Pablo Stanley - Protecting your creativity in an AI world

00:00:00There is a danger in using these tools,I think, because if we delegate too muchof to it, we're not forcing ourselves tothink and this is what was happeningwith me when I was trying to do thepixel art. I'm being shaped by the toolsthat I use. Feels like an addictionbecause it's so good. My belief is thatwe need to use it in a way that expandsour ideas. It magnifies our curiosity.It allows us to go deeper but alsobroader where suddenly maybe that thingthat was like a little spark in yourbrain suddenly can be amplified into agigantic star that is full of fire.Welcome to Dive Club. My name is Red andthis is where designers never stoplearning. This week's episode is withPablo Stanley who's one of the peoplewho I've looked up to in this industryfor a long time. So, we're going to do adeep dive into his journey buildingcreative tools. Everything from his ownproducts all the way to how he operatesas a staff designer at Verscell where hegets to build the future of Vzero.There's a lot in this conversation andPablo's a joy to listen to. So, let'sdive right in. I'm showing uh ecto whichis yet another

00:01:15design tool that nobodyneeds [laughter]that I that I wanted to make. My ideawas uh I want to make a a design toolwhere little robots are actually therewith you in the canvas and youcollaborate with the robots. And thiswas before Figma decided to do this. Iwas like what is going on? Why don't wehave agents on the ka? So like I createdmy own like a little tool for that and Iwas kind of on vacation and then I waslike ah let me make this thing. Actuallythis is started not as a as a designtool but a like more as an effects toollike with shaders and mesh stuff and allthe asy stuff that now everybody is likeif you don't have a tool like thisyou're not making it.>> Yeahright now. [laughter]>> Yeah exactly with the theater and justlike all this stuff. So, I did this likeuh last year. I was trying to do it withlayers. Like, how can we do this withlayers? Uh how can we add text andstuff? And then I was like, dude, I needa canvas. I need to zoom out andactually add stuff. And then I went downthe rabbit hole of turning this asything into well, this design tool. So,yeah, this is uh like a pretty much justlike any other like a framer, Figma,Sketch

00:02:31where you have the layers on theon the left. uh you have uh propertiespanel on the right and then you havecanvas you know and everything is codeso that's something that I wanted to dolike the part of the challenge was likea a like everything is actually code andit's very opinionated because I I wantedto make it in a way that what elementswhat is their default which is usuallyHTML and using divs and using tailwindclasses uh so I was like okay uh that'susually what uh lens love doing so I amgoing to do a tool that just uses that.So like for example here uh all theclasses here like by the default is thatyou can actually use like tailwindclasses. I don't know if this is visiblethere.>> Totally. I love that you brought that inline too. That's a nice little detail.>> Yeah, this just like a like whateverdepends on what you are uh selectingthen the tools are in line. For example,this is an image. So it just like agives you like a what is it going to bethe fit? If it's going to cover containlike a like a uh and also uh opacity ifit is a box then it gives you like awhat is the the background color thestuff in the like how is it going to bethe padding and all that stuff and allthe controls that you would expecthaving here on the canvas

00:03:47but also likefor example here it has a padding thatis uh that is something you would expectfrom let me actually change this to thelight theme I prefer light theme I don'tknow why it was in on dark theme Likehere all the controls are like I wassaying it's all tailwind it has an agentlike I was calling it jewels and thenhere you can connect it to your own forexample if you want to use your own uhCLI agent you can do it too like withyour terminal or just use the one thatis internal and you can say you canclick on anything and then just like askit like make a variation you know forexample it all creates it in the canvasso it duplicates it and then it's goingto make a variation I I I don't rememberI don't know how this is going. By theway, hopefully it's working. I shouldhave checked before jumping into this.Hey, look, it's it's changing. So, youcan see it's streaming the changes asthey're happening. And since again,since the canvas is actually areflection of the code, is you can justlike see it live happening. You know,real quick message and then we can jumpback into it. If you still haven't triedpaper, even though it kind of feels likeeverybody on this podcast is talkingabout it lately, they just made it wayeasier to get started because you cannow copy and paste from Figma directlyinto Paper's [music] canvas. All of thelayers and their properties,

00:05:02includingimages and SVGs, are imported in aninstant. It's pretty big deal, and youcan try it out today. Just head todive.club/paper [music]to get started. Framer AI agents justlaunched and it is exactly how I want tobuild pro websites. You can get thepower of working with AI agents butdirectly in a canvas that is tailor madefor shipping production sites. Forexample, I have like five different CMSsfor dive which isn't [music] exactly myfavorite thing to deal with. But now Ican have the framer agent do everythingrelated to CMS management directlyinside of framer. And of course, you getSEO, staging links, localization, andeverything else that you'd expect. Buthonestly, at the end of the day, thething that I love most is just beingable to explore, [music]design, and build with agents all insideof Framer. And if you head todive.comclub/framer, [music]you can get 30% off Framer Pro to startbuilding today. Now, on to the episode.By the way, yeah, I kept the shaders.So, you can see that this thing is alsostill the all the effects, all theshaders is still are there. And alsosomething that I did again before Figmawas doing any of this, I was also doinganimation stuff [laughter]like [gasps] I think

