00:00:00And like vibe designing and vibe coding to me is the closest thing to Flash I have seen in like two decades essentially.
00:00:07This is Tom Cracker, the CEO of Pencil.
00:00:10Pencil.dev
00:00:11And that is called Pencil.
00:00:13New app called Pencil.
00:00:14Pencil.dev
00:00:15I'm starting to feel like we're now discovering these like new user experiences for interacting with AI that were never seen before.
00:00:22You know, and these cursors and scanning and all of these things are like just the tip of the iceberg and we'll see a lot more of that.
00:00:29Like Banana or like these image generators generate like great photos, but like what about illustrations?
00:00:34And I think illustrations are going to be huge in the next half of 2026.
00:00:39Are there specific models that are significantly better than others at design?
00:00:45I think my favourite is...
00:00:48I'm just going to ask you an icebreaker question.
00:00:51Back in the day, you were very familiar with Adobe Flash.
00:00:54Do you still know ActionScript?
00:00:56Yeah, 100% like I grew up on Flash.
00:00:59It was like my first sort of break into tech and I started using Photoshop when I was seven.
00:01:09And then eventually my parents had a design agency, you know, so I would always like after school hang out there with designers and they would teach me all sorts of different tools, you know, from Photoshop, Illustrator, CorelDRAW at a time, you know, and it was all for print design.
00:01:26And, you know, Internet came along in like late 90s and I got really interested in like more interactive stuff, you know, so I started digging into HTML and it was like DHTML and all of these things, you know, long time ago, you know, and Flash came along, you know, it was like Flash version three.
00:01:45And I was like, wow, holy crap, what is this?
00:01:48You know, it's moving, it's animated, it has audio in there and all of these things, you know.
00:01:55And I immediately fell in love with Flash and started building websites in Flash and eventually won like an award at this conference in Prague called Junior Internet.
00:02:07And if you ask around, you know, people from Prague about Junior Internet who remember that, it started a career for many, many folks and eventually started building, you know, Flash was the only way to stream video online.
00:02:25So I started building video players for all sorts of news outlets and so on.
00:02:29And that was my introduction to Adobe because we started using a lot of Flash media servers and so on, you know, and Adobe saw that and eventually they offered me a job to become a Flash evangelist for Adobe.
00:02:44So that's sort of how it all started, you know, and then obviously I started building, you know, design tools and talk at like design conferences and talk to tons of designers and so on.
00:02:55So, yeah, that was the origin of my of my story really.
00:02:59So I've also seen a few videos of you doing conferences at game events and stuff.
00:03:06Did you make games on your Adobe as well?
00:03:09Yeah, yeah, totally.
00:03:10So Flash was used for a lot of interactive stuff, including gaming, you know, so eventually I became sort of a gaming evangelist at Adobe and I was helping various companies to bring their apps or games into Flash.
00:03:26So I would go and speak at various gaming conferences as well and talk to a lot of game developers and help some of these game developers to port games into Flash, you know, and like famous one is, for instance, Machinarium.
00:03:40That was like one of the first games that they ported it from Flash into iOS through Adobe Air, you know, and it was like one of the first breakthrough games of Flash running on an iPad.
00:03:55And it was like a big moment.
00:03:56It became like a game of the month, you know, or Apple for the month on App Store and it became incredibly popular and so on, you know, so, yeah.
00:04:04Were you also part of the Newgrounds community back in the day?
00:04:08No, what is it?
00:04:11Newgrounds was like the website where all the Flash games and Flash movies were listed.
00:04:18It was like, before YouTube, very popular.
00:04:21I thought that was Miniclip.
00:04:23Oh Miniclip, I remember Miniclip.
00:04:24Miniclip was the one for me.
00:04:28But I was spending like all my days on favorite website words, you know, if you guys remember the FWA and so on.
00:04:35But now we're going like really, really far back into the history of Internet and interactive web, you know, so, yeah.
00:04:43And I guess this was the early building block of what eventually led to Pencil.
00:04:49But then you also built a kind of script or a tool, I think it's called DrawScript on Illustrator.
00:04:55Can you talk about that?
00:04:56Totally, DrawScript was like a plugin for Illustrator and it allowed you to convert drawings into basically JavaScript, you know, and then essentially take that and it would paint, repaint whatever you had in code.
00:05:18So that was the idea.
00:05:19And so because I've done a bit of research on your kind of journey into kind of design to code, and I think what would kind of jump in your career, but what made you want to have designs be codified?
00:05:35Like what led to that desire?
00:05:37Well, there really wasn't any solution for bringing drawings into code.
00:05:45And you basically wanted to have like icons and other things, you know, and this was like obviously SVG was very limited at the time and maybe it wasn't even really like supported much, but you wanted to draw certain things on canvas to the canvas in HTML.
00:06:04And this was like one of the things that would allow you to do that, you know, because like once you had it in code, you could make it interactive, you know, so you could turn it into games, you know, and so on.
00:06:15And work with it, you can make it make a move, make it available for, you know, animations and other things.
00:06:23So, yeah, there was this was that sort of door to that.
00:06:29While you were at Adobe, you also worked on Adobe XD.
00:06:34What was the original vision for it?
00:06:36And was it like a replacement for Dreamweaver or how did XD came about?
00:06:42So it was about a time where like there was Sketch, Photoshop was doing also being used a lot for, you know, websites and so on.
00:06:54But we realized that it would be great to have just like a completely fresh offering that was just designed for this specific purpose, designing websites and apps, especially it was the era where like everybody was building mobile apps and so on.
00:07:12So we really wanted to take a fresh look on all these things and really come from like first principles.
00:07:18Like what would it mean, you know, if we could start completely from scratch, you know, so we put together these like principles that it should be like lightweight and approachable and fast and all of these things, you know, and that you can like draw like tons of artboards.
