's over-engineered, because it's better than it has to be, but since it's an SDK, everyone gets the benefit. it's appropriately over-engineered, ⁓ really, really fast ⁓ synchronization engine. So you can do multiplayer sessions. ⁓ And it is, how do I say that? Yeah, well, kind of like what you said. It's not designed to be manipulated. You're not moving things around. The web is a document that you're being shown, kind of like a magazine, right? Or a normal website. ⁓ you know, Notion is another ⁓ application that takes a, builds a document out of...

Yeah.

That doesn't feel like a document. It doesn't feel like a website anymore It feels like more like a structured content editor But it is still made out of normal website stuff like it is still the same building blocks as a normal website and as a result you can put like movies and gifts and things in it and and really till jobs the same way like the canvas is it's such a ⁓ Easily manipulated environment for just doing whatever. ⁓ And it has all these little primitives, like rectangles and stuff. people build wireframes out of those things. build. People do like, we have a big Chinese language Bible study community that uses Tealdraw online. People do technical interviews. They do tutoring sessions, a ton of Russian language tutoring sessions for some reason happen on Tealdraw. ⁓

Yeah. It's hard to imagine, you know, it's not something that would come to my mind. How do power people using this? This is crazy. can, you have to tell me, like, if you, if you have any like use cases, right. Show me one. ⁓

Yeah. I mean, yeah. Oh yeah, yeah, sure. let me see. Hang on. I forgot. I always forget that these are video confer videos these days. Let me, uh, let me.

Yeah, and you have... Yeah, I feel like you guys are similar. You brought up Notion. So Notion is like a document, an editor. It started out with more for language, words, but you're way more flexible. You're a blank piece of paper, right?

Mm-hmm. Yeah, I mean it and you can put other pieces of paper inside of it, which is Which is funny. I should have had a bunch of this stuff, know, you know what I can what we don't I'll kind of show you some of our own stuff as well. ⁓ Because this is a ai audience, podcast, correct?

Yeah. Yeah, yeah, AI audience, we would love to learn from you.

All right. Well, one of the things that we've done in TealDraw is always kind of like, mostly because I want to play with, I'm really curious about ⁓ like AI as a technology and the things that folks can do with it. ⁓ So that's kind of been a place where we've done a lot of exploration ⁓ inside of the company. so let me kind of clean stuff up and now I can share. ⁓

Mm-hmm. You have a lot of stuff, I would say, yeah, online. ⁓

Yeah, we, it's, so let me see, this is.

I would say you're the founder who is having most fun that I have.

Yeah, I do have a lot of fun and Tilt-Ret sort of like lends itself to the type of... ⁓

Right, right. You can also show me like fairies. I really like fairies. Like, let's come back to it. Yeah.

yeah, we'll come back to, let me come back, how long is this? This is an hour long, right? Yeah.

Yeah, if you do have a hard stop, if you don't, then we can unite.

Might have some extra time. Yeah. I don't know why my mouse is wrong. I think we have it like this. OK, cool. So the first one that we did that was really, popular ⁓ with AI was this thing called Draw Fast. And in order to appreciate this these days, you have to remember that the year is 2024. ⁓ GPT-4 with vision has just come out. And ⁓ that's the first time that developers have had a chance to play around with vision APIs for ⁓ language models for AI. so here's, I want this to be the, ⁓ I'm gonna just kind of make something. Imagine that this is a camera feed. And then there's a button down here that says, or is like, take photo. And so one of the biggest use cases for tail draw was wire framing and things like that. so we, well, actually it was a designer at Figma who first came up with this, but they used tail draw to prototype it because it was easier to do that. and let's say pixelated.

So they choose this instead of Figma or they're just using this first and then moving on to Figma.

Well, that's well, I'll show you what it never did go into to Figma because the Well, I'll show you the magic in a second So let's say I wanted to have like an interface that looked like this What would What would I do normally is I would kind of pass this to a an engineer or something like that and say hey I have this idea or maybe I'd go in and try and code it myself or something but ⁓ We ⁓ in make real you would you click a button And that button would send the screenshot, right? Because it's trying out. can make screenshots. And ⁓ you can export those. And I guess you could have pasted it into a chat GPT or something like that. But we would take this, and we send it off to GPT-4 with vision at the time. And now I think this is using Gemini 3, so we'll see how it does. And say, hey, me a working prototype based on on this image. I think the system prompt had something like, you're a senior developer, you're 4,000 years old, you work with designers and you turn their sketches into prototypes. You just got this design from your designers. Make it real. Respond with a single page HTML and CSS, whatever file. Again, gotta remember, when we were working on this, there was no lovable, there was no vibe coding wasn't a term yet. So this was like, One of the first things that anyone saw where they could, let's see if this is actually going to load. I might have to reload the page. I have a feeling it will actually use my camera. But we'll see. Oh, no. But this is like, or should be, an interactive website. And it definitely missed some of the details of the prompt. But I can stop the recording.

Did you put in a user prompt just now or this is... Okay, okay. Okay, there's no instruction, okay.

No, it just worked from the image. ⁓ But what's cool is it got it wrong. if I wanted to debug it, the thing that I would do to debug it is is I would ⁓ just draw on top of it. ⁓ And then I want to say this is the actual browser camera feed. I don't know if this will work while I'm also on the call, but maybe it will. So if I just grab all this again, that just becomes the next prompt. ⁓

Interaction. Okay. ⁓

And so the screenshot that it gets this time ⁓ is, of just the drawing, got whatever. It got this screenshot instead. ⁓ And then if you include a previous generation in the prompt, the prompt changes. It's like, hey, you've got some feedback. Basically, your designers came back with notes about how to make this thing better. Please incorporate their notes into your next result. ⁓ And I normally have some very good examples that have kind of been ⁓ cooking in the oven that I can kind of pull out and reveal that we're doing this live. like fingers crossed that this... Yeah, yeah, yeah, yeah. But this really broke the internet when it came out and I've always been very proud of kind of being one of the first...