00:06:17I went down therabbit hole of just like I want to buildmy own design tool that does it all. Soyou go to animation and then you canjust like animate anything full timeline>> but the the timeline is is I I was likeI don't want to deal with the keyframes. I I really like uh how animationis done by other tools like uh what isit called? There's an animation tooljitter. Yeah, exactly.>> That's my favorite term.>> It's actually it's more about you'reediting the animation. You're notediting the element. Uh so like forexample here, let's see if the AI Thereyou go. So like now the AI made all thisanimation for you. And then you can justlike edit almost like shapes, you know,like all these things you can I want itto be a little bit snappier. So it lastsa little bit less, you know. So you canjust like select any of this. And nowit's it's still too slow. So it's almostlike a editing rectangles, you know. Nowit's a little bit snappier. You can alsodo uh stagger just like a a like asubtle staggering you know. So so thatway everything is staggered.>> How long did you spend on this? I meancuz you've built so many full featuresbut also there are a lot of littleinteractions and thoughtful details andthis is quite impressive. What are wetalking here?>> Couple of months. Uh I think uh it'sweird because I spent a very long time

00:07:32trying to make this thing effect uheffects. I spent a solid month and ahalf. This all works in shaders and itworks in 3D and I was trying to make itwork in 3D, you know. So like I wasactually Let me see if this is stilllive. Yeah, this is still live. So youcan see that I was working with shadersand actually with an actual canvas and Iwas trying to add my own tools. So I wastrying to make it work with the canvasmeaning that uh then I can actually addtext layers here. Hi. But this is allthis all looks 2D but it's actually 3D.This is all 3D layers. But the LLM, likeI was trying to do everything here, butthe LLM gets so confused about this,dude. Because I was trying to do uh autolayout with this stuff, you know? I wastrying to do flexbox with this and justlike, hey, let's uh make it horizontallike command G, but this just breaks somuch. Oh man, look at that. No, no, no,no. It It was just like a nightmaretrying to make this work because themath and everything just is insane forit to work. and elements will just getso confused trying to make this workbecause it all of this is not is notapparent but this is like actually 3D isit in an actual 3D

00:08:47canvas and a 3D spacein it with a flat camera and because ofthat it's not real HTML I spent so manytokens trying to make this work and thenonce I said like screw this this is dumbwhat am I doing why am I trying to forcean LLM to to understand effect this wayThat is just insane. I'm just going todo it everything from scratch but nowwith actual basic HTML. And then that'swhen I jumped into this. And then it wasmagic. Then it was so easy because thensuddenly the LM was using the actual inthe canvas the actual language that itunderstands. So I spent a long time onthe wrong direction. I don't know ifthere's a lesson there like sometimesyou go down the rabbit hole but you needto know when to stop and when to saylike okay this is dumb this is notworking the usual commentary and theusual belief is like you have to keeppushing man you have to whenever thingsget hard you just have to keep pushingand it's like no no no hold there has tobe a moment where you say like no thisis dumb [laughter] you know and and thisactually happened with this too like ata certain moment I stopped working ineffectto I was like okay

00:10:02this is dumb[laughter] I don't know why am I evendoing this I was building this tooleffect in the canvas and with agents andit has design systems oh yeah it hasdesign systems and stuff and [laughter]and everything is just like uh somethingthat uh all of this is fed to the uh uhto the LLM so it's going to understandit all the all the settings all the UIit's it's here and it's something thatthe LLM understands but I was realizingthat I that I was building a tool that Iwas not using. I was not dog foodingthis tool. You know, ideally I shouldhave been using a facto to build afacto. You know, I was not doing it. Iwas just like straight up working withthe agent.>> Yeah. Why not? Why not? What washappening there? What was thedisconnect?>> Well, the disconnect is that I don'tdesign like this anymore. That's it.Like I I don't use the canvas anymore. Iuse the canvas to draw, to make mycomics, to do handdrawn stuff. Doingthings here. I was using more TL draw.This was the cam the tool that I wasusing. I I don't know like a Oh, you cansee that there there's still some old uhthis this was kind of like my designtool in some stuff. This was really whatall I needed. This I don't know what Iwas doing here. Oh yeah, look at that. Iwas doing a sketch of a dumb thing. Ithink

00:11:17for example, this is a a new thingthat I was working on on on a new tooland I was like a a like this is how Iwanted to look and this is>> Oh, I can picture that design. And Iknow what the high fidelity versionlooks like. It's cool to see thisrepresentation.>> Yeah. Yeah. Yeah. Yeah. Exactly. So,this is like a Yeah. And also like alike a a very basic uh>> Did you ever use balsamic in the in theday? It was like excaladraw and balsbalsamic. I don't know how you say it.>> I I remember about it, but I never uh isit?>> I used to use it all the time. Like itwas so great. And part of me wonders ifthis is actually the right type of toolto pair with coding agents justridiculously fast wireframes and that'sit.>> Maybe there there's a there's a comebackfor a tool like this. Actually therethere was another one I think it wascalled whimsical.>> Whimsical. Yep.>> Yeah. Whimsicle. I I I I love that thattool too and I love using it and it waslike diagrams but they also had likekind of a wireframes. They had thisthing too.>> Yeah. Like there's a comeback for toolslike this because the LLM already knows.You can tell the LLM, dude, this is justa wireframe. This is just to show youthe the flow, but use my design system.You already know what my tokens are.>> This is starting to sound like>> I don't even really care about my tokensin the canvas that much. Like sometimes

