Meng To - Codex superpowers for designers

00:00:00If you always chase the trends,you will become AI slopvery soon. Right now, the baseline isthe enter fonts, the purple gradients.That would have would have been amazing5 years ago. Like if someone were usingpurple gradients, everyone was goingnuts over that. But nowadays, AI slop,right? Very soon, you know, it's goingto be the serif fonts and the>> [music]>> you know, the same animations that AIuse all the time. You have to be alwaysa step ahead [music]of what is the baseline.>> Welcome to Dive Club. My name is Rid,and this is where designers never stoplearning.This week's episode is with Meng To,who's one of those people I like to keeptabs on to see how he's evolving hisprocess as he continues to build hisideas with AI. So, we're going to do aton of screen sharing and go deep intoall of his strategies for promptingdesign [music] MD files and littletactics that he uses to get betteroutputs from the models. So, to start,[music] I asked Meng to share more aboutwhy he's went all in on Codex since thelast time we talked.>> I used to use cursor all the time.But now, Codex is shifting the mindsetfor a lot of designers

00:01:15and a lot ofpeople who've never done the buildingdesign before or coding before. It'sIt's kind of like an evolution fromChatGPT into something far morepowerful. Your brain is on yourcomputer. All the files are there.You can build apps, but you can alsochat. You can also plan things. You candesign. You can generate charts, createHTML, create markdown files, and itconnects to all of the plugins andintegrations such as Slack, Telegram,uh you know, Discord. Uh most recently,you can now open and and continue Codexon your computerusing your phone, which is I think is isa huge deal. The first time we met,you know, we did this podcast, which wasabout 16-18 months ago.I built my very first app, you know,Dream Cut.It was using Cursor. It was using uhSonic 3.5.I never released that app. There wasvery key parts of the app that didn'tpass the the bar of quality.So, even though I thought in my mindthat I would eventually solve thoseissues,I never got to solve them.Because the technology

00:02:30was too young atthe time.It felt great to be able to to writecode an app in a week or in a month.But nowadays, with Codex,we can do all of thatin a day. But you need the knowledge.Recently, you know, I watched Mr. Beast.He was saying basically,because of knowledge, he's able to starta new YouTube channel,uh faceless one, and get like 60 million>> He said 20 million.>> 20 million. 20 million>> 20 million subscribers in 6 months. Andhe said he he he can totally do thatuh with the knowledge that he has. So, Ibelieve the same waywith Codex. Or with, you know, any othertools such as the new Cursor withCopilot 2.5 or the new Claude. But whatI like about Codex is thatI believe they were the first one. Theythey started in February, and they werethe first one to introduce this sort ofway ofbuilding everything with threads.And that completely changed the mindset,because it was so powerful that if youlook at Cursor now, that's exactly theirUI nowadays.Right? You you open Cursor, it's likeI can't Sometimes I can't even tellbetween Cursor and and Codex anymore,because they're so

00:03:45familiar. And and thedifference is just the model. But I Ithink the maturity of Codex is just fargreater at this point. Even though Ireally love Compositor 2.5, I've beendoing a lot of really, really quickiterations. In term of speed, there's noequal. So, this is one knowledge thateveryone must know is what model do youuse, what app should you use, and whatis the difference between Compositor 2.5and GPT 5.5?medium or extra high. If you're buildingapps, you need to know what the thetools are, right? Used It used to beFigma, which by the way, I don't openFigma anymore. I Yeah, sure, I sometimesopen it for a history of the work thatI've done in the past, like the logo andall that stuff, but I don't open Figmaanymore tobuild designs or to create design fromscratch anymore. I do all of that workin Codex.All of my projects are here, including,you know, the preparation for a podcast.I would just go to my agents, like,"Hey, I have a new podcast coming. Theseare my talking points that I want totalk about. Can you generate theimages?" Uh or anytime that I I find anarticle that is super cool