00:07:35And like in Photoshop, you could have just like one thing right in the beginning and that would be it. But like what you really wanted was like show the whole flow thing across all of these different screens, you know, and so on.
00:07:49And Adobe XD was the tool that would allow you to do that, you know, was designed for that specific purpose.
00:07:54You know, there was also prototyping as part of that.
00:07:57So you could connect these screens together, you know, at scrollable areas, you know, and so on.
00:08:03So, yeah.
00:08:05I'm going to jump a bit here and just go straight to Pencil, because I think it's nice to see your journey from Adobe to building this thing, which is kind of like XD, but it uses a lot of AI.
00:08:17And so at what point did you start to realize the power of LLMs and agents?
00:08:24So there were a couple of startups in between, but eventually after all of that, I sort of started digging into cursor and cloth code and all of these things as everybody else, you know, in the world, you know, and was actually building something else, you know, but realized that, you know, while I was in cursor, I was like, oh, my gosh, like, how do I like tell the LLM that this button should look like this that I have in my head?
00:08:52Or like there is a navigation bar and it should be like sticky on the top, you know, in the navigation bar, there are these buttons and so on.
00:09:00I felt like I was writing essays to LLMs of text and tried to prompt them, you know, to do something I really had in my mind visually.
00:09:09And I looked around and there really wasn't any solution.
00:09:11I looked all around, you know, the VS Code marketplace and other areas, and there really wasn't anything where you can just like simply draw something and tell LLM do this.
00:09:23And this was like sort of like pre everybody did MCPs and all of these things, you know, so like initial idea was really like I have this JSON format and LLMs are really good at reading structured data and turning that into any kind of code.
00:09:38So build a prototype, put it out.
00:09:41This was back in May 2025 and it went viral.
00:09:45It got like one million views over socials, you know, like 850 on X and like I think like 400K on LinkedIn or something like that.
00:09:53And I was like, wow, it seems like this is not just my problem, but other people probably want to do something like that and they need this helping hand.
00:10:03And nobody was really providing that helping hand of going from like a quick visual into code.
00:10:10And once I was, you know, once it got built, you know, like it gave me all of these other ideas that, for instance, like the other thing that like completely opened in front of me was like I can actually generate these designs.
00:10:23You know, it's not just for me like to draw something, but I can tell LLM, hey, draw me this, you know, and generate me this and this and that.
00:10:30And that's where like I realized like, wow, we're onto something where we're at the brink of like vibe designing.
00:10:39Was that demo the design mode for Cursor demo, the one that went by?
00:10:42Yeah, it was called design mode for Cursor.
00:10:45Yeah, and it was basically and we still have that essentially.
00:10:49It's like it's a canvas within all of the VS code based IDEs, you know, so cursor, windsurf and digravity and so on.
00:10:59And it's basically like a custom editor within your IDE that's visual.
00:11:07And you can do whatever you want, you know, and then you can just select something on the canvas and you can either copy and paste it into chat, you know, then it has a link or you can just point at it and it knows what you have selected.
00:11:21And it will start basically drawing information out of that and turning that into whatever you want, honestly.
00:11:28Like it doesn't matter like if it's React or if you use any other library, it's basically this like technology or platform or like framework agnostic representation where you can like quickly ideate, you know, and then turn it into whatever you want, you know.
00:11:45So does that mean it's using plain CSS or is it like integrating libraries like tailwind or shadcn or something like that?
00:11:52You can choose whatever you want, honestly, you know, like under the hood is just a structured JSON that basically defines like the look of whatever you're designing, you know, and then we have some baked in rules, you know, for conversion into tailwind, for instance, and so on.
00:12:10But you can write your own set of rules that fit very specifically into your own basic code base, this pen file that we call it, it lives in your git repo.
00:12:23So like any engineer and any AI agent has access to that directly and it can just convert it into whatever you want.
00:12:32So you can write your own set of rules around it and make it like really yours.
00:12:39And I think that's the key point.
00:12:41The first time I used pencil, I thought it was really magical and I'm trying to figure out how it works behind the scenes, like tech wise.
00:12:49I've seen the JSON file in the pen file, but I don't know how it's able to translate that into the UI that I can kind of interact with and like drag boxes and pull boxes and change background colors.
00:13:01And then also the agent can understand it and do the exact same thing.
00:13:05So how did you come up with that union between human designing and agent also designing the same file?
00:13:12Yeah, so multiple things, you know, like we really wanted to build this like incredibly familiar environment that you can just pick up from like day one, essentially, from like the first minute you open it.
00:13:23It's like, OK, I can orient myself in this, you know, in this like canvas where I can like zoom in, zoom out.
00:13:29You know, I have all the properties on the side, layers panel on the other side and just like make it make it like super familiar.
00:13:36But at the same time, it opened up a lot of these new things that nobody really cracked before, which was like, how do I like check with the design essentially?
00:13:47And that gave us a bunch of opportunities to innovate, you know, and the latest one that we just showed in that video was like these multiple agents swarm of agents working on the design in this orchestra, in this concert.
00:14:02You know, and so and there is actually honestly like there's so much more that can be done like every day.
00:14:10I'm like discovering new and new things that previously were not possible just because first like elements are now opening with every new model release, new opportunities to do to do more.
00:14:24And also we just keep coming up with new ideas as we talk to users and how they interact with things and so on.
00:14:32And now they want to like control it remotely and all sorts of things like this is like so much more that can be done.
00:14:39It's kind of crazy.
00:14:40I was curious what model you started on in terms of when did you realize models could actually be good at design because the first few ones weren't that great.
00:14:50And then I think they sort of focused on actually adding some into the training data.
00:14:54Yeah, it there is still a gap.
00:14:57I think it's getting better with every model release.