The person life jumps the hill, ⁓

Pieces of software that used AI to let non or people without programming skills. ⁓ Two things. All right, we're getting closer. ⁓ But.

So, you know, we are very close. What you're expecting is the camera will be working.

Yeah, Even though we built this one a while ago, you could just kind of change the models underneath it without changing anything else. It's not really optimized. There's no like agentic stuff because that wasn't around yet. But we'll see how CloudSonic 4.5 does.

Yeah, I know that this went viral right two years ago in the so if you're looking back today for the past two years, are you are people actually using it or

Yeah. Yeah, mean, an interesting thing happened is like, this got very, popular, it was ⁓ so new that it really couldn't be a product. ⁓ Sonnet 3.5 got closed. Maybe my drawing needs some work, but who knows? ⁓

It feels very close. ⁓ the way to iterate, least you can toggle different models and so maybe give a little bit more instructions, like that.

Yeah. Yeah, exactly. ⁓

Can you add like a user prompt here?

No, but if you wanted to, could just add it on the canvas itself. So like, you know, I did the, hey, the camera should work. You know? It worked.

⁓ I love that. Yeah. my.

And it know it'll it'll will respect this as part of the input ⁓ You know you could also use photos of other apps you could say here's stripe comm you know make it look like stripe comm and Yeah Be sure to edit out all those API keys Come on let's go I I want to see it happen and and the the websites also

Yeah. Yeah. Yes.

You can open them up in Replet or these other websites that you might want to edit the actual code. And here we go.

Yep. There you go. Yes, did it. Oh my goodness.

All right. Cool. Now, do these have any effect? Yeah, hey, look. Well, sort of. There, and then look, this works as well, right? So yeah, apparently you just had to yell at it a little bit. And then look, it downloaded, and here I am. And all that just from drawings. And I could, of course, continue like.

my goodness, yeah. Great.

annotating it or whatever ⁓ and keep going and we've seen some pretty cool stuff made with this.

How usable would you rate the code generated behind the scene here? Like HTMLs.

It's toy code. mean, it's not, but you know, there are some things where, especially since you can just, you know, send it to someone. Like I could send this to you. You could use this if you needed a black and white headshot for something like, yeah. ⁓

Yes, that's Yeah, yeah, yeah. You mentioned like you this this does look like the first, you know, vibe coding tool before lovable and both existed, as you said. So I'm curious, like, why didn't you become a lovable or both?

I mean... Yeah. you sound like my investors. I think there's a couple of interesting things. First off, when we made this, again, the APIs were so new that the only way for us to allow people to work with it was for them to bring their own API keys to the app. I'll just put my head on the screen so we can kind of do this. So like. I couldn't have sold access to this even if I wanted to ⁓ because each API key only had whatever 30 responses per day allowed. ⁓ And when I did eventually get on the phone with ⁓ OpenAI, they're like, how are you doing this? Are you account sharding? Are you registering multiple? I'm like, no, but kind of. It's just everyone's bringing their own account. That also meant that only about 20 % of the people who visited this ⁓ website knew enough about API keys and knew enough about what those things are in order to actually use it. So if you ever go back to the Twitter threads and stuff about make real, it was just like, my god, this is the coolest thing I've ever seen. Hey, how do I make this work? Hey, this isn't working. It's broken. It's broken. my god, this is the coolest thing I've ever seen. Tons of enthusiasm, but they were split up. ⁓ And finally is that like ⁓ Teal Draw, this did not feel like a product, this felt like a pattern, right? This idea of taking... ⁓ taking drawings and diagrams and using them as input into AI models ⁓ felt like 1,000 products, 1,000 applications rather than just one. ⁓ And that's always been my bet with Teal Draws that the thing that we sell is not an application, it's like the canvas itself. And you can do different things with it. ⁓ So just as an example, ⁓ And I think that honestly, even though I, know, Lovable and Bolt and everything are making a lot of money, you know, we're making money too, but not vibe coding up money, although we probably spend a lot less on credit or on tokens. ⁓ I think it's a bigger opportunity, like long term. So I do think that there's going to be and there's going to continue to be a big expansion of apps that want to take advantage of this type of software. And I want to be the software that they build on. ⁓ Here's a good kind of little example, a little interlude here. TealDRAW has really nice geometry system that we use for like intersections and ⁓ like hit testing. So like if you had a nice

Mm.

heart shape, the arrows would gently follow the curve of the shape. Yeah, it's super satisfying. And the way that we actually achieve that is by having an entire geometry system. I just turned on the debug mode that exists just in pure code.

That's very nice. It's very simple, but it's it's detailed. I mean, you were a designer, right? So I think have a bar on this. I mean.

⁓ pure data that we use to detect where things are and how close the user's cursor is to them and all that stuff, right? This is neat, right? ⁓ Someone else took this and his name is Grant Cott in New York. Grant had made this amazing liquid simulation in the browser, right? Like just really, really far out liquid simulation. You know, we didn't make this. This is all Grant's own kind of algorithm for doing

Mm.

WebGL, whatever, GPU accelerated liquid simulation. ⁓ But Grant used Tealdraw as a kind of an engine. Sorry, there's all this stuff on the screen. used Tealdraw as a kind of an engine for working with this liquid simulation. using that same geometry system here. mapped into the the shade or whatever. Now you can you can kind of just take this and kind of yeah, create like vessels and pour things out. In the same way, like it's the same stuff that is on Teal Joe's canvas. It's still our canvas, but it's a totally different and incredibly cool ⁓ use of that canvas because of the kind of the primitives that we've made.

It's like your finger in the water or you can... ⁓

made available, like those geometry systems, being able to drag things around and draw and do all that stuff. I think there's even a whole spout. If you make a check mark, it'll create more liquid. And if you create an X, it'll absorb the liquid. So again, like the. ⁓

this

My bet has always been that by making the canvas available, people will do really interesting things with the canvas.

