Design at the Speed of Adjectives — Paul Bakaus, Renaissance Geek, Inc.
AAI Engineer
Computing/SoftwareSmall Business/StartupsPhotography/Art
Transcript
00:00:00Let's get this going. Welcome, everyone. Thanks for coming. My name is Paul, and I'm going to try
00:00:21to save you from the mid-day slump. So let's see if we can do that. Okay, so this talk, this cold
00:00:29open of this talk will make a lot more sense if you know what impeccable is, but if you
00:00:35don't know what it is, it's a design skill that turns your coding harness into a better
00:00:39designer, and hopefully you into a better designer, and you can try it out. It works
00:00:43across all harnesses, Cloud Code, GitHub Copilot, Cursor, Codex, etc. And this talk is really
00:00:51about my approach to how to building this tool as opposed to the tool itself. But let's get
00:00:59started with a pretty typical website. So this is a intentionally very basic website, and
00:01:05I want to give you that one as context to what I'm going to show you next. So, you know, it's
00:01:11an okay website. It's not great. It's pretty bland, but at least it has some color, something.
00:01:19So now, here is the command, make it bolder, make the workflow section bolder, with impeccable. So this is with impeccable installed. So now, there are, it's not perfect. There are some problems with this. For example, the section numbers that you have there, that's a very typical AI
00:01:41loop tell. That GPT loves to do. And I mean, Claude actually loves to do as well. And it
00:01:48sometimes creeps in. So, I wouldn't say this is perfect by any means, but I think it's, I would say, bolder than before.
00:01:54Now, a lot of people install impeccable, try it out, and then they ask me, well, I, you know, like, maybe it could be placebo. Maybe the model is good enough in itself, right? Not sure if it actually works. Well, let's compare it to, this is exactly the same project, GPT 5.5 on extra high, same exact prompt without impeccable.
00:02:16So, there is a slight difference, I would say, but you be the judge of that. So, the before and after is pretty different, and I'm going to talk about why that is, and what the approach is.
00:02:35So, designing at the speed of adjectives, and to get the elephant out of the room, they are not just adjectives, they are also verbs.
00:02:44But some of them are adjectives.
00:02:49First off, the role of the engineer and the designer are blurring. I have seen more people become design engineers than ever. Designers are moving into code, engineers are moving into design, and vice versa.
00:03:02Some of the best comments I get from people who walk up to me and say, like, well, impeccable gives me this shared language of design that everybody can communicate now in.
00:03:11But really, these worlds are all colliding, in my opinion, and the typical handoff is kind of broken. Like, the waterfall process of, I start with design, maybe I start with, like, the what on the PM side, then it goes into design, and then there's, like, this handoff and this critique process, and then, ultimately, the engineer builds it.
00:03:30So, that thing is just not real anymore. I don't know about you, but I think that's starting to collapse very quickly in small and large companies.
00:03:41And when it comes to the actual work of design, I feel like there's, right now, kind of like two worlds that you can play in.
00:03:48Either it's direct manipulation, sort of like in the pixel space, right, so you're in Figma and you're manipulating margins, paddings, whatever it is, or you're in Webflow or you're in some other tool where you're directly manipulating the final output or the design.
00:04:04Now, that's fine, but then you feel like you're getting behind because everybody else is, like, agentic and going crazy with loops and loop maxing, right?
00:04:13And now you're going on the other extreme, which is simply telling your agent, please design this for me or please build this.
00:04:18But it feels like, to me at least, there's this middle ground that we haven't explored enough.
00:04:24Like, what is the exact level of control? How can we insert the human in the loop at the exact right time?
00:04:32And so, this came to mind then. It was a very popular tweet a while ago.
00:04:38So, centering a diff with Opus, yes, not a great use of Opus.
00:04:44But, you know, it kind of speaks to that problem. It's like, okay, well, when do we use which tool?
00:04:51When is which tool adequate to do what job?
00:04:56In my opinion, right now, for most work, the direct manipulation of padding, margins, spacing, and so on,
00:05:03is actually at an elevation that's too low.
00:05:08And on the other hand, if you go fully agentic, you get web pages like this.
00:05:15Now that you would get this in, like, 2022, I would say.
00:05:17So, everybody here can point out the typical AI slot patterns.
00:05:21But the purple gradients are actually way beyond that.
00:05:24There are no purple gradients anymore in most frontier models.
00:05:28Now we just get this, and it kind of wears a different suit, as Claude would say.
00:05:33But it really is slop as well.
00:05:37It's just a moving target.
00:05:39So, now we have, I call this Claude beige, we have these instrument serif fonts, we have italics.
00:05:44And it's not necessarily a bad design.
00:05:46It's just everything that comes out of Claude design looks that way.
00:05:49And when everything looks the same, you kind of get algorithmic Uniqlo, as I call it.