00:15:01I think we started it was either Sonet 3.5 or 7, maybe 7, 3.7.
00:15:10And initially you had to tell a lot, you know, how things should really look.
00:15:18And as we progressed, we basically are now at this phase where you can just say design a website.
00:15:27And it will probably do a fairly good job.
00:15:31And like the second thing that was like a big switch there as the models and agents frankly evolved was this thing called elicitation where it like asks you extra questions about like, so what do you actually want?
00:15:46And like it's really similar to like you being a client of an agency, really, because like the first thing that a designer or a project manager at an agency would do is to have basically an interview with you and try to extract as much as possible out of you so they can build the right thing for you.
00:16:10And we can't expect that. I mean, these are ones that are now like incredibly powerful, but I think this dialogue that's happening right now more and more with with the agents is a huge unlock.
00:16:22Because in the beginning it was like, design me something, it would design something you would be like, I actually didn't want this at all.
00:16:29But now through dialogue, it will it will extract what do you actually really want and how you want it.
00:16:34And then it does. It's getting closer and closer to doing the right thing for you.
00:16:40And that's that's that's a major difference.
00:16:43I imagine you've tested a lot of models with pencil.
00:16:48Are there specific models that are significantly better than others at design?
00:16:54I think my favorite is Opus 4.6 and I think just like observing like a lot of the folks in the community, that's basically the staple right now, like state of the art for for design.
00:17:10It works really, really well. It works really well with structured data as well.
00:17:15And with the layout, it makes the least amount of mistakes and things and so on.
00:17:22And for a lot of the other models, we basically have to correct some of the behaviors, you know, and so on.
00:17:28And and it's fine. You know, it's just a matter of like discovering these like nuances and making sure that we correct them or that we tell the agent, hey, you've done this wrong, you know, just like do it one more round.
00:17:42And then it usually like happens. The only challenge with Opus is that it's also fairly slow, although it's like amazing in a way.
00:17:54So like the first time I saw like real magic was when composer came out from cursor and I was like, wow, this is fast.
00:18:04This is like designing as I'm thinking about things.
00:18:08So what I'm hoping to see as we progress is that obviously there's still a gap to get like better in the quality.
00:18:19But I'm really, really hoping that these labs are now going to focus on the speed.
00:18:24And obviously it's kind of happening with Sarah Brass and Spark bottle and all of these things.
00:18:30So I'm just like really, really hoping that like the next phase of innovation is going to be focused on speed because that's going to unlock you want to stay in flow as a designer.
00:18:42And it's oftentimes more important to do more reps than like getting it right at the first time.
00:18:51Like you want to be involved in that process and these faster models allow you to do that.
00:18:58And for us, this is also like really important principle when we're building things to make sure that we keep you in the flow.
00:19:06So that's the big difference between like I go to one of these vibe coding tools and I write a prompt and then I go grab a coffee and come back 10 minutes later hoping that it was built and it built something that you didn't even want.
00:19:19Like vibe designing is different.
00:19:21You kind of want to see things happening in real time.
00:19:24So that's why we have put there the cursors that are now like moving around and you can kind of see what's going on, how the model is thinking about these things.
00:19:33Like being very communicative towards the user is really important when things take a little bit more time than they should be.
00:19:42But I'm also hoping at the same time that these models will get like so fast, it's just going to be like appearing in real time.
00:19:49And like one thing worth saying there is that parallelism is a huge unlock there as well.
00:19:57So maybe the model is slow, but maybe if it can do multiple things at the same time, like suddenly that is actually not necessarily a blocker or impediment.
00:20:08It's fine that it like takes some time, but at the same time, you kind of want to work on multiple things, multiple areas of the canvas while this is working.
00:20:17I want to work here. Also like you don't have to wait for the agents to finish.
00:20:23You can like start poking into the design sort of like under the hands of the agent, which is also really cool in pencil.
00:20:32So you can like start reorganizing things already while it's still working, you know, which again, like is this reason where like vibe designing makes a lot of sense versus like just coding.
00:20:45Because you can't really like touch the code while the agent is like working on things, you know, you might screw things up and it might not compile and so on.
00:20:54But here, because it's so separated on the canvas, you can start doing that.
00:21:01I was going to say the thing I like most about the sort of pencil journey is that obviously you have a lot of domain knowledge in sort of the design world, because I think a lot of modern vibe coding tools.
00:21:10Some people have sort of just thought, what if, like, can we do this instead of should we? So some people have like, I've seen vibe designers before where it's been sort of a terrible workflow to use it, but I like your approach of always trying to help the human instead of replacing them.
00:21:25That's sort of the AI tools that I gravitate towards too. And I think, yeah, that's a very cool approach with pencil.
00:21:30Yeah, a hundred percent. I've talked to a few of our users, obviously, and like they talk about this fact where like they now have this sort of a group of junior designers essentially joining them on their journey and they can like orchestrate things way faster and way better.
00:21:51And previously there'll be like a lot of crafting, like proper padding and proper margin and like the right font size and all of these things.
00:22:01And you would spend like countless of hours on all of these things. But in your head, the vision would be already there.
00:22:07You just needed to take this time, this extra time to run this marathon essentially. But right now you can just sprint. And because you already see it, you know, you just want to materialize these things.
00:22:20I was going to say, going back to what you said about parallel agents or parallel things happening at the same time, do you mean having multiple agents work on the same design or do you mean having multiple agents work on different designs? What do you mean by parallel?
00:22:38All of the above, actually. It's kind of interesting, you know, so like initially when we started looking into parallelism, it was for exploring different variations of the same design at different places of the canvas.
00:22:53But then I wondered, like, why can't multiple agents work on the same page? And it's actually possible. And the reason being is that if you orchestrated really well, they can split up the work in a non-conflicting way, unlike in code, which on the canvas you kind of can't do it.