So this, sound like you're just gonna stay like infrastructure for building on top of canvas, right?

I think that's, yeah, that's been the, that was the pitch when I started the company and it really hasn't changed much since then. We've gotten a lot better at knowing who needs this type of technology and how to sell it and how to connect with people. But ⁓ we've also made a lot of strange demos that show off like kind of what you can do with it. ⁓ You wanted to see the

This is really fun. Yeah.

the fairies, right?

Yeah, I do. But I'm curious, like, why Canvas, right? So there's so many infrastructure you could have built. Why did you build this? Why Canvas?

Hmm. ⁓ I mean, I have a fine art background. spent a lot of my life looking at... ⁓ working with, I guess, pens and paint and doing drawing and painting as an artist. ⁓ And I think when I got into technology, got into design. From design, kind of got into, well, prototyping and programming around design. ⁓ And then I eventually got into design tools themselves. And ⁓ Here, I'll stop sharing actually. I'm talking about the, yeah, the kind of software that is designed to create or support creative experiences, ⁓ which I think is a really hard thing to design for and a really kind of unusual problem of like, do you keep people focused on the things that they're making? How do you not put too much in front of them but...

Mm-hmm.

You know, still give them all the tools that they need. How do you, um, how do you prevent moments where something that they expected to happen one way happens in a different way and, and, and kind of breaks their, creative flow or breaks their attention. Uh, and those are a lot of, they're like, um, how do I say that? I think when I watched someone using Teal Draw, like, Number of things that are going well are like, ⁓ there's so many of them and none of them are being noticed because they're going well. The reason why they're, the thing that happens when they go well is that you're able to continue focusing on whatever you're making. You're drawing, you're diagram, you're wireframe, you're retro and that the tool is just quietly just facilitating all these things. ⁓ And boy is that super complicated to get right all the time. And it's really hard design problem, but it's also a really hard ⁓ programming problem and technology problem, especially once you get into multiplayer. And there's so much benefit from craft and there's so much benefit from attention to detail within this type of software. ⁓ Whereas I think, yeah, I mean, I like my I think superhuman is a better email client than Gmail, but ⁓ I'm also kind of fine using Gmail most of the time. I think, ⁓ yeah, there's always better software, but sometimes it matters a little bit more, and sometimes it doesn't. And Canvas seemed to be a place where all those details really, really did matter. And so, yeah, I guess it just satisfied my Interests in my way of working and just the things that I was good at ⁓ More than anything else I had ever worked on and also no one else was working on this problem in a way that Wasn't just a single product single product single product so from an infrastructure position or perspective. It was just like wow this is a This is a fascinating market. It's growing. It's lucrative. ⁓ I Think I can sell to this market as well ⁓ and I think I can

Right.

go so deep on the problem of how do you build the undifferentiated part of a canvas, build the engine for a canvas that ⁓ that it becomes an obvious choice for companies and teams to do this.

Yeah, it feels like different trends are just merging that there is a digitization of everything like information goes on web and people are globalized and we're more accessible than ever over the past 25 years to other people across the world. And this desire of having a digital version of like a blank page, maybe this is what people want it. And how do you perform? as nimblely and on the digital version of it is meant to happen, perhaps. And you're just enabling creativity. You're not the one who just creates for everybody, but you just enable all the creators. I feel that's your role.

Yeah, thank you. I think a lot of people use Tealdraw for things that they wouldn't consider creative work necessarily, but it benefits from all the same, how do I say that?

That's the wild part. It's just unbelievable. It's more than just a piece of paper. feel like, you know, people can do this, these interactive. It's beyond my imagination. I have to say, you know, like the demo.

One of my ⁓ favorite customers who uses Tealdraw is a data visualization tool. I'll pull up their website here in a second. A data visualization tool called Observable, which I've known about forever because the founders made the D3 library for data visualization. it's like a notebook if you're familiar with Python notebooks and these things. ⁓ ⁓ but for processing like large amounts of data. And one of their ⁓ new products is rather than a one dimensional notebook to expand into a canvas as well. So that the SQL queries, the database queries, the data itself and the visualizations are kind of in this, ⁓ in a canvas where they just sit on the canvas and where they kind of the flows are articulated on the canvas.

Hmm.

And changing one thing will kind of ripple across the graph of dependencies that you've created. ⁓ And that's like not, that's not at all what I had in mind when I ⁓ started building Tealdraw. I didn't never even thought that this would be kind of like possible. ⁓ But it makes total sense. And all of the things that I did while I was building Tealdraw with, you know, maybe whiteboarding in mind, ⁓ benefit in this situation as well.

This is fascinating.

the balance of structured and unstructured data, being able to draw a little circle on top of a graph seems like such an obvious thing to do ⁓ and to be able to share that and stuff. ⁓ projects like this just make me really happy to take something that, for me, have its roots in fine art or its roots in painting and drawing or design. ⁓

Yep.

and let people take advantage of those things for things that are very, still arguably creative, but just very, very different domains. ⁓

It's very different and people are very creative. can, mean, mean, I, I'm a data scientist by training. So I know that, you know, in the, development in these data visualization space was, you know, you started to have the static graphs of the data and then you started having the more interactive meaning that you hover your, your mouse over something and then you have the tips. Yeah. But after that, now this is even further that maybe my, you know, my business partner can come in and make a note on top.

Mm-hmm. Yeah, tooltips and yeah, yeah.

And then it can become even more collaborative. It can answer a question to my business partner without me being part of it. Right. That would be so cool. Oh, this is such a good idea. Maybe I should build a product on your, your, on your, on your infrastructure. And I'll sell it. Yeah. And, know, I want to ask you like who you're.

That's cool. Yeah, you should check it out. It might be right up your alley. Yeah. Well, yeah, it's extremely possible to do that. Yum.