00:05:01or a video onYouTube that is amazing, I would go and,you know, "Okay, can you create anarticle on this amazing uh library thatyou know, someone just created?" Or, youknow, "Can you create an article aboutthe writing rules?" Like, I watched uma really cool uh podcast video onYouTube about the simplicity of writingand you know, copywriting and stuff likethat. I was like, "Can you watch thevideo and then can you create and putall the main points about it? Can youalso insert Paul Graham's essays at thesame time? And can you also remember uhlike the the philosophy of it and canyou integrate that into the writing whenwe do it together with the agents?[music]>> Real quick message and then we can jumpback into it. I've anticipatedDescript's recent release for a fullyear. It's called Surfaces and itenables you to design and prototypedirectly on top of an existingproduction interface.All of the pages and flows that you needare preloaded inside of Descript asstarting points. So, you can easily makechanges, explore ideas, and then whenyou're ready, you can share yourprototype in a single click. Descript isthe only design tool that allows you todesign on existing pages and flowspulled

00:06:16directly from prod, which is apretty big deal for teams and you canget started today. Just head to [music]dive.club/descript.That's d e s s n. If you still haven'ttried Paper, even though it kind offeels like everybody on this podcast[music] is talking about it lately, theyjust made it way easier to get startedbecause you can now copy and paste fromFigma directly [music] into Paper'scanvas. All of the layers and theirproperties, including images and SVGs,are imported in an instant. [music] It'spretty big deal and you can try it outtoday. Just head to dive.club/paperto get started. Now, on to the episode.>> So, so you're not even just using itjust to create. It's actually like whereyou are storing resources almost, too.>> Absolutely. And that is just step one,right? Step two isI think designers have to build theirown tools because the tools that we haveare never good enough. So, let me giveyou an example. So, this is my ownversion of Notion. And so,you know, getting this Yeah, my ownversion of Notion. I I I will tell youwhy I I created this. So, for example,on Notion, everything has to happen

00:07:31online. The AI is kind of like itdoesn't work locally. So,what I'm doing here with Codex isworking with my files that are on mycomputer in markdown and every time wecreate a file here, it appears here. So,when I did I I did the prep and I askedmy AI, "Okay, well, for every title, Iwant you to generate an image." And thenI also have a perfect prompt for theimage, which I also create a UI for it,in which I have a bunch of presetprompts that I have gathered over theyears.Including, for example,turning a design into a perspectivemock-up because it looks just waybetter. So, if you look at this one,if you look at this one, the first thethe first version looks like this, whichis like, as I mentioned before, AI slop.But then you turn into a pers-perspective mock-up and suddenly itlooks so much better. So, this is justone prompt. This is just one example ofwhat you can do with AI. And, you know,on top of that, you know, having toolslike, "Okay, you're doing apresentation, now you can draw on it."Okay, cool. Can also watch the notesbecause this is all part of your file,so you can just like if if I'm doing atalk, you know, I have my

00:08:46notes here, Ihave my version history here, and then,you know, I can see a history of thechanges that I have made,you know, a history of the design, if Istarted from scratch or not. And then,all of this is created locally using myagents, which started from Codex.>> I've almost built this so many timescuz I see the value in having all of mycontent locally. Like, I don't want ittrapped in somecloud-based product.I don't love Obsidian. I've tried somany times to have it be sticky, and Ijust don't love it. And I've only builtit as a Mac app, but you building thislocally in the web, I hadn't consideredbefore,and I'm pretty it's pretty compelling.>> What is cool when you build your ownapps is that you figure out like, "Okay,I have all of this information here.What ifI can just ask AIto use each section as a context tocreate my image. So, you can literallyjust create like a new block like this.And then you can just generate an imagewithout any prompt and with a bunch ofyour presets. So, for example, I needa YouTube cover or an article cover