00:12:32I do, sometimes I do, but 90% of thetime like I just want a roughapproximation and then I find myselfgiving the disclaimer back to themodels. Don't use the exact designs fromthe canvas. Prioritize the designsystem. I've made that prompt 500 times.>> Yeah. Yeah. Yeah. Exactly. Like a takethis only as a guideline.>> So you have this moment in time. Youhave this moment in time. You've beenusing Teal Draw. You're realizing you'renot dog fooding your own product. Thenwhat?>> Yeah. It wasn't really for me. Honestly,I was more like a building a tool forlike a people in marketing. I I feellike my whole stick for my whole lifehas been like a trying to make designmore accessible and more I don't knowdemoc democratization of uh design in away that was my whole stick but at thesame time I was realizing no but my allof my>> my actual work is happening in theterminal and it's happening with theseagents you know so but the thing thatI've been feeling lately man is I don'tknow this stuff is uh it's prettyisolating so I built this tool by theway likeThis is my most recent tool more thananything uh I wanted to work with agentsbut uh in a collaborative way becausesomething that I that I noticed likewith my team at versel and u vzero weall are using

00:13:48these tools you know we'reall using agents and we're all usingthem on our side you know we're allusing them on our terminal on ourcomputer we use a tool called gather andwe use standups and we on slack and weshare the things, you know, there's thiskind of a disconnect from what you'redoing and and who is actually doing itbecause you you would go and it's justlike, hey, we need to work on this. Andthen you go with your agent, you know,and you go to the terminal to the blackvoid and just like, hey, dude, let's dothis. And then it does it. And then yougo back to uh Slack and say like, "Heyguys, look, I made this. You didn't makethis." Uh, and everybody knows youdidn't make this. And it's okay. we knowthat you use an agent but you'representing it here you here's a PR youknow and everybody's doing this and thenthere might be some feedback hey this isnot working or like hey we should fixthis so then you will go and grab thatand then bring it back to the agentslike hey dude you embarrass me this isnot working no mistakes you know at theend every at the end of every prompt>> my pre-prompt for everything is don'tembarrass me in front of my engineeringcolleagues>> exactly yeah like real engineers aregoing to see this I will go and then Iwill fix it and they'll come back. It'slike, "Guys, I fixed it."

00:15:03No, it was theagent. Why not just like have the agentthere, you know? So, why not uh have myagent be part of that conversation andactually see what others are saying?Like a lot of the the planning, a lot ofthe this feedback is happening withother humans, you know, but the agentsare like in isolation in in a differentplace. In the work, the thecollaboration in a way is happening inisolation too with this black voidalways a green robot. Let me show youthe the the landing page. By the way,like something that I that I hit is justa little uh here obviously because nowwe can do this.>> Also, all of this has music. So, I don'tknow if you do do you play music, man.>> I do. Yeah.>> Yeah. Okay, cool. So, maybe you have uhused a tool like a color uh like a orsin pads like a>> So, this is in the background. It'sactually a sin path. And>> I love that you brought doodles backinto it too.>> Whatever I make, I will like bringdoodles. Anyway, let's actually use it.Let me just like uh get started. It'sjust like a Discord or like a Slack, butwhere agents work with you, you know,and and you have humans working withyou, but at the same time, you can justlike talk to agents. This is what Iwanted to do. Just like, hey, can Ibring agents to do stuff, you know, andyou could connect like uh cloud code orcodeex or

00:16:18cursor and they run on yourcomputer or you can just like add anagent that runs on on a sandbox. So, forexample, let's uh Io read. Let's call itIlo Reid. You can give it instructions,but really it's just like a I don'tknow, like a system prompt if you wantedto, but it's already good, you know.Hey, uh can you add a little Easter eggabout Dive Club by Red in or homepage inthe footer? Uh shout out. Let's see whatit makes. [laughter]So, oh by the way, yeah, so this thingis connected to my GitHub. Like it'salready connected to my repo, so italready knows. And here it's just goingto create a thread and it's just like ait's coding, you know, just like it'srunning on a on a sandbox and it's justdoing everything. And I could havesomeone else's agent at the end betagged. I could have asked Eery to tagan agent to review its PR. Uh but Icould also at the end ask another agentto review the PR. There you go. It itfinished. Uh let's see. Uh uh there yougo. So like I now it uh created uh thisthing. So I can go and it's building thepreview, but I can go see the pullrequest. I can see what was changed. Sohey, there you go. Uh there's an Easteregg. A quiet

00:17:33and I guess it's going tobe in the footer. Yeah, it's on thefooter TypeScript file. Is it diveclub?>> Mhm. Nailed it.>> Okay, cool. Uh so I added that. Andwell, this is just like this is usingBarcel's. Yeah. uh like so it's stillbuilding it and over here the cool thingis that it's actually it's looking atthe preview from Barcel so once it'sready you can actually just preview itfrom here and then you can approve itrequest some changes or even reject it[snorts] I think a lot of people arefeeling this shift and I think your twotools perfectly exemplify wherepreviously yeah you're working withagents but you're on the canvas you cankind of like touch and feel things andmove things around and directlymanipulate them and you know click andjust change that padding just a littlebit. Whereas this is almost likeslightly abstracted way of working. Youknow, it's dare I say orchestration.>> It feels weird, you know. Uh I I I'vealways taken pride of the things that Ithat I make that I that I put put myhands into it, you know, that Iduplicated the artboard and then changedthings around, you know, did a sketchfirst and then I inked it and and it'sjust it it all feels like it's comingfrom my hands. So, there's a little bitof uh pride on that. And then we measureI