00:05:55So, also not great.
00:05:57Now, here's my first hot take of today's session.
00:06:01You cannot one-shot design.
00:06:04Now, many people want to, and maybe will get there eventually.
00:06:08I just don't think it's possible at the current moment.
00:06:11Maybe not ever.
00:06:13I think in order to build something that really looks and feels great and is effective design,
00:06:19you need to have something that's context-rich, which has been informed by what you're trying to do,
00:06:26what your audience is, what you're trying to build.
00:06:28And also, it will have to be multi-shot.
00:06:31You will have to iterate on your design.
00:06:33And you, I mean, even if you're a solopreneur, but if you're not a solopreneur,
00:06:38you might have different people who have different opinions.
00:06:40Users have opinions.
00:06:41I mean, design is messy.
00:06:42It's process.
00:06:43And wishing it away and just saying, like, solve this problem for me, agent,
00:06:48will not, unfortunately, solve your problem.
00:06:51Now, the other problem is, if you don't know the language of design,
00:06:57or if your stakeholders don't know the language of design,
00:07:00it gets really frustrating really quickly, too.
00:07:02So, if you've ever designed something, you kind of know, like, this kind of experience.
00:07:06And then, you know, the inner feeling builds up of, like, mmm.
00:07:11And at some point, you're like, ahhh.
00:07:16So, really, what you have to ask first is, what's the emotional territory?
00:07:20What should this never feel like?
00:07:22What's the reference?
00:07:23What's the audience?
00:07:24These are all really important questions to figure out.
00:07:27Like, you can't go to a human design studio, even if it's the best design studio in the world,
00:07:33you go to a design director, and you say, like, you know what?
00:07:36I want a design for my brand.
00:07:38And the design director nods.
00:07:39It's like, got it.
00:07:41And then walks away.
00:07:42That doesn't make any sense.
00:07:44Like, they obviously have to figure out what it is that you're trying to build
00:07:47and what you're trying to get out of it.
00:07:50Importantly, on this example, even sort of like the 2026 version of AI Slop,
00:07:55you know, quickly vibe-coded pitch, nobody decided anything here.
00:08:00Like, you just one-shotted this.
00:08:02Nobody decided anything on this page.
00:08:05This was just, I mean, it's maybe competent, but completely empty.
00:08:10So, my thesis that I make here, and what I've been trying to build over the last couple months,
00:08:17is a tool that gives you just the amount of control needed to steer your agent in the right direction,
00:08:23with adjectives and verbs.
00:08:26And these are the types of words that I'm talking about.
00:08:29I'm talking about words like, make this bolder, make this quieter.
00:08:32Distill this.
00:08:33So, basically, simplification.
00:08:35Polish this.
00:08:36Make it denser or harden this.
00:08:39Harden, for instance, would mean make sure that this design works across the board.
00:08:44Maybe the performance is bad on certain devices.
00:08:46Maybe it's not responsive.
00:08:48And so, these are words that we use in the world of design.
00:08:54And I noticed, when seeing two people attempt the exact same task,
00:08:58and one is an engineer who has never really touched design at all,
00:09:01and the other one is a designer,
00:09:03there's a stark difference, even when they use the same model and the same harness everything,
00:09:08in the output that they're getting based on the language that they use.
00:09:11So, I've been trying to put that language, basically, compress it into a skill
00:09:18and into a system to be able to express yourselves better.
00:09:22And so, for example, you could have, like, a baseline over here,
00:09:27but then you make it bolder,
00:09:28and the agent actually knows what bolder in your context means.
00:09:32You make it quieter.
00:09:33It knows what it means.
00:09:34You distill, and it knows what it means.
00:09:36But then, bolder how?
00:09:42What does bolder mean in the context of your project?
00:09:44That's a really important question,
00:09:46because, as you could tell at the beginning of the demo,
00:09:49in the cold open,
00:09:51when you just ask GPT or Claude or whatever, make this bolder,
00:09:55it doesn't know anything about what you mean.
00:09:57Now, it has its own sense of reference of, like, what bolder could mean,
00:10:01but it will do whatever it wants, right?
00:10:03It will invent new colors.
00:10:05It will invent new colors.
00:10:06It will invent new gradients.
00:10:07It will -- I mean, it doesn't matter.
00:10:10What it actually means in impeccable, not gradients, not glass, not neon,
00:10:17but actually hierarchy, scale, decisive type,
00:10:20so things that don't break your design system
00:10:24and adequately raise the attention that the user would put on it.
00:10:29And this is an actual sentence from the file that is loaded when you say bolder and impeccable.
00:10:38Show someone your work and say, hey, I make this bolder.
00:10:42I make this bolder.
00:10:43If they believe you, you failed.
00:10:45So this is something that we actually instruct the agent to do for themselves.
