Anish Acharya | The a16z Show Summary

The a16z Show | Anish Acharya, John Maeda, Paul Bakaus

1. Key Themes

Design Vocabulary as a Competitive Moat for AI Output Quality

The single most actionable insight in the episode is that prompting quality — not model capability — is the primary variable in AI design output. Designers outperform engineers when using the same model because of linguistic precision.

"Designers when using Claude, as opposed to engineers using Claude, would consistently get better results. And it's because of the language that they use... engineers don't use the words. Things like vertical rhythm or negative space or make this bolder or quieter. They don't have the same vocabulary as a designer who's been in the game for a long time." — Paul Bakaus [00:08:18]

AI Slop Is a Moving Target Shaped by Defaults

What counts as generic AI output shifts constantly, and the root cause traces back to framework defaults — not model failure. This means "anti-slop" tooling requires constant updating, not a one-time fix.

"In 2022, 2023, everybody thought AI Slop was purple gradients... The models aren't producing purple gradients anymore. The world has moved on to beige backgrounds, tinted backgrounds, Instrument Serif, eyebrow text... Slop is a moving target. Every time you move to another part of the latent space and everybody adopts it, you get something like algorithmic Uniqlo or Ikea." — Paul Bakaus [00:15:49]

"Tailwind's default theme has purple. Because of the default theme, the purple gradients were introduced everywhere... when I started jQuery UI, we created a theme framework and the default theme was orange. I colored the whole web orange overnight, and I learned my lesson. The power of defaults." — Paul Bakaus [00:18:04]

The Shift from UX to AX (Agentic Experience)

The next major design discipline is not visual at all — it is designing interfaces for agents: APIs, CLIs, error messages, and information architecture. Visual designers are better positioned than engineers to crack this.

"I've been pushing moving from UX to AX, agentic experience. And agentic experience is non-visual. It is the world of robots.txt. It's LLMs.txt. It's command line dash dash help. It's a world where we're designing for agentic affordances instead of just visual affordances." — John Maeda [00:19:22]

"Agents themselves are not really good at designing those interfaces yet. So they're not really good at designing API interfaces, designing CLIs. And that's kind of ironic because they're the user for them now." — Paul Bakaus [00:26:13]

LLMs Know the Output of Humanity, Not the Input

A structurally important limitation of current models: they learned from finished design artifacts, not from the reasoning that produced those artifacts. This means AI can approximate taste but cannot replicate design judgment.

"LLMs have been trained on the output of humanity, not on the input. So what led to a design decision is not something that the LLMs know. So it's an approximation of taste... amplifying human taste is an interesting problem — a more interesting problem than creating a replication of taste at the model layer." — Paul Bakaus [00:37:06]

Cognitive Delegation vs. Cognitive Surrender

A crucial distinction for anyone building AI-assisted workflows: delegating tasks to AI while retaining intent is healthy; letting AI set the agenda is surrender. This maps onto product design choices.

"Cognitive delegation is — I'm using Google Maps and Google Maps tells me how to get somewhere as quickly as possible. But what if I let Google Maps decide where I want to go? Now I kind of cognitively surrender to the application... with LLMs, you prompt something in plan mode and it creates a beautiful plan that's eight pages long. You're not going to read through that plan. You scroll, skim it, and it's like 'I guess the model knows what it's doing,' and then just click okay." — Paul Bakaus [00:34:39]

Raising the Floor Creates a Ceiling Opportunity

As AI makes average design ubiquitous, the last 10–20% of uniqueness becomes the only real differentiator — and the craftspeople who deliver it will finally be visible in a way they weren't before.

"If everything looks relatively polished and clean... then the only differentiator will be whatever we craft on top of that... I spend a lot of time crafting something really unique — a slider that is beautifully animated — but oftentimes that work would drown. It's the moment for these people right now to shine." — Paul Bakaus [00:23:21]

Taste Emerges from Scarcity — and AI Eliminates Scarcity

A philosophical reframe: traditional taste formation depended on scarce materials and cultural maturity. AI makes all materials universally available, which breaks the historical mechanism through which taste was generated.

"Design in the European sense came from royalty and the desire to be distinctive — I've got fur, I've got gold, I've got emerald — because they were working with scarce materials. What's interesting about this era is that this idea of taste doesn't fit when all the materials are available to everyone." — John Maeda [00:39:23]

The Bespoke Software Economy Is Finally Arriving

A prediction John Maeda made in Forbes in 2010 — that software would return to a cottage industry with many small, opinionated apps — is now materializing, enabled by AI lowering the floor of entry.