00:18:49think wrongly we measure or value orsuccess or I don't know by the amount ofoutput or the quality of that output. Ifyou are a little bit more artistic likeme how much of that is is a little bitof you. Oh there you go. Like let's seeis it hey [screaming] look at that.>> Nice. [laughter][gasps]>> Oh dude look at that. There you go. Youknow what? Let's approve it. Approvedbro.>> Beautiful.>> There you go. It approved. You youdidn't hear it, but there was a littlemusic. Oh, nice.>> Love the competitive, too.>> I I think this is a hilarious microcosmof what it's like working with AI, too,as an aside where it's like you're doinga thing, you're doing a thing, and thenit's like, oh, a new thing to review,and you go over here, and you review it,and then maybe another thing to review,and it just feels like I'm playingwhack-a-ole with the agent.>> A lot of our work is becoming more likea orchestrating, manager, like directingthese things, you know? And it's weirdbecause I've been having this uh likekind of a crisis where it's like is thiseven me? You know, is this even mine?You know, what is this if I'm not makingit with my own hands? If I'm notactually drawing the rectangle, if I'mnot writing the code, if I'm not a Idon't know like a placing the things bymyself and trying different things and Ifeel it's a different

00:20:04way of creatingit. It requires I think a different partof your brain, you know, and you couldargue that managing directing is not ascreative. And that's that's how I feel,you know, because I have seen stuff thatI did like years ago and like I I canshow you like I I'm going to show youlike a gorgeousI I I was seeing like a this thing and Iwas like, dude, this was like so good.Like I I made this page and this videoand it was all done I mean it was donein web flow you know but uh it was allalong with care you know and actuallythis these images are AI [laughter] sothere was still AI you know the imagesare these are not real people but uh andthis is well that's really me but heyall the the the screen recording how uhthe animations happen all this was donelike the separation from the backgroundfrom this. Yeah, I use AI to separate ita little bit just to fill in the blanks,you know, and to separate this photo,but all of the rest of those things likethey were made in a way like actuallymaking it, you know, and and I[clears throat] feel like I don't Ihaven't done stuff like this like thisfeels more creative than this thing. Idon't know like now this

00:21:19feels roboticin comparison to that. Like this feelslike a because I I had the help from AI,you know, to make this and I wasdirecting it what I wanted, but I didn'tmake this in Figma. I didn't put this inFinland. I was like, dude, like let's doa grid. Here's a I think I did a a TLDDRdraw. Why I wanted to do this? And thenI was like, dude, like I just put somecars, man, with the image and I made abunch of images. So just use one ofthose images and then put some of thelet's talk about these things here. So alot of those decisions were also made incollaboration you could say with the AIbut I was like delegating a lot of thatthinking also to it which is I feel isis not great if I continue on this pathwhere I'm delegating a lot of thecreation to the AI. I feel like I'mgoing to become dumber and I'm going tobecome a worse designer. But at the sametime you could say the argument againstthis a lot of artists didn't make itthemselves. They will teach others toactually make it. Like they they willhave a specific way of doing things andthen they will have students actuallymake it. Like Diego Rivera would dothis. A lot of his murals he didn'tactually paint at all. Like uh even II'm pretty sure that Michelangelo didthis like he had painters

00:22:35like actuallydoing it. He was directing otherpainters doing it you know and he hadthe sketches and everything. He had theidea. He had the vision. Like alsoRembrandt, he actually some of thepaintings from Rembrandt, he didn'tpaint them. He would actually directstudents to make this. Andy Warholwouldn't even hide it. He would justcall it the factory, you know.[laughter] He was like, "It's a factoryand I'm not making I'm just signing itat the end, you know." And but it wasall, you could say, his direction, hisjudgment at the end. It was hissignature at the end. It was his, Idon't know, his reputation, too. I don'tthink this is new, you know. It it'sjust uh it's accepting that using AIlike how much of that creation you wantto give it and at what point do you wantto give it where you feel I am at thatlevel I'm so I'm at that level of AndyWarhol I'm at the level of Rembrandtwhere I will give these AI agents theability the agency to do it for me withmy direction you know I don't know ifyou you know this artist Saul Luitt uhhe did a lot of uh minimalist but heused these things call uh drawinginstructions. Did you do you know aboutthis?>> No, I don't. No.>> This guy was the original prompter, youknow, because this is how he would doit. He he he

00:23:50believed that uh uh ideasare are kind of a machine, you know, andand a machine that creates art. And thiswas what he would give to a museum, youknow, just like, hey, uh, on a wallsurface, any continuous stretch of wall,just use a hard pencil and put a lot ofpoints around at random and then makeall those lines connect to each other.And that's what he will make. And thenat the end, he will say, "That's my art.I that's that's my art. Yes, I only gavethe instructions,>> but even when he was dead, some of hisinstructions were still being used andhe they were still being said as this isa soul lu it you imagine that likethere's another one that is like aboutlike a like doing 4 in for 4 hours.Let's Yeah, this one within fouradjacent squares that each 4x4 ft youwill pay these draftsmen $4 an hour. Idon't know. He had a thing for four, youknow, it's like, hey, it's a four incheslong with four colors. And then he wouldcreate a stuff like this where it's likea it's all lines, you know, and hedidn't make it, but he wrote the prompt.And this was in the ' 60s, by the way.>> That's cool cuz that's the thing is