00:10:49And then the agent oftentimes reflects and, like, hmm, I don't know.
00:10:54I think I did some bad work here.
00:10:57An adjective is a light word.
00:11:00And a light word, such a German term, and Matt Pocock actually talked about this recently,
00:11:06who also makes some really excellent skills.
00:11:08And so this is stolen by one of his tweets.
00:11:11But I really think it makes a lot of sense.
00:11:15These are words that you basically imbue with meaning, that you infuse with meaning,
00:11:21and they mean something.
00:11:22They already mean something to the model,
00:11:24but then you translate this into the area that you're interested in.
00:11:29And the right altitude really shifts.
00:11:34Now, I don't think you can solve any problems at that altitude.
00:11:39That's super important, too.
00:11:40There is still room for the very beginning.
00:11:43For instance, if you just want exploratory work and you just want to get something on the screen,
00:11:47or you just want the last amount of polish.
00:11:50I don't think, right now, AI is good enough to replace humans for the last 5%, 10%, maybe even 20% of the work
00:11:57to get it from good to great.
00:12:00But it does actually make it better.
00:12:04Now, I've built tools for more than 20 years.
00:12:07I guess I'm dating myself here.
00:12:10But whenever I built a tool, I actually map the workflow first.
00:12:13I think about what is the actual workflow of the target user that I'm trying to help.
00:12:17And the workflow of design is not linear.
00:12:20It's messy.
00:12:21It has many different stages.
00:12:23There are different people involved.
00:12:25And so, I kind of mapped every part of the process and thought about,
00:12:28well, what are the injection points along that route?
00:12:31And how can we help along that route?
00:12:34And so, you start with initialization, with the shaping process of design.
00:12:38You go over to crafting, to iterating, with all these verbs and adjectives,
00:12:44and then to the actual hardened and polished phase.
00:12:46And finally, you kind of bring it back to your design system.
00:12:49You do some tech that clean up, design that clean up.
00:12:53Now, here's one example of a command that I wasn't actually sure about shipping.
00:12:58And it's called overdrive.
00:13:01I basically built this command in impeccable that tries to create something completely over the top.
00:13:08And it was kind of like a half of a joke, maybe?
00:13:11But people absolutely are rabid about this, about this thing.
00:13:15And they love it.
00:13:16And so, this is just one example of, I mean, this is not one-shotted.
00:13:20This is a sort of realistic event horizon shader that I've built for my shader library, radiant shaders.
00:13:28But I did actually use the overdrive command for this to overdrive it even more.
00:13:33And make it more ridiculous.
00:13:34So, sometimes I come up with these commands that I'm actually not sure about.
00:13:38And I test them with the community.
00:13:39And once it sticks, I realize, okay, well, there are more people who are getting joy out of this.
00:13:44Oh, here we go.
00:13:49Here we go.
00:13:50Okay.
00:13:51An adjective with nothing behind is just a nicer prompt.
00:13:54So, you really have to tell the agent what you mean by the word that you're using.
00:13:59It has so many different meanings within its training materials that you really have to tell it.
00:14:05Now, here's another thing.
00:14:08I've gotten so many requests for an automatic way of using Impeccable.
00:14:14So many.
00:14:16It's so frustrating.
00:14:17Like, at least once a week, somebody tweets at me and says, you know what would be cool?
00:14:21Just let Impeccable do all the work by itself.
00:14:24I don't have to decide on any commands anymore.
00:14:26That's not the point.
00:14:27The point is to give you a way to steer what you want to end up with.
00:14:31It's never going to be a tool to one-shot design.
00:14:34That's not the intent.
00:14:36And I'll never -- I mean, there's a pull request standing out right now that I'm not going to close because of that.
00:14:41There is no auto and there will be no auto.
00:14:46In my opinion -- and this is a bit awkward maybe for some of the next speakers -- but I think taste cannot be amplified.
00:14:54Sorry, I think taste can be amplified.
00:14:56I think I call this amplified craft.
00:14:58And I think you can work with tools to sharpen your taste and bring out more of your taste.
00:15:04I really do not think it can be lab-grown.
00:15:07I think it is taste by definition sort of is contextual.
00:15:12It's cultural.
00:15:13It's scarce.
00:15:15And once everybody replicates the same thing, it starts to become really muddy.
00:15:20And we don't think of it as taste anymore.
00:15:23That's my opinion.
00:15:24I know this is like a hotly contested topic, but I'm not trying to self-taste with my tools.
00:15:32That's what I got.
00:15:33This was about finding the right level of control for design.
00:15:36Thank you very much.
00:15:37And I think I have time for a few questions if you have any.
00:15:40Thank you.
00:15:54We'll see you next time.
Community Posts
No posts yet. Be the first to write about this video!
Write about this video