"The software industry is poised to embrace its craft heritage. By 2020, software will return to a cottage industry with bespoke apps made by many. We will discover the value of authorship." — John Maeda, quoted from Forbes 2010 [00:31:26]

"Products with small but still compelling markets — the million-dollar TAM products, the $100,000 TAM products — because they were economically infeasible before... a digital main street that can support a product that may have smaller scale, but be more opinionated." — Anish Acharya [00:31:03]

Human Trust as a Premium Signal in an AI-Saturated Market

As AI design floods the market, consumers will pay a premium not just for craft but for provable human accountability and trust embedded in a brand.

"The reason why someone's willing to pay that much more for that bespoke, better thing is they're paying for human trust and accountability. So for brands and products where people demand that level of 'can I trust this?' — that human smell could become even more valuable than ever before." — John Maeda [00:31:06]

2. Contrarian Perspectives

The Next Miyazaki or Billie Eilish Will Not Come from AI Tools

Despite enormous investment in creative AI, no AI-enabled tool has yet produced a culturally transformative creative voice. The thesis that AI democratizes creativity has not produced ceiling-level creative output — only floor-level volume.

"We've seen so many AI startups that are trying to enable creative work — whether it's music or video or whatever. But we haven't really seen the next Hayao Miyazaki from Runway or some other video startup. We haven't seen a music AI startup's next Billie Eilish or Coldplay. A lot of focus right now is raising the floor. There's too little time spent on raising the ceiling." — Paul Bakaus [00:21:53]

Motion Design Remains Genuinely AI-Resistant Today

Counter to the prevailing narrative of AI capability across all design domains, temporal and motion design are structurally unsolvable with current vision models because they lack true temporal understanding.

"Right now there's no good feedback loop for motion. I had a shader animating rain upwards after working with the model for a bit. It's not the model's fault — the model takes screenshots and thinks 'I guess that looks like rain.' But it doesn't understand temporal resolution. Gemini does a little more — it samples frames at about one frame per second — but even that doesn't really give you the full picture." — Paul Bakaus [00:27:34]

High-Craft Design Will Survive AI But Its Market Will Shrink

Rather than becoming more valued, elite human design craft will become a niche cottage industry — like letterpress printing — admired but not economically scaled.

"The number of people who will raise the ceiling will not be a high number and it won't be that valuable in the mass market. So I think it'll never be automated — and that's good news for people who want to stay in the craft. Only problem: the customer base will be smaller. It's like letterpress printing — not many people will pay for it. So it's that market mechanics." — John Maeda [00:29:44]

API and CLI Design — Not Visual Design — Is the Urgent Design Problem

The most under-resourced design problem is not consumer UI but the interfaces that agents use: APIs, command-line tools, and error messages. This is where design thinking is most needed and least applied.

"Agents themselves are not really good at designing those interfaces yet. And that's kind of ironic because they're the user for them now. Information architecture is another great example — it's a certain type of designer that's very good at crafting a good navigation, and they have to think differently." — Paul Bakaus [00:26:13]

3. Companies Identified

Impeccable

An open-source AI agent skill for design quality, built by Paul Bakaus. It injects professional design vocabulary into agent prompts, adds a quality/anti-slop layer, and provides a visual iteration mode. Currently integrating with GitHub Copilot.

"It is an agent skill that's completely open source. It has lots of subcommands to steer. And then I kind of built a visual iteration mode on top of it and a quality layer that removes slop... it does stop the overfitting of the models. So not everything is Claude beige, as I call it." — Paul Bakaus [00:10:08]

GitHub (Copilot App)

Microsoft's developer platform, specifically the new GitHub Copilot app, which is live in GA and being designed to a high craft bar. Impeccable is being integrated as its first design skill partner.

"We were launching this GitHub Copilot app that's already in GA... The craft bar is very high now for applications. So we wanted to set the bar with this app... Paul is going to be the first one to be in our integrations of this powerful design skill." — John Maeda [00:13:05]

Suno

AI music generation company that began as a consumer vibe-music tool and has evolved into a professional music production platform used by real producers.

"Suno is actually a great example of something that started for beginners and sort of vibe sound designers that wanted to just one-shot a quick song. But now it expanded very much into real music production. There are producers using it for all sorts of things now. They've built a whole studio around it and they are really trying to push the last 20% now." — Paul Bakaus [00:28:31]

Tailwind CSS

CSS framework whose default purple color theme is argued to be the root cause of the "purple gradient" era of AI slop — a powerful illustration of how defaults shape entire design epochs at scale.

"Tailwind's default theme has purple. Because of the default theme, the purple gradients were introduced everywhere." — Paul Bakaus [00:18:04]

jQuery UI

jQuery's UI component library, whose default orange theme caused Paul Bakaus (its creator) to inadvertently color large portions of the early web orange — his personal lesson on the outsize power of defaults.