00:10:01andthen I just generate on the fly. I don'tneed any context because the context ishere. So, when I'm going to begenerating this, it's going to talkabout what Codex actually did. That'sthe beauty of building your own tool isbig because you're not only doing thatout of frustration because something ismissing. You're not only doing thatbecause it's the only way to do itlocally because Codex works really welllocally and agents can do it far morepowerful than any private walled gardenthat exists out there in the cloud, butalso as you create those tool, yousuddenly discover so many new ways todesign such as this for example. I'm oneof those designers who discover as Iwork, right? I don't plan things. Idon't like have a giant blueprint whereI just plan all the ideas in advance uhbecause when when you work solo, this isyour superpower like this is how youmove away faster than anyone else. Andso, when you generate like this and thenyou can, you know, turn into a differentversion. So, now I can uh turn this intoperspective and you know, I can turn thevideo now. So, this becomes a video.Wow, it's all part of the same workflow.It's all part

00:11:17of the the tool thatyou're making which again starts fromCodex. I want to talk about the processof building the tools and actuallywriting code because I I totally forgot.Like of course you're using Sonnet 3.5last time we talked which now I have somuch respect for. Like at the time, thatfelt like magic, and now you look backand you're like, "Oh my gosh, it's sorudimentary." So, I salute thee forgetting as far as you did with Suno 3 5.My question then is, how does Codexshape the way that you approach actuallybuilding software now? How has yourworkflow there changed or as a result?>> Let's say you you start from scratch.What do you want to do at this point isto go to the plugins, and then you wantto do to find the toolsthat are part of your workflow, becauseeveryone is different, every computer isdifferent.And so, you want to download, of course,the main one, like computer use, butthen you have a lot of things like, I'msure you know Granola, if you work withthat, or Canva, or Figma,that you can install. If you want tomake videos, you want to use Remotion.If you want to use video again, you haveHyperframe. And then, if you like to usevoices, AI voices, or

00:12:32AI avatars, theycan be really, really powerful. Pluginsintegrations are like the same asplugins in Figma, is the powerful onescreated by startups or individuals thathave spent a lot of time integrating it.Skills, on the other hand, are the onesthat are kind of like prompts. A skillis is basically a prompt, sometimes alarge prompt. You can think about whatis the best like what is the bestpractice. It's more like a style guide.What is the best practice for taste,which font to use, you know, how do youguide AI to not create AI slop. So, theskill would be great for that. Okay? Thething is that if you always chase thetrends,you will become AI slopvery soon. So, my suggestion to you isalways be a step ahead. It's likeplaying chess. You have to be always astep ahead of what is the baseline. So,right now, the baseline isyeah, the AI slop, the enter fonts, thepurple gradients, and the you know, thatwould have would have been amazing 5years ago. Like if someone were usingpurple gradients, everyone was goingnuts over that. But nowadays it's AIslop, right? Very soon, you know, it'sgoing to be the serif fonts and the youknow, the same animations that AI

00:13:47useall the time. So you have to always beahead. And the only way you can be aheadis just to to to spend that last 10%.So, the way that I like to do it,especially as a designer or as someonewho haskind of like developed taste over somany years, is that I usemy memory. My memory is all of thethings that exist on my computer that Ihave gathered over so many years of mycareer, the fonts that I know, the filesthat I have created, the Figma designsthat I have done, the code that I havedone. And because it exists on mycomputer, I can just reference it andgive it to AI as a starting point. So,when you have a starting point, your AIwill stop making AI slop becauseyour starting point is you, and you aredefinitely not AI slop because you areunique as a human. For those who, let'ssay, are not designers just gettingstarted, take screenshots. If you lookat my at my Codex, right?Uh or or or cursor, right? Cuz I've beendoing a lot like really really uha lot of iterations.It always starts with a screenshot.Screenshot, screenshot, screenshots.