00:25:05like, am I atrophying muscles as adesigner? Yeah, in some ways I have tobe. There's no way that I'm not. Butthen every once in a while like I willhave an idea that I would not havepursued in like two years ago, 3 yearsago, you know, but now I feel like I cando literally anything that I can thinkof. And so I have a concept that isnovel and then I execute it and yeah,maybe I one shot at it, but I don'tactually care because it was the ideathat was novel, but it starts it gets somuddy because for every time I'm doingthat, maybe it's five other times whereI'm like, "Hey, show me a variation."Actually, show me six variations. andthen I just pick one which is somethingthat I that I'm pretty sure that Warholddid too. It's like guys I don't knowlike he would just be directing the sillscreening and he will be just likehere's a photo that I took now enlargeit and put it there and and paint it anddo it in multiple variations. We'regoing to select some of them and thenI'm going to put my signature at theend. It I mean he also did it in a waythat was kind of a counter against justlike art, you know? He was saying like,"Hey, look at this thing that is justrepetitive and that is that you couldargue that is not art and I'm going totell you it is art. Who decides what'sart, right?" But you could say that abigger example to what we're goingthrough is like Rembrandt.

00:26:21A lot of hispaintings it wasn't him who painted it.And still there's historians and peoplein museums who don't know which ones areactually his like which is his guess,you [laughter] know. But this was thedeal. The deal was he would teachstudents how to paint like him. So atthe end there was the student who gotsome knowledge out of it. You know therewas there was a kind of a I don't know atransfer of that knowledge to thosestudents and those students like DiegoRivera too. He would teach them how todo murals and then at the end thoseartists would be able to create theirown murals too with that style or with asimilar style you know. So there wasthis transfer of knowledge that washappening too and maybe yeah at the endthe credit was going to the artist toDiego Ria or to Andy Warhol or toBrembrand but those artists were able toget something in it which is somethingthat I feel we're losing with juniordesigners because that's something alsoyou would do with junior designers rightyou would hire a junior designer who isstill like learning and you will teachthem like hey this is how we use oursystem this is how we use our tool thisis how like they will try something andthen you will try to guide them on howto do that. I feel like that we'relosing that because now we're delegatinginstead of to junior designers, we'redelegating

00:27:36these kinds of things to theLM to those things that and we willguide the LM to do. No, no, not likethat. Not that purple gradient, bro.Then you start manually you're guidingthis thing to create an output that isokay, you know, and that is doesn't haveall that AI slop. But all that knowledgethat you might have and all thatjudgment and filtering that you will doas a potentially a more a designer thathas had some years under your belt isnot going to a junior designer anymore.It's just going to another model. AndI'm pretty sure these models arelearning from this stuff too. So nexttime they're going to not make thosemistakes anymore. But then what's goingto happen with the junior painters, thejunior designers, the the artists whoare still learning and who are theretrying to be mentored? what could happenis that they could use AI as theirmentor and then that's that's where Ithink we're going where those who aregoing to prevail are are going to bethose who just like use the tool in away that transforms new creations andthat don't just delegate everything tothe tool but actually come withcuriosity come with maybe an idea ormaybe just a question and then they usethis infinite machines

00:28:51to just ampliflify those ideas or take them indifferent directions. And I feel I'm I'mexcited about that. I feel like it mightalready be happening and we're notnoticing it. And I'm pretty sure thatthis is happening with younger people.We say AI native, too. That's the funnything. It's like we like, well, now I'mAI native. It's like, no, you're not AInative. You're a millennial. The AInative person is 18 years old andthey've literally grown up in this andthey're going to fly by all of us.>> Exactly. Yeah. In the 80s when hip-hopstarted it was criticized and and musicmade with synthesizers. I mean now weaccept like hip-hop is just like one ofthe top styles of music types of musicyou know and and electronic music is thesame you know but there was a time whenwe would question that and artists willquestion that this is not real music.You're just using samples from othermusicians and you're mixing them up andyou're supposedly creating something newbut you're not. This is not real music.Oh, you're using you're not using a ananalog real instrument. You aredelegating the sounds to a machine thatis just like synthesizing sounds. So,this is not real music, man. A lot of usthat that are our age, we question AI. Iquestion it. I question myself using it.You know, I I use it all the time. I Iwould be dumb

00:30:06not to use it, honestly.[laughter] But at the same time, I Istill have that uh old thinking where II question I was like, I don't know ifthis is real. You know, a lot of mywriting on this is very I want to staypositive, but at the same time, my mynegative side comes. [laughter]It's like I need to I need to work onthat. I need to talk to my therapist[music] about that. There's one questionthat I can't stop asking myself. What ifcompanies applied to talk to you ratherthan the other way around? And thatquestion is the foundation for theall-new dive talent network. And it'sworking. Like right now, I'm helpingmany of the most exciting startups thatI know to hire the designers andbuilders [music] who listen to thisshow. So, if you're curious what mightbe out there, and if you want to get onmy list, or maybe you're even lookingfor your next designhire, head todive.club/talentto join today. Maybe for a second wecould actually talk to the youngerdesigner who is listening, who iscurious, who wants to go all in on thisway of working, who views AI as a tool,as an instrument, and yet they don'thave a bunch of reps under their belt.They don't have the surroundingenvironment and encouragement that comesfrom a place like Verscell that is veryAIDS.

