The People Who Taught Us Structure
The structure AI work depends on now is the kind a small group of early web designers, writers, and information architects taught the rest of us. Product design started with no manual, so they built one piece by piece, in articles, books and talks, and gave it away. I learned much of my job from them, for free. This is my thank-you to them.
The first time I sat at a table with my heroes, I hadn’t earned the seat. It was an early Information Architecture Summit, and Karen McGrane had invited me to dinner. The dinner is a bit of a blur: I recall (but could be wrong) that Maria Giudice from Hot Studio was there, Chris Fahey from Behavior and Dan Brown from EightShapes. I knew most of the people at that table from their websites or books. Their work had been my on-the-job training for years before I met any of them, and most of what they taught me was structure: how to organize information, what to call things, and how to make a site hold together. Now somebody had pulled out a chair for me at the grown-ups’ table.
These days I spend a lot of my week building the structure AI design work runs on: specs that tell an agent what good looks like, loops that run the same check a hundred times, evals that catch it when quality slips, and rules that turn “that border looks off” into something a machine won’t let through again. Models want structure. An agent reads a web page through its accessibility tree, and a company’s AI finds answers by the headings in its documents. It turns out the people in this story taught me how to give them that. I came up through art, where nothing had a grid and anything was possible, and got into web design through its constraints: grids, code, and content models, most of it figured out by women and men a few years ahead of me and posted for anyone to read. Plenty of people still give away what they learn, and a lot of it is excellent. There are still dinners like that one, but they’re harder to find than they used to be, and some of them charge admission now. So I went back to see how the first ones got built. Rewind with me.
The web came with no manual. CERN put it in the public domain on April 30, 1993: use it, copy it, change it, pass it on. The unwritten message was simpler: see what you can do with this. The spotlight these days mostly falls on the people building AI models. Back then, a lot of the structure that shaped how people actually used the web came from designers, writers, and information architects. They pieced the manual together themselves, in magazines, small books, conference talks, and free templates, and gave it away. Here’s who built what, in order, as I remember it. It moves fast, so follow the links for the backstories.
1997–1999The polar bear book
The web was a few years old, and the dot-com money was starting to pour in, mostly to founders and investors. The people working out how to design for it were largely self-taught, and the job didn’t have a name yet. What they did have was a habit of writing down what they’d learned.
Jeffrey Zeldman and Brian Platz started a mailing list in 1997 for people who made websites and called it A List Apart. A year later it was a magazine, published on the web, about the web, written mostly by people who had taught themselves.
Nobody taught you how to make a website back then. There were design degrees and a few HCI programs, but no bootcamps. You figured something out, wrote it up on your own site, and if it was useful, people read it. Paul Ford started Ftrain. Jason Kottke started kottke.org. Peter Merholz took Jorn Barger’s “weblog” and shortened it to “blog,” and that same year Meg Hourihan’s company launched Blogger, so anyone could have one.
O’Reilly published Lou Rosenfeld and Peter Morville’s Information Architecture for the World Wide Web in February 1998. Everyone called it the polar bear book, and it made organizing a website something you could study. Karen McGrane joined Razorfish that year as its first information architect, back when the title needed explaining in every meeting.
I was in a loft in Boston, helping build one of the first 3D printers at Z Corp, an MIT spinout, and running a business reproducing frescoes on the side. Out of a Montessori-sponsored curiosity, I took apart a lot of computers. One day I saw my loft mates on CU-SeeMe, an early video chat tool, and that sent me down a rabbit hole of design blogs until I wanted to make a website of my own.
Today, when a company points AI at its own documents, it splits them by heading and section, so the hierarchy Lou and Peter’s book taught is the map the model reads.
2000–2004The Elements diagram
The money left, and the work got serious. With the hype gone, the people who stayed started asking what a website was for and who it was for, and that question turned into a profession: user experience and information architecture.
The bubble burst in the spring of 2000, and it seemed like everyone got laid off. Razorfish went from about $57 a share to under a dollar within a year and a half. I was living in Brooklyn at the time, and by 1 p.m. the bars were full of people living on severance and unemployment (aka the $405 plan), and the secondhand furniture stores were full of Aeron chairs. A small group kept working anyway.