Your main customers are developers and you're just helping every, I think you mentioned last time when we talked that you are helping every developer who needs a KEMAS. Is that your

Yeah. mean, TealTros is like kind of the way that we sell this is a little bit unusual. We don't sell to individuals. We sell to teams. And we don't sell, it's not a SaaS product in the sense of like, I want to have like five seats. I'm going to have five developers working on this. ⁓ We have one price. If you're

developers.

under 10 people and then if you're more than 10 people, we negotiate a price. ⁓ And that has ⁓ like essentially just a ton of flexibility and my salespeople are as close to me as you can get while being an account executive, like very creative in terms ⁓ of how to find the right number for the right team. But we've tried to keep it really, really simple and acknowledge that just different companies are using this for different reasons, at different scales, et cetera. But yeah, the types of companies that are buying Tealdraw are, we have a lot of AI design tools, things like UX Pilot or Variant.ai or Magic Path, where they've kind of operationalized.

Mm-hmm.

some of make real actually the kind of the first demo that I did and said okay what does that actually look like for customers like how does that function in a product in 2026 ⁓ or you know how do we take advantage of the canvas to let people design ⁓ websites that just work you know ⁓ another big category is just general productivity tools especially increasingly internal productivity tools or internal collaboration tools So these are companies that are like, ⁓ yeah, we used to be using Miro, but like, we can't get all of our data out of Miro or we can't plug it into our AI agents. so like, you know, we're, you know, we, talked about just vibe coding our own, but like, it's too much to even vibe code. like, let me just code on top of what you guys have built. ⁓ And, and so that we can create our own internal ⁓ collaboration suite.

Yeah.

So a good example of that actually Shopify does that. So Toby has talked about on the pub podcast about their internal tool set. yeah, give us a little shout out. Every meeting, you get ⁓ all the contacts from all the people in the meeting. And you also have a teal draw where you can do your collaboration and your whiteboarding. ⁓ And yeah, that's probably the most interesting growing category, especially motivated by. ⁓ So that's definitely at the top of my mind. ⁓

That's very interesting. When you're showing me so many different demos from the end user, it's just amazing. It's hard to imagine. Where is the boundary? What if, as a developer, I just don't know what I could or I should use a canvas for my use case? Do you have any suggestion? Because I don't know. This is like a new paradigm.

Yeah. Yeah, So we have, let me show you something. I mean, this is not a, I'm not gonna. do my whole pitch deck or anything. yeah, so if you're a developer, could go to children.dev to learn about the project. But one of my favorite things that we have, which I guess is on the docs, would be the fastest way to get there. So this is our documentation. And we have starter kits, which are kind of fun. For example, if you wanted to use tealdraw inside of the context of a chat, like, hi, what animal is this? And then we'll do a duck rabbit just to see what it comes up with.

you

I mean, that's a type of interaction that, honestly, I wish that more like both selfish. I wish they used Teal Draw for it, also, all right, I didn't draw it well enough. I also wish that just more chat apps took advantage of the fact that people like to draw and they can just paste images on there and whatever. ⁓

Yeah, this is surprising. But once you show me, feel like, oh, I get it. It makes sense because, you know, inside of my chat, I also want to showcase something. And instead of, I could have used, I don't know, another whiteboarding thing and draw it and take a screenshot and send to Claude, right? It doesn't happen like on the spot, but when you show me, it kind of makes sense. Okay, why not just do it right there? Right? You can draw something, pull up a whiteboard. That's the hard thing.

Yeah. Yeah. and that.

The hard thing is when I have a use case, I don't know what I don't know. Like, should I use a canvas? Or you know what I mean? That's what's, ⁓ yeah, that's my challenge of thinking. Your users are so creative. I guess I'm not creative enough. Do you have any suggestions?

Yeah. I can see that. No, think any of the times where... ⁓ ⁓ Where you you want to have collaboration around unstructured data like that that's always a good place to reach for the canvas ⁓ Anytime that you want to have a kind of like a dialogue that involves like visual references Like a kind of like a You know like kind of I guess a one-way Visual reference is like something like a slide deck or like a PowerPoint But if you really want to go back and forth about the same stuff if you want to have a conversation especially remote

Mm-hmm.

⁓ And you want to have that leave a record then Canvas is great for that ⁓ Increasingly, it's like having a dialogue also with with AI models just also to make it ⁓ specific for your audience here ⁓ One of the things that we were seeing a lot of folks try to do with tealdraw is use tealdraw inside of a ⁓ inside of a kind of like a tutoring scenario where I'm gonna try and remember how to.

Why do I do a tutoring scenario? felt like it remotely relevant for a canvas. Why? ⁓

Wait, is the oxygen the bigger one or the little one is the, is this water or is the H smaller? I forget.

It's through all I think that you have to it and the one own

yeah, of course, this is a... Yeah, there we go. All I thought it was the big one in the middle. Clearly ⁓ I did go to art school. ⁓

forgot about our high school math, chemistry.

where you might want to be able to kind of have a chat with, for example, like an agent and have it see your stuff. ⁓ And also change your, yeah, right?

that makes sense. Yeah, that makes sense now. It's an afterthought. That's my challenge, right? When you show me, like, this is right. I need it. But it just, yeah.

I mean, that's the job is that like if I'm making something and I'm telling you, hey, you can do cool stuff with this, then it's sort of incumbent on me to also show you the cool stuff that you can do. So in this case, we have the agent, the kind of like cursor style agent also not only looking at what I've done, but making stuff of its own on the canvas. ⁓ And this is just like, it's so fun where you are able to have this kind of conversation. with model, but also using this other thing as the surface of that stuff. if I'd said, all right, well, hold on a second. Well, wouldn't be a show if I didn't get a bug. I did a, well, why is that? That's interesting that that's, I'll have to look into that after here.

Yeah. Yeah.