00:15:02See? Every single time I start with ascreenshot. The reason why is verysimple, it's because the screenshotis the fastest, easiest way to buildwithout having to voice it, withouthaving to type it. So, for example,let's say I want to fix an issue here.So, what I do? Well, in Codex, we have ashortcut, which is command command.Boom, it sends a screenshot right awayto Codex, just like that. And so, itmakes the whole thing so much easier. ItIt took a screenshot of your browseror whatever window that you have active.And then from here, you can just starttyping. You can just say, "Okay, I wantto fixuh the top left left part of this. Iwant to fix the video timeline." So, thequestion is, how knowledgeable are youof your product? Because if you're notknowledgeable about your product, forgetit. You're not going to do anythingmeaningful, and someone else in 3months' time will do exactly what you'redoing in one prompt. But, I canguarantee you that nobody can make DreamCut in one prompt just because I wentthrough so many iterations and so manyfeatures. And here's another question.How How many prompts do you think I usefor every project for any of these

00:16:17projects that I that I launched?>> I'm sure it isthousands.>> Yeah. At least 10,000 each.>> Yeah, I believe it. I totally believeit.>> Right. The one secret I use when I redidthis video editor, I did it in the day,by the way. It was ugly, but I did it ina day.And the secret was I searched thelibraries that were video editors, and Iwanted the ones that I that had videorecordingthat well, video editing, but alsoscreen recording, zooms, and to be ableto resize and all that stuff. And thenyou you spend the last 10% to do all ofthese extra features that didn't existbefore. So, for example, drag and dropor you know, to be able to re-reposition it, you know, and theselittle dots that grows>> What are like those dots? Yeah, that'snice.>> Right. And this is inspired This isinspired by Apple, but you will neverfind something like this in an existinglibrary. Why? Because it's not part ofthat personalityor that experience that you you yourselfhave by watching Apple work or bywatching your best friend designer workand then giving you that idea and thenknowing how to prompt it into existence.>> The only way cuz I know that you have an

00:17:32engineering background, right? Likeyou're not intimidated by code. So likeif you're working on the web app versionwithin that 10% are you writing any codeat all or are we prompting everythingfor the finer details?>> Right. So remember at the time I wassaying 95%of my code was written by me. Now0% of the code is written by me. 0%.Yeah, we have gone that far. And soyeah, I see I just noticed like a littleproblem here. It's it's overlaying. So Icould just take a screenshot and then,you know, based on the resolution of myvideo editor, we have these edge casesthat AI will never be able to fix. Ispend so much of my time testingeverything and just reporting bugs. Butnot just reporting bugs, but sayingexactly the right thing so that AIunderstands me. So part of the skillsnowadays is to be able to talk to AI inthe most efficient way possible. And oneof the ways is just yeah, takescreenshot, use your voice because whenyou use your voice you can explainthings much clearer instead of saying gohere and because you're lazy to to typeall of that stuff, you don't add all ofthose details that you can speak uh likethree or five times faster.

00:18:48>> There's one question that I can't stopasking myself. [music]What if companies applied to talk to yourather than the other way around? Andthat question is the foundation for theall new Dive Talent [music] Network andit's working. Like right now I'm helpingmany of the most exciting startups thatI know to hire the designers andbuilders who listen to this show. So ifyou're curious what might be out thereand maybe you want to get on my list, ormaybe you're even looking for your nextdesign hire, head to dive.club/talentto join today.>> I'm encouraged by your 0% answer,because I do think there's like a wholegroup of people that I have on the show,actually, where, you know, it's slowerto describe that you should change thepadding on something than to just go inand change it. And I'm like, yes, Itotally agree. But I think it goes backto something that you said earlier,where the big shift with these tools,for me, too, is I'm just doing way morethings in parallel. Like, I don't careif it takes more time for AI to changethe padding than if I wrote it manually,because I'm not sitting there waitingfor a loading state. Like, I'm allergicto loading states now. If I'm looking ata loading state, it means that I I needto spin up another agent, you know? Andso, I I'm okay with kind of doing asimilar workflow, as long as I'm able

00:20:03tobe making many different polishimprovements simultaneously.>> Yes, exactly. It's how precise you are.You know, back in the day, you know,when I say back in the day, I mean, alot of people still do it, you know,using Figma and moving rectangles andsort of like creating components, youknow, creating variables, and exportingcode, and doing everything manually.That skill is still super useful. So,let me give you an example. Nowadays,you can also annotate. So, now you canjust uh point something that you want tofix, and then you can say, you know, fixthis, uh fix the padding, et cetera.Instead of taking a whole screenshot andexplaining that, oh, this the part thatI want to fix is in feature templates,you annotate, and then boom, you can dothat. The best people are the editors,right? So, I I also think myself as amore of an editor than as someone wholikes to scratch from scratch. And also,you can totally say something like,okay, can you create, you know, a simpledesign generated in GPT image toolnowadays, right? Like for example, ifyou're creating like a 3D design, youwould start with this, you would startwith a screenshot, and then you you turnthat into a design. If you're creating