That March, just as the market peaked, Jesse James Garrett posted a one-page diagram on his site. It stacked user experience into five planes, strategy on the bottom and the visual surface on top. The parts nobody sees, the structure, the sequence, the words, were design decisions too, and somebody had to own them.
Over the next few years, people started specializing in those layers, and each one got its own advocates making the case in public. Garrett co-founded Adaptive Path in 2001 with Peter Merholz, Jeffrey Veen, Indi Young, Janice Fraser, Lane Becker, and Mike Kuniavsky. Christina Wodtke launched Boxes and Arrows. Code had its own version of the fight. CSS had been an official standard since 1996, a way to keep a page’s structure separate from its look, but browsers supported it so badly that most sites were still held together with spacer images and layout hacks. In 2002 Douglas Bowman rebuilt Wired News in CSS to show it could be done, and Zeldman’s Designing With Web Standards, a year later, turned the argument into a book. Over at Stanford, David Kelley and a few colleagues founded the d.school in 2004 and taught design as a method anyone could learn.
I started my agency, EatMedia, in late 2004. Our first big client was SAP. We started in content strategy and IA, the work that diagram said somebody had to own, and grew into a full-service shop working with DAV, McKinsey, JinglePunks, and Laureate University.
Today, an AI agent driving a browser often reads the page through its accessibility tree, which is built from the semantic markup the standards crowd fought for.
Thank you, Jesse, for the diagram that told a lot of us what our job was, and where to focus.
2005–2009A free Photoshop file
The web started to feel like software, and then it moved into your pocket. Pages updated without reloading, anyone could publish, and by 2008 the iPhone and its App Store handed designers a new screen to figure out from scratch.
Garrett gave the technique behind that a name in 2005: Ajax. Zeldman and Eric Meyer ran the first An Event Apart in Philadelphia that December. Rosenfeld started a publishing company, Andy Budd, Richard Rutter, and Jeremy Keith opened Clearleft in Brighton, and Vitaly Friedman and Sven Lennartz started Smashing Magazine.
The money came back, too. Google paid $1.65 billion for YouTube, a company less than two years old with almost no revenue. That money went to the platforms. The designers got paid a few years later.
Then Apple opened the App Store in July 2008, and one person could sell software to millions. A month later, Teehan+Lax put out a Photoshop file with every iPhone interface element in it, free, and a big share of the early iPhone apps were mocked up with it. Nobody asked them to.
Jessica Hische drew an ornamental letter every weekday for Daily Drop Cap and let anyone use them. Dan Cederholm and Rich Thornett opened Dribbble. Jeffrey Veen’s Typekit made real typefaces work on the web. And the field started to formalize: Tim Brown’s “Design Thinking” ran in Harvard Business Review, and Liz Danzico chaired a new MFA in interaction design at SVA.
Content got its turn in the back half of the decade. Rachel Lovinger made the case for content strategy to the web crowd in Boxes and Arrows in 2007, and Kristina Halvorson’s “The Discipline of Content Strategy” in A List Apart, in December 2008, gave it the definition people still quote. In 2009 I was part of the Content Strategy Consortium at the IA Summit, a small group working out where content strategy fit into the design and development process. It was the first time I helped write part of the manual instead of just reading it.
Today, content broken into types and fields, the way content strategists taught, is content AI can find and reuse without guessing.
Thank you, Teehan+Lax, for the file half the industry built its first apps on.
2010–2012The web bends
The sharing peaked. The web had to work on every screen at once, and the community answered with short books, free frameworks, and talks that spread in days. It was also the last stretch when most of that knowledge lived on independent sites.
Ethan Marcotte published an article in A List Apart in May 2010 arguing that a website shouldn’t be built for one fixed screen. It should bend to fit whatever it lands on. He called it responsive web design, and within a few years it was just how the web got built.
A Book Apart launched that year: short, cheap, one idea each, meant to be marked up, not shelved. In 2011 Marcotte wrote one on responsive design and Luke Wroblewski wrote Mobile First. That same year, two engineers at Twitter, Mark Otto and Jacob Thornton, released Bootstrap. Nobody said “design system” yet, but that’s what it was.
The community taught the business side in public, too. Jessica Hische posted a flowchart called Should I Work for Free? Mike Monteiro got up at CreativeMornings in San Francisco and gave a talk called “F*ck You, Pay Me.”