00:31:22So, can we just zoom in on partsof your workflow and what yourday-to-day work looks like working onthese models? Maybe we could use, youknow, one of the recent releases. I knowyou just did like design mode and vzeroa ton that went into that like help usunderstand how you are collaboratingwith these models on a daily basis. Whatit looks like to explore differentideas. I use this terminal ghosty. I useclaude and then dangerously skippermissions. [laughter]I'm like don't ask me questions justjust I'll let you cook, you know. Sosomething that I do also is I use agentsand which is it creates a tool thatcreates a list of the things. Right nowI'm just in my root folder but from hereI can just ask it anything. So it'sreally just a chat you know help usimagine a type of kickoff session thatyou would use when you're working onlike a medi [music] project at versell.This is the black void right we're justlooking at the black void. You don'teven have a visual of the productsurface area that you are working on.Where do you even start? How do you evenpoint it correctly? One of the thingsthat I use is for example I use TL drawa lot. I use it to just like a a forexample this was for design system thingthat where I was trying to just like puttogether like how to display thedifferent fonts and

00:32:38how type scales andalso like the pallets how to display theUI for this. And since I'm alreadyworking on the repo and in the repothere's already a design system. I don'tfeel like I need to go into highfidelity as much because uh a lot ofthat stuff like the fonts, the spacing,all of those already rules in the designsystem with the agent. All I need isjust like a direction of what I want todo. So this will be artifacts that Iwould give for example an agent justlike hey I'll write a prompt uh justlike a what I want what is does mysuccess criteria of what I think itshould do I would also ask it to to askme questions too sometimes I wonder ifgiving it a screenshot is is to if ifI'm still on the ideation part Iwouldn't give it a screenshot and I willuse it more as a collaboration toolwhere it's like hey I have this ideawhat do you think we should do you knowand then it come with some ideas and Ihad this idea I'm thinking it should belike this but what do you think you know>> so then it will give me some optionssometimes it will validate my idea whichmakes me feel really nice just repeatingit back it sounds like the the vastmajority of the chats or new sessionsthat you are doing with AI you're notjumping right into like an execution itis very much so I

00:33:53kind of want to workon this maybe you have some preliminarythoughts what do you think like thatquestion you said it twice is thatsomething that you're asking most of thetime sometimes Depends>> depends like some stuff is just thetypography is wrong. It should be thesize and you need to think about thisyou know and it's there's a bug this isnot working it depends on at what stageyou are if you are for example here Ithink I was really just starting youknow I had an idea like how this shouldlook but like a hey one little headingand and I want a drop down you know uhwith a search which I didn't make but II just described that I need to draw ityou know uh so in in that case I Butsometimes just to start like let's planit. No, actually not even plan it. Let'sjust collaborate. Let's uh let's dobrainstorming, you know, in a way. Andalso ask it to ask you questions isreally good because then it will ask youthings that you didn't even think about.I want to get a bit more context for howyou're working at Versell. So like onthis spectrum from you know ceiling ofone line tailwind fixes all the way toshipping full-blown features. Like howmuch of the front end are you actuallyowning as a designer? there aredifferent uh types of builders or makersor people collaborating

00:35:08in productdevelopment teams I think and I don'tfeel is the same as designers and UXdesigner and the copywriter and the PMand the front engineer and the back endengineer I don't feel those division Ifeel those divisions are getting blurredyou know and and and those types ofroles are I don't know if they're goingaway it's just like there's so much orstuff that has happened for years that Idon't think they're going away butactually in the actual work those thingskind of disappear and and it's moreabout like who is prototyping that isone that I think a lot of designers aredoing that and maybe PMs and prototypingas in like I have this idea let's put ittogether and let's uh let's test it outyou know>> uh and it could be just a wild idea andyou do a lot of those prototypers are Ithink making a lot of weird stuff ortrying a lot of different things. So toyour question, how much of the codeyou're doing? I think I'm in that areawhere I do a ton. [laughter]You know, I would put a lot of code inour GitHub. That doesn't mean that allof it gets chipped though. Somebodyshared a screenshot. I don't know ifit's still the same, but at least inthat moment, I was the one putting more

00:36:23code.like after Claude code it was me andthen another designer uh uh Tom John uhwho is also working in in Bzero. So thetwo designers we were even making morecode even more than other softwareengineers in our codebase. We wereputting more code out there. How much ofthat was actually shipped? Not a lot.[laughter]And that's because we were just tryingdifferent ideas. you know, we're like,hey, what if we do the composer thatanimates in? What if the homepage hasmore stuff? What if the homepage is justcleaner and sin? Like, what if thesidebar has this and what if the sidebarhas that? All of that stuff like nowit's happening in code. So, it's kind oflike instead of having a ton ofartboards and Figma files, now it'shappening all straight in code. andyou're having a ton of PRs and differentthings that you tried that are in draftmode or maybe ready for review butthey're never reviewed. I feel likethat's kind of a good place for us to beand and PMS. Then there's the ones whoperfect that code that might take thatcode and then cherrypick some of thatstuff to turn it into a PR that isactually going to be shipped. Forexample, this is another one that Ichipped

00:37:38is um what is it called? That'suh annotations and then design mode thatis over here. So annotations I added ina way that is like hey you just want toinstead of uh having to describesomething it's like hey uh make itbigger you know. So now I didn't have tosay hey the hello world the hello Idon't know span in the heading let'smake that bigger. I can just point at itand say make it bigger you know. Andthen the annotation is here. So it'sactually attached to the node and it'sdoing that. So and then another one iswell design mode. design mode is morejust like a more here but actually justlike a using the things that you wouldexpect actually a lot of the stuff fromAlcto after I was>> just about to say there's that there'sthat contextual little menu there at thetop it's cool.>> Yeah. Yeah. Yeah. Like also the layerslist uh which is more like a the blocksand the elements list the nodes on theleft and then you can rearrange and youcan move around and yeah like contextualmenus depending on what you haveselected all of that stuff. I built itin a gigantic PR because like this noneof this existed. This doesn't didn'texist. This didn't exist. Annotationsdidn't exist. A lot of the sidebar, thisis not the sidebar. Uh the inspectorpanel that I made, this is we're stillworking on this. It would be weird