00:21:18abrand or a slide, then you start withthat. And then you sort of like slowlyand surely walk into the things that youneed, the inspiration that you need togather, and eventually get in get intothe final result. So, generating>> Do you ever do that for interfacedesign? Like are you ever exploringideas? Okay, so talk about that for alittle bit cuz I'm I'm doing a lot withlike spinning up quick HTML websitesjust to visualize ideas quickly, but I'mnot doing anything with GPT. But everytime I play with it to create aninterface, even if it's just as like amock asset for a website, I'm alwaysreally impressed, but I haven't used itas like this exploration mechanism yet.So, I'd be curious how that fits in foryou.>> GPT image two has become so good. If Ifyou look at the templates that we made.By the way, we're we're probably thelargest library of templates. We have5,000 templates. But you know, nowadays,most of these templates are created fromGPT image twoscreenshot to HTML.>> No way.>> is Yeah, yeah, yeah, this is this isthis is insane. The same way that I'mI'm doing that here. So, I would start,let's say, the content, I would startmaybe the PRD or the article, and I youknow, I get the the bullet points going,and then

00:22:33turning that into uha you know, a landing page or a herosection. So, for example, so let's say Ihave this one. I don't have uh you know,let's say let's assume this is this isan app. So, I would go insert an image,and I would be like, okay, because Ialready have the context of let's saythe app that I want to build. So, I'mgoing to go here, and I'm going toselect landing page, or I want to selectmaybe the the web design hero. And I'mgoing to use the taste skill if forthose who don't know, taste skill iswhat allows the AI to sort of pick theright fonts, the right colors, designsystem, and so on. I don't need thereference here, but just go with this. Icould have said, "Okay, my my app iscalled Aura, and then, you know, it'sit's a chat app, etc." And then it'sjust going to create a design for you.You can say, "Oh, I want a testimonialsection. I want the hero section." So,there's so many ways to generate So, forexample, I need a 3D version of theCodex logo, and I give the flat one, andthen I ask, "Hey, can you turn that intoglass? Can you turn that into, you know,3D perspective?" And so on. So, you canexplore a lot of things. Illustration,so so good. Slides, interior design,wireframe. I love wireframe.>> Okay, I have another question. Cuzyou've clicked on enough of these

00:23:48pop-ups now where like the way that thatpop-up enters is more drawn out andelegant than almost any I've ever seen.So, talk to me a little bit about that.Like, how do you That's harder to dowith just screenshots, right? Like, howare you finessing that level ofinteraction and animation with AI?>> I'm going to give you another secretsauce that you can do today. This usedto have you needing to learn all aboutthese libraries, these animation terms.Now, you know what our team use? Theyuse a video.They give the AI a video, and they askthe AI,"How should I prompt this animation?"Isn't that incredible?>> It's such an important tactic becauseit's like everything comes down tolanguage at the end of the day. Like, Iwas working on that with shadersrecently. I don't have any of thelanguage for shaders. It's like I almosthad to use Claude as an intermediary tobe like, "Give me some I'm trying toachieve X. I have no words for it. Like,maybe ask me some questions, butultimately I need you to write thisprompt. Like, I'm not going to be ableto write this prompt." And I love thattactic for something like interactiondesign. Makes a ton of sense.>> More than that, if you look at the wayAI works, it's not getting any smarter.Like let's take an example