At EatMedia, the work was changing. More of our contracts were fixes to things someone else had built, budgets were tighter, and everyone wanted it faster, even though good work still took the time it took. The community felt different too, a little more competitive, as more people were trying to make a living from the same work and the same attention.
Today, when a tool generates an interface, it assembles components, parts like the ones Bootstrap gave away.
Thank you, Ethan, for the web that fits in our hands.
Thank you, Jessica and Mike, for teaching a generation to send the contract first.
2013–2016Reader goes dark
The independents got absorbed. Platforms took over the reading, and big companies started buying the studios and hiring the people who had taught everyone else. Design moved in-house.
Google shut down Reader on July 1, 2013. After Reader, the reading moved to Twitter and a lot of the writing moved to Medium. The blogs were still there, but fewer people showed up. Anil Dash had seen it coming a few months earlier in “The Web We Lost,” his list of everything the open web was giving up to the platforms.
Design had become a differentiator, and the big companies went shopping. Facebook hired all of Maria Giudice’s Hot Studio. Google hired Wroblewski and his Polar team. Capital One bought Adaptive Path. Teehan+Lax closed and its partners went to Facebook. At IBM, Phil Gilbert hired his first 60 designers in 2013 and kept going into the thousands. After fifteen years of giving their best thinking away, people got paid for it, and paid well. The trade was that more of the thinking happened inside one company, for that company.
I was starting to share the stage with some of my heroes, at Lean UX and SXSW. Looking back, I think we all got a little too excited that design finally had a seat at the table. To keep it, a lot of us learned to speak the language of the business, and that part was right. What we let slip was the balance, advocating for the user as hard as we argued for the business. The seat was for both.
Thank you, Adaptive Path, for thirteen years of showing companies what UX was.
2017–2020Design gets a number
The argument that design belonged in the business was mostly won, and it came with maturity models, design systems, and design ops. What had been passed between people became infrastructure inside companies.
In 2018, McKinsey published “The Business Value of Design,” tying design to company performance. InVision surveyed more than 2,200 companies the next year and sorted them into five levels of design maturity. Five percent made the top.
Sketch gave up ground to Figma, which put the whole team in one file, and the component became the unit everybody organized around. That idea had been given away in pieces for over a decade: Yahoo’s pattern library in 2006, Bootstrap’s code, Brad Frost’s Atomic Design vocabulary, Salesforce’s Lightning Design System, Jina Anne’s Clarity conference. By 2020 most large companies had a design system, and plenty published theirs.
Not everyone was impressed. Pentagram’s Natasha Jen gave a talk called “Design Thinking Is Bullsh*t,” arguing that sticky notes had replaced actual design work, and it went viral. Capital One retired the Adaptive Path name in 2019, and UX Week went with it.
Around this time I got a seat of my own: I worked at IBM during the Phil Gilbert era, and later landed my dream job at Mailchimp. Designers felt empowered. But if I’m honest, you could sense the business’s patience wearing thin with Post-its and empathy maps. And the business wanted all of it faster.
Today, AI design tools read a system’s tokens, the named colors, spacing, and type, to stay on brand.
Thank you, Phil, for hiring designers by the thousand when most companies were still hiring them one at a time.
2021–2024A new reader arrives
The institutions that held the community together started closing. Conferences, magazines, and publishers that had run for well over a decade wound down, and generative AI arrived in the middle of it.
ChatGPT came out on November 30, 2022. Two weeks later, An Event Apart held what turned out to be its last conference, in San Francisco; Zeldman later wrote that COVID had done it in. InVision, the company that had graded everyone’s design maturity, announced in January 2024 that it would shut down. That March, A Book Apart stopped publishing and began handing its authors their rights back. In December, Zeldman wrote about the endings, and mentioned that he and his partner had earned nothing from An Event Apart in its last five years and closed it owing six figures.
Meta called 2023 its “year of efficiency,” and flatter and faster became the goal almost everywhere. IDEO, the firm that had carried design thinking into the business press, cut about a third of its staff that year, and Fast Company called it the end of the design-thinking era. Servant leadership and design thinking, both everywhere a few years earlier, mostly dropped out of how product teams talked. Most of the design leadership jobs I saw were at Series C or D companies whose last raise had been a down round, or at companies private equity had just bought. About one in five venture rounds in 2023 was a down round, the highest share in years, and private equity took a wave of software companies private.