"When I started jQuery UI, we created a theme framework, a theme roller, and the default theme was orange. I colored the whole web orange overnight, and I learned my lesson. The power of defaults." — Paul Bakaus [00:18:04]

Notion

Mentioned as a destination for alumni of GitHub's design team, specifically Max, who led design at GitHub before going to Notion — signaling GitHub's historic design culture and its broader influence.

"There's Max at Notion who used to lead design at GitHub. There's all the Brassell people. They're all over the place." — John Maeda [00:46:11]

Adobe / Photoshop

Referenced as the historical analogy for how great API design (via Apple's Quickdraw) enabled an entirely new category of software — and how the right primitives, not just features, determine platform success.

"Photoshop happened because of the design of Quickdraw's API. It was impossible for there to be a Windows version of Photoshop because it didn't have the same API shape and DirectX." — John Maeda [00:40:45]

Runway

AI video generation company, cited as an example of a creative AI platform that has not yet produced a culturally transformative creative voice analogous to a major filmmaker.

"We haven't really seen the next Hayao Miyazaki from Runway or some other video startup." — Paul Bakaus [00:21:53]

4. People Identified

John Maeda

VP of Design at Microsoft / GitHub. Former partner at Kleiner Perkins, former president of Rhode Island School of Design, MIT Media Lab alum. A pioneering thinker on computational design, author of the annual Design in Tech report, and originator of the "UX to AX" framework.

"I've been pushing moving from UX to AX, agentic experience... a world where we're designing for agentic affordances instead of just visual affordances. That force multiplication will occur once we can retire more of UX." — John Maeda [00:19:22]

Paul Bakaus

Founder and CEO of Impeccable. Former creator of jQuery UI (which colored the web orange). Serial engineer-designer hybrid. His Impeccable tool is the first design skill integration for GitHub Copilot.

"I noticed that whenever I would tell Claude or Codex 'make this better,' it would do a really terrible job. Because it didn't know what angle of better... I brought that vocabulary to the actual agent harness. And that was the first iteration of Impeccable. And that already made a huge difference." — Paul Bakaus [00:08:18]

Muriel Cooper

MIT Press and MIT Media Lab designer (historical). Predicted the desktop publishing and electronic publishing revolutions. Founded a lab at MIT combining AI and design automation — effectively the original "auto design" researcher.

"She was the one who imagined a world where people wouldn't want to use the terminal and might want to use this thing called Helvetica to look at things on the screen. She predicted the whole desktop publishing revolution... She also had people from MIT's AI lab asking how do we create things by machines that humans are good at?" — John Maeda [00:02:59]

Bill Atkinson

Apple engineer (deceased) who created Quickdraw, Apple's graphics library. His API design choices made Photoshop possible on Mac and impossible on Windows — one of the most consequential API design decisions in computing history.

"Bill Atkinson, the engineer who recently passed away, worked on the Macintosh, specifically Quickdraw, which was Apple's graphics library. Photoshop happened because of the design of Quickdraw's API. It was impossible for there to be a Windows version of Photoshop because it didn't have the same API shape." — John Maeda [00:40:45]

Kai Krause

Creator of Kai's Power Tools, the Photoshop plug-in that dramatically expanded Photoshop's addressable market by enabling non-photographers to do algorithmic image manipulation.

"When something called Kai's Power Tools came out, it blew up the TAM for Photoshop because people could do weird algorithmic things with their faces or all kinds of filters." — John Maeda [00:10:44]

Eric Spiekermann

Renowned German typographer, designer of the Meta typeface, founder of MetaDesign. Cited as an illustration of design operating at the level of abstraction — designing the system for design, not just design artifacts.

"There was a famous typographer, Eric Spiekermann, who designed a very popular sans serif typeface called Meta... His company in San Francisco was called Meta Design. At the time, one of the top design firms in the world." — John Maeda [00:20:06]

Donald Knuth

Computer scientist and creator of TeX typesetting system. Cited alongside Muriel Cooper and PostScript creators as an exemplar of the rare combination of mathematical and design thinking.

"Donald Knuth, before that with TeX. He was an incredible typographer, mathematician. So TeX was a miracle in the combination of function and form." — John Maeda [00:12:08]

John Warnock

Co-founder of Adobe, creator of PostScript. Cited as an engineer who understood graphic design deeply enough to encode its primitives perfectly, enabling the desktop publishing revolution.

"PostScript was a miracle because the Adobe guys, the Warnock — the manual for PostScript is beautiful, incredibly well designed. Those engineers understood graphic design better than most graphic designers. PostScript properly encoded the perfect set of primitives to implement visual graphic design of the day." — John Maeda [00:11:40]

Jay Parikh

John Maeda's boss at Microsoft/GitHub. Cited for articulating the concept of "conviction" as the defining quality of a leader with genuine taste — distinguishing local-maximum decisions from global-maximum bets.