I don't know what molecule that would be, but maybe it'll tell me. But you can see it's thinking just like, oh, that's N2O2. Come on. N2O2, And then this was actually the foundation for the.

Well, it doesn't exist, I guess. Yeah. Okay, great.

Well, let me finish the page. If you're interested in kind of seeing these different demos or seeing these different starter kits, what's fun is that you could basically just run a terminal command and you have this whole source code just ready to go. ⁓ And then you can kind of fork it or kind of manipulate it, change it ⁓ based on this as a starter kit. And we have a couple of other ones as well, like for branching chats or for collaboration or for weird...

Right.

shaders where you want to have like a fluid simulation. So there's a lot to play with and there's some good places to start from.

Oh, like, yeah, yeah, that looks really fun. Do you think, you think, know, I don't tell me like, tell me if this is the trend you're seeing. Do you think Canvas is becoming the new UI for AI? Like chat was the default, 2023, like after ChatGPT everybody's used to chat and you see a whole blossoming of products that build on chats. Everything is chat, but you have a dashboard. Do you think Canvas is the next one?

out. So I think that Canvas has a place, ⁓ just like think terminal apps have a place. ⁓ I guess it's question about for whom and in what setting, and what are the different constraints of those different platforms. What I think is really cool about ⁓ the Canvas is that all of the things that make a Canvas good for collaborating with other people make it good for collaborating with AI. So as an example, when I have a conversation with a model or something, if I'm doing some work with whether that's ChatGPT or whether that's like a coding agent or something like that, there's not a lot of room in that experience for anyone else. Having group chats with an AI actually isn't a very good experience. Group chats in general aren't a very good experience because they're one dimensional. Like every message kind of pushes to the end of a big list. ⁓ the canvas, why the canvas is one of the reasons why the canvas is so good, why it's so popular in situations like Figma or Miro for collaboration is that everyone can kind of be working at the same time in parallel with each other. And there's really no limit on how many people that paradigm can support. Because if it's busy, you just go somewhere else, right? If everyone's working over here, you just start working over here. ⁓ And we did an experiment towards the end of the year using ⁓ AI agents that were spatially located, that lived on the canvas. ⁓ And they were kind of like virtual collaborators, except they were definitely not as good at anything as people would have been. ⁓ So instead, we decided to make them into fairies. ⁓

Yeah, I bought a fairy for my daughter and then yeah, they were playing with them like, the leaves are so cute.

⁓ awesome tension. Love it. I love it. I love it. I'm so happy to hear that ⁓ You can you can choose their hat and you can choose how long their legs are Those are the only and I guess what color their head is ⁓ But that's I'll make this one a short-legged ⁓

I have to say I feel kids are more creative in using these tools than an adult like me.

⁓ A lot of the design of these things is heavily influenced by whether my daughter can play with it or not. ⁓ So these are AI agents, just like the ⁓ AI agent that I showed over on that other website. they're located. They're on the canvas. You can move them around and you can select them. And you can have a conversation with one of them, or you can have a conversation with all of

Okay. ⁓ All of

And when you think about the parallels, and we did a lot of thinking about the parallel, you know, ⁓ thinking of these as agents, like, that type of... conversation like with multiple agents at the same time or the way that they self-organize and things like that I think is these are really hard problems to solve in any situation ⁓ but they some of them are problems that like have been solved on the canvas already ⁓ so some one of the questions is like where is my agent you know what are they working on or something like that and and on the canvas they're working on wherever they are like

Yeah.

And if two things are near each other, there's a kind of presumption that they're working on the same thing. But let me show you this, like draw a cat, right? Very similar to the chemistry stuff that we were doing before. The other thing is like, is an agent doing at this moment? ⁓ Exactly. And those states actually represents real things that the agent is doing, right? Are they ⁓ editing their to-do list, or are they working? And they'll kind of like do this when they work.

They're taking notes. looks like they're taking notes.

⁓ And or are they you know reviewing their work? They'll kind of like you know really put their hands on their hips and kind of like think about it ⁓ And even though these models are absolutely not trained on how to be ⁓ artists at all ⁓ You know we can we can have them be be creating on the canvas in the same way that ⁓ that I can and ⁓ You know and it can do the things that that I can do move the cat onto the box, ⁓ and it can kind of understand the canvas and read it ⁓ in the way that a collaborator could do, which is great, because they can also kind of work on top ⁓ of each other's work. So put the shoes on the cat. ⁓ I'll let them move the legs over. ⁓ So even though this agent didn't make the cat drawing, it can still see the cat. It can still navigate the canvas.

The agents are moving, the fairies are moving them one by one, one stroke at a time. That's so silly.

There we go. All right, cool. She was on the cat. And now if I wanted to do something with all three of them, I could do that as well. Make a few more animals. So when working together, when addressing all three of these things, they work a little bit differently. This one, one of them gets elected as the leader. And it creates a plan.

Mmm.

their wing color changes to indicate that they're part of a group. ⁓ And then now they can, ⁓ this one will begin giving tasks to the other ones, which they will complete ⁓ in parallel. And actually we do have an IRC chat, a kind of a running chat and activity log of the different ferries as well, if you want to find it this way. So we have a chicken. We have ⁓ another cat, and there we go. ⁓ Now the first ⁓ one is kind of like reviewing everyone else's work and deciding whether it's done or not. And it decides it's ⁓ Agents on the canvas working together. ⁓ It's fantastic. ⁓ And I can even give these folks pretty complex stuff.

Yeah, visualize. This is visualize the Asians.