Inside companies, design got quieter. The design systems the community had given away were mostly run by engineering now. The CMS came with templates, and product managers owned the roadmap. A lot of what designers had fought for was built into the tools as defaults. Design was still in the room, but it was less often the voice asking what the thing was for. The magazines, books, and conferences that had made that case in public were closing, and nothing took their place.
Thank you, Jeffrey, for keeping the conference and the books going long after they stopped paying.
2025–nowVibe coding
Making things got cheap. AI can produce a screen, a component, or a working prototype in minutes. Knowing what’s worth making is the hard part now.
Andrej Karpathy coined “vibe coding” in February 2025 for building software by describing it to an AI. Claude Code and tools like it followed.
There’s more free design knowledge now than at any point above. Figma’s community is full of free files and plugins. IBM, Google, and the UK government publish their design systems in the open. People working with AI post their prompts, workflows, and agent setups within days of figuring them out.
What’s different is where it lives. It’s spread across thousands of feeds instead of a few places everyone read. That opened it up. The old network was small, and its heroes were mostly people who could afford to write for free for years. Now more people get to share, but fewer get found. Small rooms are easier to set up than they’ve ever been: a group chat, a meetup, a shared repo. Someone just has to start one.
AI-native design is still early. There are guidebooks and pattern libraries, like Google’s People + AI Guidebook and Emily Campbell’s Shape of AI, but no agreed practice and no credential anyone trusts yet. It’s being figured out in public by whoever bothers to write it down, which is where the web was when I found it. Some of the same people are at it again. Paul Ford calls himself a “fun Cassandra” about what’s coming. Wroblewski is bringing the same patience to agent interfaces that he brought to web forms.
Hiring is starting to catch up. When everyone has the same tools to research and produce, what separates the work is taste, craft, and the design lens that decides what’s worth making. I’m seeing more postings for creative directors and brand leads than I have in years. And companies that cut people for AI are reconsidering: Forrester’s Predictions 2026 report found that 55% of employers regret layoffs made because of AI. Not every job comes back the way it left. Nielsen Norman Group’s State of UX 2026 found the market steadying, with senior practitioners coming back first, and argued that surface-level design won’t be enough anymore.
So my colleague David and I are setting out a few chairs of our own, a meetup in Chapel Hill for people working with AI across disciplines. It’s just getting started.
Why it matters now
AI gives better results when what it reads is well structured, and the people in this story worked out much of that structure years ago.
Stack it up the way Garrett stacked his planes: the hierarchy, the markup, the content models, the components, the tokens. That’s much of what AI reads today.
None of them were building for machines. They were building for people, and it turns out machines need a lot of the same things.
The web grew up because people wrote the manual as they went. AI moves faster, and most of the teams I work with are building before anyone has written theirs. That’s the work I care about now: the decisions, the patterns, the shared sense of what good looks like, written down where people and agents can both find it.
Where are they now
This is my thank-you to everyone in this story, and a few more: where they were, what they gave us, and where to find them now.
In this piece
Then Razorfish, Autogram
Karen McGrane and Jeff Eaton
Gave us: Content modeling for content that goes anywhere
Now Contentful
Then Hot Studio
Gave us: Rise of the DEO, and a studio that grew designers
Now Coaching creative leaders at a relaunched Hot Studio
Then Behavior Design
Gave us: graphpaper.com, and a high bar
Now 1stDibs
Then EightShapes
Gave us: Communicating Design
Now Curious Squid
Then A List Apart, Happy Cog
Gave us: Web standards, and a place to publish them
Now Automattic
Then Ftrain, Postlight
Gave us: A whole issue of Businessweek on “What Is Code?”
Now Aboard
Then kottke.org
Gave us: A blog, every day since 1998
Now Still posting
Then Adaptive Path
Gave us: The word “blog,” and Org Design for Design Orgs
Now Consulting, and Finding Our Way
Then The polar bear book
Gave us: Information architecture you could learn from a book
Now Rosenfeld Media
Then The polar bear book
Gave us: The same book, and findability as a discipline
Now Semantic Studios
Then Adaptive Path
Gave us: The Elements diagram, and the name Ajax
Now Finding Our Way
Then Adaptive Path, Typekit
Gave us: Real typefaces on the web
Now True Ventures
Then Adaptive Path
Gave us: Mental Models
Now Thinking Styles
Then Boxes and Arrows
Gave us: A magazine for the people doing the work
Now Stanford
Then An Event Apart
Gave us: The CSS books a generation learned from
Now Igalia
Then Clearleft
Gave us: The UK’s first UX agency
Now Seedcamp
Then Teehan+Lax
Geoff Teehan and Jon Lax
Gave us: The free iPhone interface file
Now Lightspark (Teehan)
Then Daily Drop Cap
Gave us: Should I Work for Free?