00:23:15You know, they just need to have the same understanding of what you're trying to achieve, you know, they will split up the work and they will work in a different sections of the website, you know, so like one agent can work on a hero section, the other agent can work on like the midsection and third agent can work on like a footer or something like that.
00:23:32And then you can run like multiple agents for each page, which is something I was actually showing in that last demo that I posted on X, where you had three pages and two agents for each page, basically. Yeah.
00:23:50I think there's a paper by Cognition that says don't run multiple agents because of the context issue. They don't all have the context of what they're trying to work on. And they gave an example saying if you were trying to build, like a Super Mario game, one could work on the level and one could work on the characters, but they could have kind of different gravities because they don't have the main knowledge. And so how have you got that to work with with Pencil?
00:24:15Magic. Well, it was a lot of iterations, you know, you just have to like keep digging into that and like eventually and like honestly, we just scratched the surface. I think we just showed like the direction of where things could be going.
00:24:32And honestly, like there's so much to be done also in terms of like how you work with that shared context, optimisations around that, again, like going back to speed and token usage and all of these things, like there's so much that can be done.
00:24:49And like we barely just scratched the surface, really. Like there's so much exciting stuff ahead of us. You can have different roles, like you kind of mentioned. You can have agents, I think I feel like one exciting thing for me is definitely like having the same role versus like different roles, because like usually when people talk about teams of agents, they would have like different roles.
00:25:14And that's cool in a way, but they're ultimately doing like different things and they're waiting for each other. Like the parallelism I'm like really excited about is that these agents are not waiting for each other. They're like truly working in parallel, just like GPU, essentially.
00:25:30So considering there are so many tools out there, so many models, so many agent harnesses, when you're working with Pencil, what is your, what tools do you use and how do you go about back and forth between the agent and the tool?
00:25:45And do you mean like what kind of tools, like specifically in Pencil or?
00:25:49So both, so specifically in Pencil, or do you use Codex? Do you use Claude Code? Do you use, I don't know, Gemini? What do you use?
00:25:57All of the above, like there was also like a big surprising moment for me when we launched Pencil, like how many different combinations people want to use it with.
00:26:08And like immediately our discord blew up and everybody was asking like, can I use it with like this model blah that I never heard about, you know, before.
00:26:17And can I use it with this agent harness and this CLI and this and that. And like, I was just like surprised how vast this world already is today.
00:26:30And like, we're really just at the beginning of it and we'll see all sorts of different combinations.
00:26:38I think it's going to be, right now in Pencil, you can run one frame in Codex and another one in Opus, third one with Gemini, et cetera, et cetera, et cetera.
00:26:48You know, it's kind of incredible. Actually, you can run all of these models at the same time. You can compare and contrast the results.
00:26:56They can collaborate with each other and so on. So it's kind of interesting.
00:27:02Curious how you benchmark the models yourself internally for that design capabilities. If you do that, do you have a sort of system set up every time there's a new model release, you can sort of check it for if it's been a great improvement.
00:27:14Yeah, I mean, that's sure. There is like some automated way, but there is obviously like a lot of manual work involved in that just because we are like so early in this thing.
00:27:32Like when the models come out, like the first thing I'm like always excited about is like, OK, let's see what is there. And sometimes it's just like nuance. You can't really put in words, but you see it and you're like, there's just a tad bit more detail, like going from Opus 4.5 to Opus 4.6.
00:27:51I can't even describe it. Like people ask me, like, so what's the difference? I wish I could tell you, but like when you try it, you can see it. You know, there is like a little bit more detail here and there. And suddenly it just makes a lot more sense.
00:28:03And I think that's going to be this iteration that we're going to be going through as we go to maybe Opus 5.0 and so on. There is also now new image models coming out, obviously.
00:28:15There's again, like more a little bit more detail, but like you can't really put it in words. It's just like once you see it, you know it.
00:28:22So at the moment, if I wanted to get some designs for a website, what I could do is I could go to Manobanana Pro 2 or whatever, and it could design me some websites. And then I guess from there, is there a way to take that design into Pencil and then design that or is it best just to start from scratch in Pencil?
00:28:48100%. And in fact, you know, like what you can actually do is, so first of all, like you can drag an open image into chat in Pencil. You can drag multiple images there.
00:28:59You can also drag images and like MD files or text files with your own set of rules or guidelines for it. And that's what like a lot of people are now essentially building around it.
00:29:12It's like a set of rules plus libraries plus images to get the results they want. You can also take an image of a fruit or a flower and use it as a sort of like vibe image.
00:29:28That's how we like came up with a bunch of our style guides was also like, let's just use some kind of vibe image, you know, like a flower or something, you know, and let's see like what color palette is going to get extracted from that vital model.
00:29:43And I think we'll just like generally see a lot of that more and more of that, like exploration and combination, because like just like human, you walk outside, you maybe go into mountains or woods or by the sea and or just walk in the street and you see some inspiration and you get inspired.
00:30:04And you're like, aha, you know, this is what I would like, but like in your hat, you know, when you were like manually designing before you wouldn't obviously even like take a picture of it, which is like a take a mental snapshot of that into your hat.
00:30:17Go back home to your computer and like start designing and they will be influenced by something you have seen during the day.
00:30:25But now you can just pass it to the AI and let's see what comes out of that.
00:30:30Is it also capable of designing some crazy SVG animations or something if you prompted with, I don't know, an SVG asset or an image?
00:30:42SVG is not necessarily just yet, but I have seen some SVG models that recently came out.
00:30:49This is like really interesting what's happening on that front. And also, I think animation is also like a big next frontier.
00:31:01There is like Remotion, right? And Lottie and all of these things. And Replit just came out with animation, right?
00:31:12So I think there's going to be, this is just like the next big unopened frontier.