00:38:54tojust do this for example first>> and then do the the layers list and thendo like a lot of that stuff. You need todo the whole thing, you know,>> especially when you don't know whereyou're going sometimes, too. Like a lotof you're still exploring. I find thatso tricky where I end up making PRs thatare way too big, but then I want to shipa lot of them because I don't have a bigteam, but I'm like, well, I didn't knowhow to split it up when I started, youknow, like this has replaced meexploring on a canvas.>> Yeah. So, even these things that likethese things that are like manually youcan just like the gap between the words,the gap between the elements, thepadding, all of that stuff does deserveto be their own PRs, you know. But forme testing it as you're saying and forme actually doing it, I had to leave agigantic PR. A gigantic PR where I wasdoing it was just design mode. Itactually still more advanced than whatshipped because it has the sidebar hereand it has some interactions here thathave haven't been shipped. I was the oneprototyping this and then engineer willbe the one cherrypicking stuff andactually making a reality. It was like,okay, I'm going to cherrypick this thingand then I will put in reality. I willdo the contextual menu and then I willmake it a reality. So there's that rolethat is uh of the person who maybe

00:40:09and Isee that others have mentioned this thatis that is not the prototype but is theone that is makes it for real more in ateam and and this would be a a softwareengineer you know and this depends onwhat you're making real. If it is justlike front end stuff, then it's going tobe someone who knows enough of the frontend and and potentially some of the backend to just make it real and then soit's performance. So it doesn't break.So all the dependencies are there. So itdoesn't uh like you're not introducingsomething that is going to create a bugthat you didn't even realize because youwere just trying to make this work, butwhile doing that, you just likeintroduce something that is breakingother parts of the codebase that youdidn't even know because I'm a designerand I'm a dummy. You know, the bigprototype that you're making, is thatoff of production codebase or do youhave a separate playground repo thatyou're building off of that theengineers then cherry pick?>> Oh, no. It's it's it's on top ofproduction code base.>> Oh, okay. Cool.>> I also want to be able to be using orreal tooling or real everything real.You know, there are some things that youcould do that is just like on a like ona sandbox or like in a HTML or whateverwhen you're just like playing around. Ibelieve at least in a tool like this youwant to connect to the I don't know toreal projects to the real agent you knowhere and that way you will test it

00:41:24theright way which is something that a a PRin a way is just something that is notat least depends on how the restrictionsand the permissions that you have inyour repo but it's not going to bechipped unless someone else approves itand then unless it passes all the testsso it's okay to just create a PR andthen just to be safe put on draft andjust play around with it and just usethe real thing. And that's that's a partthat just makes a tool like Vzero sopowerful because you're connecting to tothe repo, you know, you're connecting tothe to GitHub and you're working on theactual codebase and you're not justworking on a static artboard where justmimicking and you're just like a tryingto think of all the edge cases, but atthe end it's hard to think of those. Soyou end up just creating the happy path,you know, or just the static pathbecause you're not testing and you'renot finding out the things that uhusually happen when you're testing thething with real data or with the realtools or your real components. You know,there's another implication of howyou're building and one big prototypethat I'd love to learn a little bit moreabout which is like this is a mediarelease, right? There's so muchstrategy. There's so many UX decisions.There's so many overarching productdecisions that I would imagine a

00:42:39bunchof people had opinions on. So, how doesthis way of building change the way thatyou share ideas with the team? Uh, getfeedback, bring people along from thejourney, make sure that everyone's onthe same page so you can get thismonstrosity of a release out the door.You can go into a rabbit hole like AIpsych psychosis rabbit hole if you'rejust working with terminal in isolationyou know because you go and and thisthing that will always say yeah yeahlet's build that they will always agreewith all your terrible ideas then uh yougo build a monster that nobody wants youknow so you want to avoid that but yeahby having humans bring the friction thatwe all need real creation and I don'tYou know, things that that feel moreoriginal are going to come from multipleperspectives, having friction and havingpushing you pushing back and pushing youto do to think differently. And LM's,they're not that good at doing that. Andand I think humans, we still we stillhave that. We can be jerks sometimes,[laughter]you know. That's our superpower that wecan be like, "Ah, that sucks." you know,and

00:43:55it's okay, but it's good to sharethat stuff and put it out there. It'slike a record a video of your thinkingand then ask for for advice, ask forfeedback, put it on Slack, put it in aloom, share also share like now withthis thing, the cool thing is you canshare a preview link, you know, like aRELE preview link and then people canactually open it and click around andthen find the things and then they couldtell you where it broke, you know.That's usually how it works. We willhave a channel that is specifically forPR reviews and then we'll have channelsspecifically for projects. It were likefor example the design mode project hadits own Slack channel and we're justlike talking about it and we're justpushing ourselves to just like do betteryou know because I will put somethingit's like hey this is cool right andthen I will have a PM or engineer orsomeone tell me like ah what if we dothis hey this kind of sucks and thenyou'll fix it. So, like it's all aboutcollaboration and that shared knowledge.>> I want to talk about the Loom piecebecause you've made like you've made aheck of a lot of videos, right? Like youhave a lot of reps under your belt morethan a lot of average designers who havenot put themselves in front of a cameralike you have. So, I know there's likethis 101 set of advice