Here I'm going to ⁓ give just a bunch of mermaid diagram text. So this is kind of like a notation language ⁓ for graphs and stuff. in order to do this, they need to know a lot about mermaid, ⁓ which I didn't tell them about. They need to know about what

Mmm.

understand what it is and the relationships between all these different things. ⁓ But this is much more ambitious task. So they're starting by getting rid of the... clearing the canvas a little bit. And then now they are getting to work on the various parts of the diagram. And you can kind of see... ⁓

Do the clean up.

also see the limits of their abilities as canvas artists. But let's see. We have a recording and these are, yeah, you all the arrows are connected to the right stuff. ⁓ But I think I'm going to have to follow up with a, hey, make it bigger. ⁓

Nice. Is the project still alive?

No, they were only ⁓ live on tealdraw.com for December.

Can I ask why? Like why are your some of your projects like fun projects? Those you you charge $25 but limited time only.

Thank Yeah, this is the first time we've ever tried that. And the main motivation there was ⁓ that we wanted to give this to as many people as possible as a thing to experience. ⁓ We never charged any money for these experiments. We haven't talked about it yet, but Teal Draw Computer, this whole thing that we did with Google that uses Gemini Flash to create these executable graphs and everything like that. Everything has always been free and that's been fine, but ⁓ this was the first one where it was like, wow, we're kind of burning a lot of tokens here. ⁓ I don't think we can give this away for free. ⁓ But also I was just curious to see ⁓ to whom this would matter and who would be really like, ⁓ yeah.

Can I ask if it would turn out to be profitable?

If you ignore all labor, then no. It was still not profitable, but it ⁓ was pretty close. think we spent more on, if it wasn't for labor or lawyer fees, I think we'd be fine.

I can't Is it a matter of marketing or do you feel actually enough users were using it? It's just still not enough.

No, I think, well, there's kind of two. Most of these projects ⁓ in my head are marketing. think marketing by building is a really great way to market a product that I'm hoping that people build on top of. ⁓ And sorry, this is making an ⁓ extra teeth for sharks as a product. So it's.

Mm. Computer is very interesting. looked into this and I saw some demos, but honestly, I don't know what is it for. It's connecting the dots and it's doing a lot of things. But what can an end product be?

It can be a lot of things. can be an invention, inventor. I think of this as like, ⁓ again, kind of like a toy. It should, for a certain type of person, a product-minded software developer or something, ⁓ this product should inspire like, I never thought, you know what, you could probably do this. You know, this is kind of a joke, but like if you wanted to do this for real, ⁓ you could absolutely blah, blah, blah. Like, you know, if you put this in a classroom, like how would that work, you know, and blah, and kind of get people thinking more about what is possible with AI.

We'll you I feel like it's kind of liberating in a sense, because for learning how to use the ⁓ internet and then learning how to get used to the website, the applications on your mobile phone, then get used to the drag and drop, the menus. then consumers, we are used to these. I feel like we are used to being bounded by these things. But this is like, you can do anything basically. Like you're only limited by your own imagination.

think if I've done my job right, you should be pretty able to compose.

I have some sense. I'm inspired, for sure.

⁓ If people want to play with this, by the way, it's computer.tldraw.com or tldraw.computer and it gets weird and it gets really, really fun to create these kind of like little machines, sort of. Have you gotten a good demo of this at all?

Mm. I saw some of your demo on your site. Yeah, so it's pretty cool. ⁓ It's connecting all kinds of stuff together. yeah. But if you want to show one, like show part of this graph for my audience, they'll know.

That's good. Awesome. All right, let me show a quick. It's a good film that is recognizable back to the future. ⁓ Right. So these are all little components on the canvas that can run, can execute. ⁓ And you can think of them as steps ⁓ in a ⁓ workflow. And each one of them does something and then pushes the result out into one or more other things that are connected by arrows. So the way that this one works is I'm going to click Play on this. It's going to pass Back to the Future in as the of the arguments to this block. Write 10 detailed visual scene descriptions that tell the story. And the agent is going to get this together with this. And so it'll do that for Back to the Future. ⁓ And then it's going to push. one of them into this image, but this image will create an image prompt and then create the image for the image prompt, but it's going to receive scene one and the list of 10 instructions, right? And those are going to be its input. Hey, create an image based on this list of 10 things in the words scene one. And when it's done, it's going to trigger this thing. which is gonna receive that scene one ⁓ and output either yes or no. ⁓ If it outputs yes, this better have worked, if it outputs yes, ⁓ then it will increment the scene, right? This one's also receiving scene one, but it's writing back into the scene one, which will cause this to run again, looping until this scene goes from 1 to 10. So we just saw it do that analysis, is seen more than 10. The answer was, ⁓ or less than 10. The answer was yes. So this ran, so this ran. And so now this is running again. And thus, we have a little machine that will tell us, or show us 10 scenes from Back to the Future ⁓ in order, I hope, based ⁓ on just ⁓ the title. So it's like a little machine. It just kind of goes. the thing that is going, the thing that is happening ⁓ is done just with AI. And you get to do these. ⁓ I this is probably the world's most expensive counter or something like that, or like loop. ⁓ But you can do things like, is this image good enough? Or is this image funny? Or does this job description match this role or something like that? ⁓ You can do these kind of squishy logical things as part of the machine as well.

Good point. ⁓ Can you take in data inputs? Like you mentioned, I can compare, let's say, resumes to job descriptions, see if this is a fit. Then I have a data set with all the resumes. I need to pass it in here. ⁓

Yeah. Yeah, you can. Yeah, we have a lot of different ⁓ components here on the right. ⁓ And one of them is, ⁓ let's see, ⁓ data. Yeah. And so this actually, you can even get hit an endpoint ⁓ and pull things this way. You can also get a file, and you can add like CSV files or other things ⁓ here.

⁓ cool. You know what, do you know NAN? that's a, this remind me of NAN. So which means that I can just use NAN endpoint here as a data input, whatever I process from NAN, I can visualize it just using this. Cause NAN doesn't have this, right? They don't have this. But now I'm creating another layer of my data processing pipeline. sorry.