"I was having dinner last week with Jay Parikh, my boss. He used this word that I hadn't heard in a while: conviction. Someone, a leader with taste, has conviction. It's not just design conviction, not just business conviction, not just engineering conviction — it's that combination. When they have conviction, it is a bet aimed at the global maximum, not the local maximum." — John Maeda [00:43:36]

Steve Jobs

Apple co-founder. Referenced twice: once as the world's greatest editor who said no to more than yes, and once for his tactic of stress-testing employee conviction by feigning disagreement to see who would cave.

"Steve was one of the greatest editors in the world. He said no to so many things as opposed to saying yes to so many things." — Paul Bakaus [00:05:04]

"Very often somebody would walk into the room with Steve and Steve would just test them and say 'I think that's a stupid idea' even though he might think it's a really great idea — just to test them. Will they just cave and walk out of the room? Or will they actually double down?" — Paul Bakaus [00:45:17]

Hayao Miyazaki

Legendary Japanese filmmaker and animator. Used as the benchmark for ceiling-level creative output that AI-enabled creative tools have not yet approached.

"We haven't really seen the next Hayao Miyazaki from Runway or some other video startup." — Paul Bakaus [00:21:53]

Max (GitHub → Notion)

Former head of design at GitHub, now at Notion. Cited as part of a lineage of computational designer-engineers who came out of GitHub's design culture.

"There's Max at Notion who used to lead design at GitHub. There's all the Brassell people. They're all over the place." — John Maeda [00:46:11]

5. Operating Insights

Teach Design Vocabulary to Engineers Systematically — It Compounds

Teams that install Impeccable across both designers and engineers report better cross-functional communication — not just better AI output. Design vocabulary adoption is an organizational capability multiplier.

"I've had people walk up to me and say 'we're now using Impeccable as a default thing that we install on all of our engineers and designers' computers'... engineers are using design language more and designers use engineering language more because the tools, by using this careful vocabulary and approach, actually teach you what works and what doesn't." — Paul Bakaus [00:33:40]

To Sell a Design Decision Upward, Bring the Leader on the Emotional Journey First

Rational arguments for design instincts fail when leadership is in a business-outcome mindset. The tactic is to meet them where they are and shift the conversation toward shared vision.

"It's your job to present to them in a way that makes them dream the way you're dreaming about the future... you have to bring them along to where you want to go and get them on the same emotional journey. Once they see what you see, once they see the type of future that you want to create, then you're on the same page." — Paul Bakaus [00:42:09]

Default Settings Are Product Strategy — They Shape Industries

Whoever controls the default theme, font, or component in a widely-adopted framework shapes the aesthetic of an entire era of the web. This is a lever of outsized influence that product and platform teams routinely underweight.

"I colored the whole web orange overnight, and I learned my lesson. The power of defaults." — Paul Bakaus [00:18:34]

Stress-Test Conviction Before Betting on It

Borrowed from Jobs: before committing organizational resources behind an instinct or design decision, test whether the person holding the conviction will defend it under adversarial pressure — or fold.

"Steve would walk into the room and just test them and say 'I think that's a stupid idea' even though he might think it's a really great idea — just to test them. Will they just cave and walk out of the room? Or will they actually double down?" — Paul Bakaus [00:45:17]

6. Overlooked Insights

Evals Are the New Design Primitives — and Almost Nobody Has Noticed

In passing, John Maeda equates building model evals with the act of meta-design — constructing the latent space itself. This is not a metaphor; it means that whoever writes the evals for AI design systems is performing the highest-order design work. This reframes who should own eval development inside AI product teams: not ML engineers, but designer-engineers with the most developed aesthetic judgment.

"When we talk about evals, evals live in latent space construction as literal meta design." — John Maeda [00:21:23]

"We have the harness and the models coming together, we have a nice combo platter plus the evals. We have the feedback loop. That's actually really brand new. And that's why I'm so excited." — John Maeda [00:06:38]

The Quickdraw Lesson: API Shape Determines Category Winners, Not Features

Mentioned briefly and then dropped, the observation that Photoshop was Mac-exclusive not because of features but because of Quickdraw's region-based API primitives is a principle with direct application to anyone building developer platforms or agent tooling today. The team that designs the right primitive shape for agent-facing APIs will do what Quickdraw did — make their platform the only viable substrate for the breakout applications of the next decade.

"Photoshop happened because of the design of Quickdraw's API. It was impossible for there to be a Windows version of Photoshop because it didn't have the same API shape and DirectX... The Macintosh Quickdraw had this routine built upon this idea of regions — things like flood fill were super-duper optimized for speed. Great APIs are not just about shape but performance." — John Maeda [00:40:45]