00:25:03of Fable 5,right?You use Fable 5right here, and you ask it, "Can youcreate a beautiful landing page?" Thereis zero difference between that landingpage and one created byOpus 4.8 or Sonnet 4.6or 4.5.The reason why is because it's likeyou're using someone far too powerfulfor a simple task. The only way you canutilize these super intelligent modelis by giving a giant piece of prompt,literally. So,you would select all of these contexts,all of the, you know, the sections thatyou want, all of these pre-prompts thatI already have in my template, and thenyou put all of these giant prompt withthe design system design.md, and thepreview of the HTML, of the reference,or the screenshot, or you can import theURL, and then you can fully utilize amodel like Fable 5. Otherwise, there'sno point. You're paying for nothing.>> Can we talk about the design.md piecefor a second? I I think it's interestingthat you've fully productized that inAura. For people who hear it, maybe theysee it on Twitter, you can intuit whatit is, right? But like how much strategyactually goes into making

00:26:19an effectivedesign.md file?>> 10 years ago, 20 years ago, all thedesigners, they have a style guide. It'skind of like their guiding principle oftheir mastery of a designer. Right? Yougo to school, you you work on 10,000designs, you have embedded in your bloodthat design language, that style guide.The design.md is basically all of thatsweat, tears, and blood into onemarkdown file that the AI can read. So,that's what design.md is. So, why is itpowerful? Like I mentioned before, ifyou're using a model like Mythos fablefive, then you give something likedesign MD, it will understand it and itwill be able to take it to the nextlevel. If you use a base model, it won'tunderstand it. So, that's kind of likethe difference. So, design that MD isthe design language, the design systemthat holds all of that beautifulinformation about the colors, thetypography, the spacing guide, the thecorner radiuses, and you know, is itwhat surface and it goes into greatdetails. And Google started this thingbecause it needed to standardizethe language so that everyone followsaccessibility

00:27:34and composition and itneeds to include all of these,you know, segments and all of thesetokens that we designers have kind oflearned over the years, but if you're anew designer, you kind of lost all ofthat context. So, the design that MDholds [snorts] all of that informationso that you don't have to.>> Yeah, this is something that I think Icould probably take more seriously in myown work, honestly, cuz I I think I relyon Claude to almost piece it togethereach time I need it. Like if I'm likeworking in paper, like I do a lot ofvisualizing ideas in paper and I'malways like, you know, use the designsystem, use our typography and I almosthave to kind of list things out. I'mlike, I should probably just wrap thisall into a skill. Say, use the design MDand then do the thing that I've asked.This is my to-do list that I'm addingfor myself after I hang up here.>> I mentioned as a designer, I always workwith a previous work that I've done inthe past. So, whether it's the fontsthat I gathered, whether it's the colorsand the Figma files that I have done inthe past or the HTML and project, I doexactly the same withyou know, with Codex, which is that Ithink I would I would say about 50% ofmy prompt is me referencing another chatoranother

00:28:49page that I do. So, I mentionedbefore that if I start a new projectlike a a new iOS app, I would referencean entire folder. Right, all of these,by the way, are folders in yourcomputer.When you start a new project, youreference an entire folder. You can saysomething like, "Okay, I want an iOS appbased on my Mac app that I already donethis in this folder." But, when you doiterations, this is where you startto to to reference another page. So, onething that I love to say is"Create this page or this model or thisbutton in the style of another buttonthat I created in this page." And so, itwill do exactly that. Or, you can alsobring the design.md, but then again,design.md is powerful at the beginning,but when you're doing a lot ofiterations really quickly, you don'twant to always reference that. You cantotally save that into your agents.md.For those who don't know, agents.md isthe file thatis always read by by the AI for everysingle prompt or anytime that you createa new thread or a new task. So, this isa project, this is a thread, and thenyou have like multiple tasks within athread. I'm not a person who likes toput