Yeah, yeah, yeah. You could try. Yeah. ⁓ I would love to see it. And then, if you wanted to add more to it, if you wanted to create a 3D model out of your pipeline or something like that, you could. ⁓ Yeah. I think when it comes to, yeah, these are very, very conversational, very kind of like, it's not a, yeah, we worked on this app for like six weeks, ⁓ but it is, ⁓

Yeah.

It is surprising how far you can get and how fun it is to kind of click these things together. I have a blast using this app.

Yeah, that's why you know you are the number one most having fun funder, a funder I've talked to is again like I have to have to acknowledge that you're the only probably you're the only one who's having so much fun. Yeah.

You Well, the creativity balances all the other challenges and stuff like that.

It's crazy. ⁓ Do you think, do you think you're not going to just be limited to developers? lot of these things, it seems to be fairly easy to navigate for vibe orders for PMs. Do you think that you could be for everyone? Like cameras could be for everyone.

Yeah. I think we're gonna, like my guess is. Probably yeah, I mean we have tealdraw.com is like an app that you can go to and like it's it's very good

Going to tld- ⁓ tldraw.com I don't get any of these. I don't feel like I-

No, you don't because it's like these are all these like little side projects. know. It's 3D model. ⁓ Because they're not like they're not really not all of these are really like products in the sense of like solving a particular problem. ⁓ They're more of like playgrounds or demos, which ⁓ you know, the problem that those things are selling is that

I have to dig it up or talk with you to know about these things. So cool. Wow, yeah. Right.

Kind of to your point earlier, it's hard to imagine what you could do with a canvas, given how conservative the uses of the canvas have been in the past. Yeah.

Exactly. People are conservative.

Well, canvases have been so difficult to build for the last whatever that I think a lot of the use cases that are out there are just very safe. mean, they're whiteboarding. ⁓ Yeah. So I do want to of crack open the category a little bit. it does, ⁓ yeah, it takes some, it's my job to also kind of inspire people to make some stuff around it.

Yeah, it is. It's probably, yeah, yeah, yeah. You have to be the inspiration for all of us who wants to build something. You know, I was interviewing a CEO from a company called Hi Touch. And he shared with me, just this Monday, and he shared with me that within his team, designers are now the first group beyond the engineer team, shipping features within their company.

I've heard that from a lot of teams, yeah. ⁓

Are you seeing anything like that among your users, non-engineers building with TLDraw, like designers?

I mean, especially with Make Real, that group who, I don't know where it went. ⁓ Make Real was great for PMs who just wanted to create user interfaces. I'm not surprised that remains like the killer audience for Loveable, et cetera. ⁓ With this one, this one got really a lot of interest from like people who are doing marketing or needing to do multi-stage prompting around text to say, ⁓ OK, what about if I, here's a thing that I wrote, ⁓ how would I adjust this input for various ⁓ other audiences, right?

Yep.

How would I adapt this to this type of a customer profile? Or how would I adapt this to this type of company? ⁓ And building these very complex, kind of multi-stage prompting ⁓ flows on Tealdraw as a way of prototyping those outcomes, but also kind of prototyping the thing that they were going to go build and code. ⁓ And that this was Yeah, this thing of like, yeah, how would you do that? Okay, let me just knock it together in Tealdraw. And then I can either refine the idea or pass it off. And so it kind of became like like computer specifically, it's not like a way of writing, ⁓ like kind of load bearing programs, but it is kind of like a design tool for multi-stage prompting, or it's kind of like a, maybe even a design tool for, ⁓ for doing some agent stuff as well.

Yeah, yeah, it feels like a canvas visual layer together with something like NEN. You're pulling these steps together, but with the visual layer.

Yeah. And you know, and if you need to write and draw, you totally can. And that's the fun part too. You're always able to kind of break out of this structure.

I'm. Yes. Yeah. Hi touch was telling me, ⁓ know, this is what they're doing. They're doing like they have, they have the data about ⁓ brands, customers, have customer data, and then they're using customer data to build personalized marketing campaigns for their clients. Right. And part of it is generating images like on brand that would bring their ⁓ assets.

Mm.

for ads, like creating ads, including the copy and including the image, things like that. I mean, I could have built something like that on your infrastructure so that it's even more flexible. I guess, right? Because it's having the image for the ads copy. I'm sorry, I feel like I'm still bounded by my creation. I have to be more creative, but this is the one like immediate use case that came to my mind.

Yeah. I don't know, go for it. ⁓ This type of input into many, different types of visual outputs, think, is one of those things that Teodra computer is such a cool ⁓ way of exploring how you would set that up. And sometimes, least, you could also just use it as the place where that stuff happens. ⁓ But definitely, for deciding what goes into that, that type of workflow or what it needs to be. GeoDraw is like a great kind of creative space for that. So in this case, I'm going to ask it to like, make me a Dungeons and Dragons character sheet and then an image from that and then like make a story and then show me all the different scenes from the story and like, yeah.

Yeah. I feel like I can have my daughter try to use this pipeline for maybe writing a book. She's like really into reading.

Go for it. ⁓ I don't want to contribute to the amount of ⁓ bad AI-generated content out there. as a thinking space, ⁓ Tealdraw is fantastic. All right.

I mean, you've built so many visual tools. As a designer, engineer, and a dad, how do you think about raising your kids in the age of AI, or just about raising a kid in the age of AI? Do you try to make your daughter AI native, or are you more cautious about it?

Yeah, in fact, have a, here, I'll show you the, I don't think the camera is gonna focus on this. of the, yeah, we spend a lot of time kind of with tools like TLDraw Computer or, you know, with, I wouldn't say maybe a lot of time, but this is something that I think is, I'm very happy to be able to know enough about the technology to kind of, safely bring someone or like give some experience with this because I think it will end up mattering quite a lot. ⁓ Whether someone is like comfortable with these tools or whether that's afraid of these tools or whether they're afraid of you know or they think that they can do too many things or too few like I think there's a like having a familiarity a kind of like a numeracy with AI, think is something that I can give my kid that I realize that not a lot of parents can. And so I certainly should. ⁓ Even if it starts off with ridiculous ⁓ images and stuff like that, it's still a good, ⁓ I think it's an important thing to be doing. But yeah.