00:45:10for what makes agood Loom video. Giving the context,talking about the problem, whatever,yada yada yada yada. I want to ignoreall of that. I want Pablo's 2011 advice.like what are you intentionally doing tomake your videos more compelling? Nobodywants to suddenly see being tagged on avideo that is 10 minutes long, you know?It's like, oh, dude, really? You know,so you you want to keep it short unlessif you really need it to be long, thenyou better be telling a good story, youknow? That that length better worth it,you know? And whenever I I realize, oh,this is going to be like a longer video.So, I would try to tell a story, youknow, like start with, okay, why am Ipresenting this? Where did it all comefrom? I would try to sometimes I wouldsketch the stuff. For example, here uh Iwas uh talking about a robot. This is aproject that I made some years ago. Andit would tell like uh how things cameabout you know like before it wasvectorized like how it was actuallyfirst a sketch and a lot of it well forexample like a back in the day I waswere building a map and stuff and I wastrying to convince people of an idea youknow so I made this map and this waswhile I was doing the loom video of this

00:46:26I was like walking people through thismap and I was like telling them okayguys like I I felt like a like Charlieand it's always sunny in Philadelphialike showing like all the diagrams andhis conspiracy theories. But I wouldtell him like guys, okay, so we startwith this, you know, and then this uhfrom here this becomes this and then weget some tokens and then we get some uhprices for people. This is going to betokenized. So only these people aregoing to be allowed. But I was I wasusing like a first a graphic my ideathat was still very conceptual. I wastrying to guide people through this tothis vision. You know,>> it just points at the fact that youspike in so many different directions,which I appreciate. Like you got the UIskills, you have like a little bit ofcode background. You're you're you'redoing all the technical things, you'reliving in the terminal, and yet then youhave all of this artistic ability andand the illustrations. And I think itfeels sometimes like we're entering thisworld where everybody is being pulledmore technical. I'm sure a lot of peoplelistening feel that. I know that I do.So maybe before I let you go, one finalquestion I have is how do you thinkabout protecting your creativity as wehand more of our output to the modelsand we're doing more of the things incode and more of our decisions

00:47:41are atthe orchestration level. Any thoughts onthe future of staying creative assomebody who's really built a career onthat skill set?>> This project I had never done uh pixelart before. I made this I don't knowlike two months ago or three months agoand I made all the different things tocreate just like I don't know I thinkit's like 10,000 characters or somethingor uh that you can do with combinationsyou know and then I forced myself to doit by hand. I didn't use AI to do thisstuff. So this was like me learning apixel art tool that is not Procreate orPhotoshop. That was a challenge for me.It was a challenge for me to understandlike how to do this gradients, you know,with this pattern to just like simulateshading. At every turn, man, my brainwanted to use AI. It felt like a drug,you know. It felt like I was I'm hookedto this tool that can do it all. So nowmy brain wants to do it that way, youknow, wants to use that tool. Andinstead of trying to do the pixel artstuff that I wanted to do, my brain keptgoing to like why don't we create apixel art tool with AI and code insteadof doing the pixel art. If you do thetool then you can do

00:48:56later the art. Somy brain wanted to do that. I don't knowif you ever heard uh uh Mluhan he saidthat we create our tools and after thatthe tools shape us too which is greatyou know like a new technologies changeour way of thinking too I think it wasSocrates who was against writing he wasagainst writing because he thought thatwriting was going to erode our thinkingor ability to memorize things he thoughtthat by putting in writing we're justlike suddenlydelegating our thinking to the writtenword instead of trying to use our brain.Th this thing goes back for a while.That's why I'm also still very positiveand very optimistic about AI even when Ihave my own existential crisis be a lotof the times while using the tools. Atthe same time, I am like we're going tofigure it out because back like I don'tknow thousand thousands of years ago, wewere having this debate where thesmartest people were against writingbecause they were afraid it was going todilute our thinking. In a way he wasright because it did change our way ofthinking but he was wrong and that itwas going to just like make us dumber.you know there is a danger in

00:50:11usingthese tools I think because if wedelegate too much of to it then we'renot forcing ourselves to think throughand this is what was happening with mewhen I was trying to do the pixel artI'm being shaped by the tools that I usefeels like an addiction because it's sogood I think that you want to be able toforce yourself to not delegate all ofyour thinking to this tool and use thistool the right way is that we need touse it in a way that expands our ideas.It magnifies our curiosity. It allows usto go deeper but also broader wheresuddenly maybe that thing that was likea little spark in your brain suddenlycan be amplified into a gigantic starthat is full of fire. I'm soappreciative that you ended on thisPixabots example. I'm challenged byseeing it even, you know, because it isit's so tempting to just use the models.Use the models. I don't actually need tothink until Claude finishes its responseand then I'll pick up thinking again,you know, and so seeing you do this byhand, embrace the friction. I think it'sbeautiful and I find it inspiring. I'msure that people watching will beinspired as well. And honestly, Pablo,

00:51:26just thanks for coming on and sharingnot only what you've made and how you'reworking, but also just what it's like,you know, living in this world today andeverything that's changing and embracingthis this new way of working. I superappreciate you and have looked up to youfor a while and just grateful for beingsomeone who is continuously sharing whatyou're learning and thinking about inthis industry. [music] Oh>> man, thank you. Thank you. Appreciateyou.