Now Lettering, and children’s books
Then SimpleBits, Dribbble
Gave us: A place to post the work in progress
Now SimpleBits
Then SVA
Gave us: A graduate program for interaction design
Now Microsoft AI
Then Razorfish
Gave us: Content modeling, taught in public
Now Publicis Sapient
Then Brain Traffic
Gave us: Content strategy as a practice
Now Brain Traffic, and Button
Then A List Apart
Gave us: Responsive web design
Now University of Washington
Then Polar
Gave us: Mobile First
Now Sutter Hill Ventures
Then Twitter
Gave us: Bootstrap, with Jacob Thornton
Now The Pierre Computer Company
Then Mule Design
Gave us: “F*ck You, Pay Me”
Now Mule Design
Then Six Apart
Gave us: Movable Type, and blogging for everyone
Now antitech
Then IBM Design
Gave us: Design thinking at the scale of a 400,000-person company
Now *Irresistible Change*, and speaking
Then Atomic Design
Gave us: A vocabulary for components
Now Design systems for agents
Then Salesforce
Gave us: Design tokens, and the Clarity conference
Now Microsoft
Then A List Apart, A Book Apart
Gave us: The look of the small books
Now Loud Room
Then Predicate
Gave us: “The day two problem”
Now Bucket Studio
Then Fresh Tilled Soil
Gave us: Design Leadership, and honest advice on running an agency
Now Coaching leaders at Second Harvest
Other vanguards
Then Revision Path
Gave us: Eleven years of interviews with Black designers
Now Eames Institute Curious 100
Then LeanUX NYC, AgileUX NYC
Gave us: The conferences where Lean UX got worked out in public
Now Strategy and foresight
Then 37signals
Gave us: Getting Real, free online
Now 37signals
Then Lean UX
Gave us: Lean UX
Now Sense & Respond Learning
Then gotomedia
Gave us: Web ReDesign
Now gotomedia
Then Mule Design
Gave us: Just Enough Research
Now Mule Design
Then A List Apart
Gave us: The Elements of Content Strategy
Now wreckage/salvage
Then Advanced Common Sense
Gave us: Don’t Make Me Think
Now Still teaching
Then useit.com
Gave us: Alertbox, and the ten usability heuristics
Now UX Tigers
Then Apple, NN/g
Gave us: The Design of Everyday Things
Now The Don Norman Design Award
Then Plain language
Gave us: Letting Go of the Words
Now Semi-retired
In memory
Then Web Standards Project
Molly Holzschlag
Gave us: More than 30 books, and a web built on standards
Now Died 2023
Thank you
These are the people who made room for me.
- Karen McGrane, for that dinner, and for introducing me to half the people in this piece.
- Kristina Halvorson, who brought me into a small group while content strategy was still being defined.
- Jeff MacIntyre, whose depth, delivery, and all-around Canadian-ness made me up my game.
- Andy Budd, who always answered my questions and emails, and gave sound advice.
- Richard Banfield, for his advice on running an agency, and on being a good human.
- Rachel Lovinger, who brought me into Razorfish and is still a great sounding board.
- Peter Merholz, for Org Design for Design Orgs and for honest advice.
- Chris Fahey, who set a high bar early for what good thinking and good design looked like.
And to the people doing it now: the mentors who take the free calls, the teams who open-source their systems, the designers posting what they learned about AI last week. You’re in the same line.
Nobody has written the polar bear book for AI-native work yet, or drawn its five planes. There’s no A List Apart for it. I got an invite. If you’re starting now, don’t wait for yours. Write down what you figured out this week. Answer the question when someone asks. And when you find your people, make room for the next one.
Highlight any sentence and three thinkers weigh in.
Ian Alexander
Design leader & design engineer — writing on leadership, AI product strategy, and building teams that ship.