There's ⁓ a huge amount of uncertainty and I always think that there's a fair amount of fear, especially in the US market. for, I don't know if you've heard in the US market, last year, there was a lot of layoffs in the fan companies. That used to be the fresh graduate. They all want to go there for the tech companies, Google of the world. ⁓ But all of them are laying off people, lot of uncertainties.

Yeah.

How are you preparing your kids for this, for her future? Yes, future of work, future of living.

for the future of work. I don't know, I'm kind of like positive on, it's probably an unpopular view, but I think that what I've seen with, even in like the last two months with people coming on to Claude Code 4.5, which if you're viewing this in the future, seemed like a tipping point moment for coding AIs and kind of. and more than just coding, like, ⁓ boy, was that, has that been like a change. like, ⁓ I think what I'm recognizing is that like, wow, there's actually a lot of work to do. Like there's like, there's a lot of stuff that like we were never trying to solve. Like in my startup and my team, in my mono repo, in my like life, like a lot of like tedious knowledge work, a lot of like things where, wow, if I had like 15 really smart consultants, like I could do a lot with that. And I have a feeling that, well, yeah, and also that like, you know, now that, whatever, just it's better at programming than anything else right now, but like now that programmers have these tools, What do they do with them? Well, they start engineering on top of them and with them. It would be like, ⁓ someone's handed me just like God in a box. And I'm like, all right, well, how do I make it faster? How do I run two of these things at once? ⁓ How do I optimize my God box in order to make this thing more powerful? That the ⁓ impulse to innovate, the impulse to engineer, the impulse to like, it's just like.

Do more.

Yeah, do more and to do better and ⁓ to go deeper and stuff like that. ⁓ I think what we're gonna find is that we've actually been having a really shallow engagement with the amount of knowledge work that could be done historically and that it's probably.

Hello.

much, much, much deeper than we've thought about. And that, the kind of the shallow engagement, know, the work that we've been doing right now, ⁓ will probably be able to do using fewer resources. that like, wow, it turns out that like, we weren't nearly boiling the ocean. we could have been doing so much more. Like, our business decisions could have been better informed, our kids could have been better educated.

Yeah.

are politicians, could have had better data, everything ⁓ that could benefit from essentially just applied engineered knowledge work ⁓ will be done and there will still be more to do. So I'm slightly booster on it, yeah.

Thank you. Thank you for being so positive. I mean, ⁓ the world will become a better place, hopefully. Yeah.

I hope so. ⁓ Or else, you know, paper clips. I hope, yeah, hopefully not.

Are you using any AI tools day to day? What are the favorite ones? What do you use?

tons. yeah. ⁓ Cod code is my main one, but cursor and ⁓ I use a lot of Gemini Flash. Gemini Flash is a really great model. ⁓ Multi-modal, super fast, cheap as hell. ⁓ I highly recommend Gemini Flash. I go to Gemini Flash for a lot of programmatic, analytical type of stuff. sharing here.

I have a traditional closing question. ⁓ If you would recommend a book or some form of wisdom, a book, a podcast, or any content to let's say the last new grads starting their career, maybe one to yourself today, someone like yourself today, me, ⁓ what would they be and why?

Gosh, okay, that is a big one. ⁓ I, This is gonna be kind of a boring answer, unfortunately. Well, you know what, I'll do this. I'll do the kind of the sci-fi one. ⁓ I really liked ⁓ a series of books written by Kim Stanley Robinson called ⁓ Red Mars, I think is the first one. I him in university. They're big chunky books. He's a geologist, so there's a lot of text about rocks in there. But the thing that I liked about this ⁓ series of books was that as a narrative conceit, People lived a long time. The first page of book is like, yeah, don't ask us how, but we've extended life to 500 years. And what I really liked about the book is that because folks just had more time, ⁓ there was a complete fearlessness about ⁓ learning something new. The idea that, well, I've been a physicist my whole life, but now I need to solve a problem that requires me to have a PhD level understanding of biology.

Thank you.

All right, I got time. Let me spend the next 12 years studying biology in order to be able to solve this problem and being able to think of those long-term thoughts and again, the fearlessness about spreading yourself too thin or not being able to engage deep enough. Who cares? Life is long and I think that that is That's always been a very, very inspiring thing for me.

I guess the current technology advancement, tools like TeoDraw or ClockHolt are kind of extending our lives in a sense. It's a matter of you go try it out, right? Maybe not for a proof-deed, but you can try other things.

I think. I think the, I mean, I've never been a, I never thought that I would be doing firmware for little development boards, but I'm doing that now thanks to AI. yeah.

you Amazing. Yeah. Thank you, Steve. I mean, what strikes me? This is one of my favorite conversations again. You what strikes me most about this conversation is how much you care about the craft. I think you care about how a line feels, how it curves, ⁓ and getting all the details right. And not only that, it's the vast amount of imagination, of creativity that's potential there. You just show me. I feel like...

Thanks.

I need to push my boundary of how I imagine things, how I can create things. And it's not pen and paper. And it's not just like the drag and drop menus on websites. It's none of those anymore. I also feel like, you know, know, a lot of people are chasing the next AI hype cycle. I feel you're adamant about

I think it can go a lot further.

building infrastructure that lasts. I really respect that. Yeah. And I don't know if Canvas really become the next UI paradigm after chat interfaces. You are going to be at the center of that thing, right?

Thank you. ⁓ I'm building for it one way or another. ⁓ Yeah.

I cannot wait for you to see what you ship next. Thanks for going deep with me today and all the demo, all the fun demo. So fun. Thank you so much.

Thank you and take care.

Don't don't