00:31:18They just barely scratched the surface again. It's like it's nice, but it's still not like after effects level, like a pro level person kind of designer doing like amazing like animation with it.
00:31:32But I think there is a lot of potential in that space as well.
00:31:36You know, I saw the new SVG model and it looked very cool from the demo because current models do struggle a bit with SVGs.
00:31:43They're getting better and better. But yeah, that specific SVG model was incredible compared to what's currently possible.
00:31:50It touches on this like very interesting topic, which I think we are getting to the point where like layout is looking pretty okay.
00:31:57But like we know that websites are not just like good layout, right? It's this emotion that you get from it.
00:32:04And that could be reinforced by an illustration, by breaking the text into like very bold, you know, and then like maybe like subtitles.
00:32:16And so like it's a composition of text in like this interesting or funky way.
00:32:22Maybe there's a video that's that's moving, but it's sort of like this feeler only, you know, it's not necessarily like it's like abstract or like abstract image that's like moving a little bit or like things that like slide or move with scrolling and so on.
00:32:41So, yeah, I think those are the things that are very likely start to make it into vibe designing as well, especially on the illustration side, because then you can like, what about 3D?
00:32:54Like a lot of these websites now have like these renditions of like there's like cubes, you know, and they kind of like show the structure of like, I don't know, maybe like some websites that are trying to explain like how bytes come together into like packages and so on.
00:33:11Like we've probably seen like a bunch of these websites and that's like the next missing thing because like right now, sure, like the banana or like these image generators generate like great photos, but like what about illustrations?
00:33:23And I think illustrations are going to be huge thing in the next half of 2026.
00:33:30Now I'm excited for that because I've struggled with illustrations in the past.
00:33:33I've never been great at designing illustrations with Adobe tools and stuff.
00:33:37So yeah, it is a great point, by the way, you know, I think a lot of us can appreciate what good design feels like.
00:33:47We can even see it in our head.
00:33:49It's just that there's this gap between us and computer to materialize it.
00:33:55And like what excites me a lot is to just see more and more of these tools helping to close that gap between you and your vision.
00:34:04Do you still think sort of the human touch is the fact that obviously if we trust AI to do the design entirely, everything's going to look quite generic.
00:34:13Do you think that human touch is, as you said, getting the inspiration as a human from somewhere outside in the world and coming with your idea and actually steering the AI in that direction?
00:34:22Is that sort of why you still need a human designer essentially in the future?
00:34:27Yeah, I mean, a lot of great products that we use on a daily basis have that like taste element that's undeniable.
00:34:39And it's really important because I think what we will be getting from AI on the code side and also on the design side is this like cookie cutter result, you know, and it's going to be improving.
00:34:51It's going to look even better and better and better, of course, but it still might be missing that last touch of someone.
00:35:00And I think that's going to be still really important.
00:35:02And that last touch might not be even like last touch.
00:35:04It might be like a process in between and a lot of iterations where you stay in the loop.
00:35:11It's just that things like parallelism and these like vibe designing tools and vibe coding tools will just allow you to get there faster.
00:35:18You know, so like the amount of reps you do will allow you to produce things and not even faster, but like ultimately better.
00:35:28And if you look at the Internet today, not everything looks amazing.
00:35:34And my hope is that we're just going to like elevate the overall user experience all over the place, everywhere, just like by an order of magnitude.
00:35:46So that's that's really my hope for like at least like the future.
00:35:52As far as I can see, like that's probably like the next frontier.
00:35:56What comes after that?
00:35:58Like, who knows?
00:35:59We're still discovering a lot of these things, frankly, on like daily basis these days.
00:36:05Would you have any advice for sort of a junior designer going into the role on sort of how they can keep up in this world?
00:36:12Yeah, I think Marc Andreessen actually described this like really well.
00:36:18I'm just going to like go back a couple more steps and then we're going to answer your question.
00:36:23You know, but like he described it, he described it as this like Mexican standoff where like PM is pointing to a designer and engineer and engineer to designer and PM.
00:36:35And everybody thinks they can be the other person.
00:36:39And I was thinking like, so historically you would go to a school and study design for like four or five years.
00:36:46You would go to an engineering school and you will start engineering for four or five years.
00:36:53And as a PM, maybe you went to business school or you went to some of these things, you know, and you just really were like obsessed about building things and trying to figure out like the user behavior and so on.
00:37:06But like I'm actually honestly asking myself.
00:37:11What is like Gen Z and like Gen Alpha going to think about when they're going to be picking up like their school or profession?
00:37:21Like that's for me like the biggest question, because like I think everybody is now becoming a maker, essentially a builder.
00:37:29And we don't think about ourselves as necessarily or at least as my hypothesis, you know, as like I'm just an engineer.
00:37:39I'm just a designer like this new gen, you know, I'm not talking about like the current gen, you know, but like the new gen of folks.
00:37:46What I think we'll see more and more is like problem solvers.
00:37:50I have a problem that I want to tackle and I will just pick up the tools that I want.
00:37:55I've always operated with this mindset and this mindset with me was really started with flash.
00:38:01You know, back to that story because like flash was really the first tool that allowed you to be a designer and an engineer in like one environment.
00:38:11And it was really magical.
00:38:13Obviously, it didn't scale.
00:38:15It broke on many levels later on when like new frameworks came out and so on.
00:38:20But I think what I always dreamed about, there was going to be a comeback, something like that, where people who just have a vision will pick up whatever tools necessary to realize it.
00:38:33And vibe coding to me and like vibe designing and vibe coding to me is the closest thing to flash I have seen in like two decades, essentially.
00:38:44I wanted to ask about your style guides.
00:38:47How do you manage those and how do you build those?
00:38:51Because like with design, we know that each year Pantone comes up with like, oh, this is the color of the year or they're like design trends that, you know, everyone uses at this moment.