00:30:04too much stuff in agents.md.Therefore, the way that I work is justtaking a lot of screenshot and thenreferencing a lot of files. You know,like, this is where my file is, this iswhere the file the page is, this is thebutton that I created. And then, justremembering the most important terms,design terms and details. You know, youask about the model. Eventually, if youwant to stop having tofeed a video every single time youprompt, you have to remember that. Or,another thing that I love to do is like,I can say something like, "From now on,I want you to remember this workflow inmemory or this design system in memory.They could have a agents.md inside aproject, which is kind of cool.Uh, sometimes I call it a runbook.md.A runbook is more like a workflow.Sometimes I call it workflow.md, butgenerally speaking it's agent.md. Youhave a main agent.md for all of thethings. So, you can totally go tosettings and go to your configurationand then put that into you. That it's abit dangerous because it's red for everysingle problem or every single newproject. So, instead what I do is I Iask the AI to remember in memory aboutwhat

00:31:20we just did, right? So, let's saywe just created this this model and thedesign or this button. I can say, "Iwant you to remember this and save it toyour memory." And then usually what itwould do is it would save that in in inthere in the agents.md for this specificproject, so that the next time that it'screating a model, it will remember toto do exactly in this style.>> We've covered a lot of ground, a ton oftactics. I want to make sure that we'veyou're not missing anything, basically.So, like when you think about the thelittle tools or tactics or workflowsthat designers can adopt to moreeffectively collaborate with and buildwith AIs, or anything else that you wantto make sure that we talk about beforewe move on.>> I'm usingGPT 5.5 extra high. So, we haven't talkabout this too much. Like, what is thedifference between the model? So,obviously when you first start, youmight want to start with medium becauseit's not too costly. It's not going tohit your limits.Not everyone's going to have like thesort of $200 or $100 plan.So, what I would do is I would startwith medium because it doesn't eat a lotand it's reasonably good. It's going tobe able to do most of the things, butmuch faster as

00:32:35well. Another thing youcan do is to go to cursor and usecomposite 2.5, which is super fast. Andbecause it's so fast, I can just focuson one project at a time. But when I useextra high is, by the way, to me,GPT-5.5 is like probably the best modelthat I've used this year.Without counting the new mythos and allthat stuff, but they're very expensiveanyway, so I would probably not use themat this time. But I've been using extrahigh and it's so, so good. It's able tobuild anything you can dream of today.But most importantly, have a an endlessamount ofof curiosity about what's going on andhow you can perfect your own workflowbecause it keeps evolving. And becauseit keeps evolving, it always requiresyour attention. I've never been work sohard in my life. I've never put so manyhoursinto the projects. I've never built somany projects at the same time. I'venever doneanything of this magnitude solo.Usually,you know, when I first startedDesigncode, right? I was traveling for 2years and you know, I would spend all mytime sort of like working on like smallpockets of the content creating my firstvideo. And I remember it took so long.Eventually, it made enough

00:33:50money. Ibuilt a team. We had a team of editors,video editors, and you know, to save metime. But today, you can do everythingyourself because your editors are youragents.Your designer [clears throat]is your agent. Your coder is your agent.Your personal assistant, of course, butanything you that you can think of, youragent can do it. And the only thingthat's stopping you is you opening tothis world of apps and plugins andlibraries anduh prompts and design.md. And if youhave that, it's it's your greatestweapon.>> I can't think of a more inspiring way towrap this up. And maybe even justsomething that I appreciate you likescrolling through your Twitter feed. Imean, there's a lot of people who areobsessing over the workflow, and thenyou kind of have this question about,like, well, okay, but what if youactually shipped, you know? And man, youare the complete opposite, where you'rejust a shipping machine, you know,you're just putting really interestingand valuable things out left and right,and I have so much appreciation andrespect for what you're doing assomebody who's not only curious, butjust goes and runs through walls andmakes things happen. You're superinspiring to me, and I know that a lotof people listening feel the same way,so I appreciate

00:35:05you coming on and andjust sharing a little bit about howyou're thinking about things, how yourworkflow's evolving, what you're workingon. I think we should make it a maybelike an annual checkpoint, becausethere's always so much to learn fromyou.>> This is such an honor, and I would loveto come back like next year and give youan update on what's going on with withAI and sharing my new workflow at atthen, because, you know,16 months ago, we used Sonic 3.5, nowwe're using GPT 5.5 and Codex, but wedon't know what's going to happen in 12months.