00:39:05So do you imagine like pencil being also updated as these trends evolve and change?
00:39:12Yeah, as a matter of fact, we just updated our style guides with the latest release.
00:39:16And we're always like refreshing them.
00:39:18But I never thought about our style guides as like, this is it.
00:39:22I always thought about it as like, this is just a little bit to give you an idea what you can achieve here.
00:39:30Also really fix that call start problem of blank canvas.
00:39:37And like, I don't know how to use the tool.
00:39:42So that was really the point here.
00:39:46But our goal is to actually document it and open it up so people can make their own and contribute and start sharing these across the community.
00:39:56I was never like in a position where I would be like, we want to be the pacemakers or whatever.
00:40:03Not at all.
00:40:03Nearly an inspiration for others to build on top and get inspired.
00:40:09Myself and back to Flash actually, you know, when I started working with Flash, I studied tons of Flash files, how people build different things.
00:40:19And that was for me the source of inspiration and how the first of all, like how to create these things, you know, but also like all these like intricacies in there and so on.
00:40:33So I think I just hoping that our style guides are just merely that, an inspiration for others to build on and remix and change and make them their own.
00:40:45I was going to say, it seems like other companies are figuring out what Pencil's done.
00:40:51So I know Sketch have got the MCP server or tools and so does Figma.
00:40:56Figma also have their own MCP tools.
00:40:59How are you going to differentiate or stay ahead of the curve and not let people like those other companies take your market share?
00:41:07Well, I think we just have to like, it's an interesting question, you know, but like, how do you differentiate between like Cursor and Copilot?
00:41:17Maybe like, I'm going to think it back at you.
00:41:20So like there is a nuance, right?
00:41:24There is a difference.
00:41:26Like, can we put it in words probably if we like dug deeper, but like there is a major difference, obviously.
00:41:33So I'm just hoping that we can keep that, you know, stay ahead of the curve essentially, you know, there.
00:41:45And like I said, you know, like I said multiple times here, there's just so much to be done, you know?
00:41:50And I just want to make sure that we're like front running that innovation on the Pencil side.
00:41:57So Pencil, for it to be made, you went through, I think a program, is it A16Z Speedrun?
00:42:05Yeah, I'd recommend it for everyone to consider it.
00:42:11So there's obviously Vysee and Speedrun.
00:42:15Those are like the most famous ones, I would say.
00:42:20But there is like many, many accelerator programs around the world.
00:42:24Speedrun was a great experience for me because multiple things.
00:42:31So during COVID, everybody sort of like hunkered down and like didn't get out much, right?
00:42:40And I think Speedrun was a great opportunity for me to like reconnect with fellow founders again on like regular basis.
00:42:48So, and this amount of energy is like underrated, I think.
00:42:53I think everybody should go through that and just like spend time in like a tight group of people
00:42:58who have like a same vision to keep pushing the boundaries of what's possible.
00:43:03And so that was like number one.
00:43:06Number two, I would say it like really shapes you as a founder in a really positive way.
00:43:13Like first of all, like the great support from the Andreessen Horowitz team there is just undeniable.
00:43:20But like the other thing is, and anything like basically discussed like tons of things with them
00:43:25and they have like great experience because they have seen so much.
00:43:27So they have this like great instinct about like what might work, what might not work, you know?
00:43:32And they're just great advisors.
00:43:35But also there's this moment where you have to present at a demo day in front of like thousand
00:43:42investors. And this year we were like Speedrun number five.
00:43:49It was in Yerba Buena Center of Arts in San Francisco.
00:43:53It's this place where like Steve Jobs announced ITAT and other Apple products.
00:44:00And like you are now standing on the stage.
00:44:02There's some magical energy there.
00:44:05Like it's insane.
00:44:07And I can tell you, like, I think everybody should go through it.
00:44:11Like this is such a great experience and it will force you to really sharpen your pitch.
00:44:18You have just two minutes to present.
00:44:22And I can tell you, like talking for an hour about pencil, easy.
00:44:28Talking about pencil in two minutes, incredibly hard, especially if you're standing on a stage
00:44:35where you are blinded by the lights when you come there.
00:44:38Because like you're coming from like this backstage sort of like dark, super dark place.
00:44:44You know, there's like no lights, nothing, you know?
00:44:47And suddenly you get on the stage, super bright light shines on you.
00:44:52And like you have to start talking about whatever you're building in like very excited way.
00:44:58And it just has to keep coming like this, you know?
00:45:01So I went through like a bunch of rehearsals for that, obviously.
00:45:05But like overall, I think for everybody there is such an amazing experience to just like
00:45:10get out and like really pull the best out of you.
00:45:14Yeah.
00:45:16I always think people who are in California or San Francisco have an unfair advantage
00:45:20because you're in the hub of technology and you kind of feel the kind of power of AI and
00:45:28like all the tech hubs and stuff.
00:45:29Do you feel the same way?
00:45:31Yeah, a hundred percent.
00:45:34I mean, I used to live in a cell for many, many years.
00:45:36I don't live there right now.
00:45:38I'm like dozing it myself because there's also like, frankly, is great energy.
00:45:44There's also some noise related with that, you know?
00:45:47So I think it's at least like for me, that's the balance that I found is to just like go
00:45:53there on regular basis, but don't necessarily be there like on daily basis.
00:45:59It also like inspired me, inspires me in a new way every single time I go there.
00:46:05So that's what I found working for me.
00:46:08But I spent a decade there, you know?
00:46:10So like, I know like every bit and piece of the city.
00:46:14But yeah, it definitely is some unfair advantage.
00:46:19And you can talk to anyone like who flies to SF, you know, how they come back from the
00:46:24trip always like incredibly re-energized.
00:46:26And so, yeah, I mean, it was like one thing I would do is like just take your bag and like
00:46:33fly to SF and just like meet people and talk to them and just like, you know, jam on things.
00:46:38Is Pencil.dev, is there a team or is it just you at the moment on Pencil.dev?
00:46:42We're like 10 people right now.
00:46:45Yeah, so and yeah, it's a great team of folks.
00:46:50You know, we've known each other for some time.
00:46:52So yeah, it's actually a really complex product if you really think about it.
00:46:57So it definitely needs more than me now.
00:46:59So yeah.
00:46:59Are these people that you met during your time working for Adobe or the startup you had or
00:47:06these people that you've known from the speed run?
00:47:09Some folks are from like the previous startup we worked on.
00:47:16And yeah, some folks are new.
00:47:18So it's a mix of people.
00:47:19Yeah.
00:47:20Oh, cool. And speaking of the previous startup, are you allowed to talk about that?
00:47:24Yeah.
00:47:25Like one of them was called Around, for instance, and it was like a video conferencing app.
00:47:30I don't know if you guys have seen it.
00:47:33It was, have you guys seen it?
00:47:34The one we used until it.
00:47:37Yeah, we did use it.
00:47:38Yeah, I already used it.
00:47:40We used to like it a lot.
00:47:41We were very sad when it went.
00:47:42Yeah.
00:47:44So what was your impression of the tool, you know, if you don't mind sharing?
00:47:50I'm kind of curious, actually.
00:47:51It just felt a bit more sort of fun and a better experience than like Google Workspace or Slack
00:47:57and things like that.
00:47:57It just, yeah, I don't know.
00:47:59It felt like it did exactly what it was designed to do with some cool features in it.
00:48:03It was always like our intention was to just make it like a little bit more fun, especially
00:48:07like during COVID times.
00:48:09Like, I feel like everybody was trying to make like HD video and 4K video and larger and larger
00:48:15and larger.
00:48:15And we were like, let's think different here.
00:48:17What if we made like small circles and put them on the desktop and actually made other
00:48:23people see less of you in a way?
00:48:25So we put these filters there and so on, and it made everybody suddenly feel more comfortable
00:48:31in all the team discussions.
00:48:33And I think that was like a big, big, big factor of that, you know?
00:48:38And I think with all of the products I worked on, you know, I'm trying to always find this
00:48:46like little edge that's just like completely different from everybody else.
00:48:51And it's always like seeking that versus like, oh, somebody built this, you know, let me try
00:48:58to redo it, you know, remake it or copy it or whatever.
00:49:01You know, I think it's really important to find like that new thing that you feel passionate
00:49:08about because ultimately this is a marathon and you have to be personally like excited
00:49:15about building this thing, you know?
00:49:17And with Pencil, you know, like when I saw these cursors flying on the canvas for the
00:49:23first time, I was like, this is it.
00:49:25It's amazing.
00:49:25You know, like I couldn't stop watching it.
00:49:27I was mesmerized by it, you know?
00:49:29And then I was like, oh my gosh, like I can't wait to share this with everybody else.
00:49:35So, so it's these little things that it started like initially we had this like animation
00:49:40that would like scan the things, you know, we didn't have to do it.
00:49:43Like we could just rat the JSON, you know, and that's it.
00:49:47But the fact that it like scanned the frame, I was like, wow, this is like, you're feeling
00:49:56you're feeling it.
00:49:56You're feeling the AI.
00:49:57So, and I'm sort of like hoping that like a lot of people have been talking about this
00:50:04a lot, like the chat being the main interface for AI.
00:50:08I'm honestly, I'm starting to feel like we're now discovering these like new user experiences
00:50:14for interacting with AI that were never seen before, you know?
00:50:19And these cursors and scanning and all of these things are like just the tip of the iceberg.
00:50:24And we'll see a lot more of that.
00:50:25Just like guiding you visually throughout your creation.
00:50:29Like it really makes it feel like that AI isn't just some like LLM text chat, but like
00:50:36something behind it, something creative behind it.
00:50:40I don't know.
00:50:40Like I can't really find the right word for it right now, but it's starting to feel like
00:50:45that.
00:50:46You know, I think that's especially true that sort of chat is a bad interface for things
00:50:50that are creative, like designing and all of that.
00:50:52It's, it's really hard to get into a prompt sort of what you want to see on the page.
00:50:56Um, yeah, and I think if we want more people in the world to use AI, we'll have to figure
00:51:03out better interfaces for that.
00:51:04And sure, like voice could be like one of them, but like, I don't know, like how many times
00:51:10do you guys use voice to interact with AI?
00:51:13Like rarely, you know, so, uh, I think we'll have to find some new interfaces for that as
00:51:19well.
00:51:20Yeah, I've been thinking about that, um, on and off and I think if there's a way that AI
00:51:24could listen to what I'm thinking, because sometimes I don't want to talk, I don't want
00:51:28to type, I just want it to understand what I'm thinking and just do it.
00:51:31I don't know if that's neurolink or whatever, but if there's a way to do that, that would
00:51:35be a cool interface.
00:51:36Yeah, you should, you should actually consider inviting a good friend of mine to this podcast.
00:51:44Jakub Zeczolka, he's a, he's a, uh, well-known tree designer, you know, just like up and coming
00:51:50designer and he just recently posted a video of his latest prototype.
00:51:55He's doing eye tracking on iPhone and he's basically just looking at different things
00:52:02on the screen, you know, and just talks like what should happen with that.
00:52:05He doesn't have to describe like this image on the left.
00:52:08He just looks at the image and says like, change it to something.
00:52:10And the AI knows what he's looking at.
00:52:13So these are the kinds of interfaces that like are going to open like so many new things because
00:52:19like if the AI knows what's on top of your mind, like the eyesight and the direction of
00:52:24like where you're looking is like one of these clues.
00:52:26And hopefully we'll find a lot more of these clues.
00:52:30Then it's just going to feel a lot more natural.
00:52:33Just like you were standing here next to me in my room and I wouldn't have to like tell
00:52:38you look at this and like specifically point to that thing.
00:52:42You would already know that just from like the context of the setting.
00:52:48That is super cool.
00:52:49I was just going to say I am actually one of those people who use voice a lot because sometimes
00:52:55I'm just tired of typing.
00:52:56I just want to talk to the AI.
00:52:58So if I can talk to it and it knows where my eyes are, that's a cool product.
00:53:03I like that.
00:53:04Yeah.
00:53:05Or even like imagine that it could track a hand and you can like point at things on display
00:53:10or like do this, you know, like sometimes you know how you are in a car and you just say
00:53:15like go here and you don't say like right, you don't say left.
00:53:19Just like turn here.
00:53:20And like I wish I could like tell the AI you know in the car or like autonomous driver like
00:53:28turn here and like the person, the people in the car with you, they would know.
00:53:33The AI has no idea.
00:53:35You would have to be like turn right on this intersection.
00:53:37I'm like, oh my gosh, who's going to say that?
00:53:40You know, so it's just too mouthful.
00:53:43So yeah, do you have any hot takes on design or AI or the tech industry in general?
00:53:50It's such a loaded question.
00:53:53Although like it's very simple.
00:53:55We need the clip, you know, we've got to get the clip.
00:53:59Yeah, what's a hot take?
00:54:02Well, I think honestly like this, this past thing we've been just discussing is it's a
00:54:06pretty hot take that I'm just hoping that we're going to uncover these new interfaces for
00:54:12interacting with AI that are going to feel more humane, more natural.
00:54:16And the AI will just know from the world context, from your own context, from your point of view,
00:54:26from your focal point, what you actually had in mind.
00:54:30So you don't have to like handheld it so much.
00:54:34Just you have to do today, you know, and going back to my initial idea of like, I just wanted
00:54:40to draw something versus like describe it with words is the exact same thing.
00:54:44Like for me, it's oftentimes easier to draw something or to point at something than describe
00:54:52it with words.
00:54:53So I think this is like sort of like the next immediate step that is going to start changing
00:55:00in the AI.
00:55:01Also, like if you, if you think about like chat GPT, like few years back, like you would
00:55:06have to like really ask these like long questions and prompts right now.
00:55:10You just ask something small and like does the whole analysis of things and gives like
00:55:1420 options and so on.
00:55:16It just like keeps going.
00:55:17It like kind of knows where you're like going.
00:55:21And maybe, maybe it's informed by like other chats people had there and it cached some of
00:55:30these like thoughts and answers and so on, you know?
00:55:33So yeah, I think that's also like the other thing that like a lot of the times the, you
00:55:41know, like a new LLM prompt is like a start of the new world.
00:55:45Like it starts from scratch.
00:55:46Like I'm hoping that we'll see a lot more of that like shared memory.
00:55:51And like having, I think the big difference is like LLM please summarize this and like
00:56:01LLM being a colleague on your Slack, having all the context of like all the conversations
00:56:06that happened there and doesn't have to like dig it out every single time to like think
00:56:11about it because it kind of feels stupid.
00:56:13So yeah, I think we'll see a lot more of that.
00:56:16I was going to actually say going back to around, we really miss our sound effects and the mini
00:56:21game we would play at the end of the call.
00:56:22It's like our biggest loss was sound effects.
00:56:25We loved them.
00:56:26Yeah, totally.
00:56:28And also like it wouldn't be cool if you could just play the game right now.
00:56:31Yeah.
00:56:32Yeah.
00:56:33Yeah, that was my favorite.
00:56:35Yeah.
00:56:36Cool.
00:56:40Fun times.
00:56:41I think we use Slack now.
00:56:42We use Slack huddles and it's not the same.
00:56:44Yeah.
00:56:45It blows my mind.
00:56:46It doesn't have a record feature Slack.
00:56:48Like just let me record the call.
00:56:51It seems so simple.
00:56:52But anyway.
00:56:54Yeah.
00:56:55Yeah.
00:56:56Cool.
00:56:57Are you allowed to talk about what happened to Around?
00:57:00Yeah, Around got acquired by Miro and and got fully integrated into the Miro wideboarding
00:57:08experience, you know, so it's there.
00:57:10Okay, fine.
00:57:12So it still exists, but you have to use Miro for it.
00:57:15Thanks for your time, Tom.
00:57:16Is there anything that you want to plug in?
00:57:17Anything you're working on other than Pencil that you want to talk about?
00:57:21Yeah, I mean, well, first of all, like go check out Pencil and I'm here for everybody.
00:57:27You, you know, and all the users that tried out general Discord.
00:57:31People share a lot of interesting insights, you know, using Pencil and also screenshots
00:57:40and files of what they created.
00:57:42It's a pretty vibrant community of thousands of people at this point.
00:57:46And also feel free to DM me either on X or Discord if you have if you run into any issues
00:57:56or you have any questions or feature requests, you know, just just want to make sure that
00:58:00like, you know, that I'm like here for you.
00:58:02Nice feature request I have is a custom UI kits.
00:58:07Because sometimes I want to make my own UI kit without using Shadsian.
00:58:10Totally, totally makes, makes a ton of sense.
00:58:13Thank you for listening to the Better Slack Podcast.
00:58:16Find us wherever you listen to podcasts.
00:58:18So Apple, Spotify, give us a like or subscribe if you're listening on YouTube.
00:58:23And for me, it's a goodbye.
00:58:25Goodbye from me, goodbye from me and goodbye from me.
00:58:28Thanks, everyone, for listening.
00:58:31Really, really enjoyed it.