The episode opens with a polished OneTrust ad that frames the central tension of the AI era: innovation demands speed, but risk demands caution. OneTrust positions itself as the bridge — an AI-ready governance platform that promises organizations they can move fast without moving recklessly. The Hard Fork Live 2026 sponsor slate is then read, naming IBM as the premier sponsor and Everpure, Pure Leaf, the University of Notre Dame, and Atlassian as associate and supporting sponsors.

Kevin Roose and Casey Newton kick off the episode in buoyant post-show mode, recapping what listeners are about to hear. The headline guest is Dylan Field, CEO of Figma, whose conversation they describe as fascinating. Dan Powell — NYT composer, Hard Fork theme-song creator, and the show's DJ for the San Francisco event — is teased as a 'legendary' performer. The hosts also look ahead to Friday's Hard Fork Live episode, which will include a conversation with Dwarkesh Patel and a live audience Q&A. Casey's deadpan question about whether the crowd questions will be screened in advance is met with a flat 'No' from Kevin, setting a playful tone.

The pre-interview ad block returns with a second OneTrust read emphasizing that moving fast without governance is a liability rather than a strategy. Framer follows with a pitch aimed at design and product teams who want AI-assisted website building that stays editable and on-brand. Rippling closes the block with a workforce-data angle — its AI is pitched as actionable rather than merely analytical, because it draws on live global HR data. The Framer offer of 30% off a Pro annual plan at framer.com/hardfork is the only listener-facing promo code in this segment.

Kevin Roose pulls off a perfect opening gambit: reading a Facebook message the teenage Dylan Field sent him in 2009 after picking up his book at the Brown library. The message is earnest and slightly nerdy — 'maybe I'll meet you in Providence someday' — and Roose spends a beat apologizing for never responding. Field takes it in good humor, revealing a charming follow-up: they did actually meet, because Roose came back to hang out with the a cappella group that Field had joined as a freshman. It's a warmly personal opening that grounds the CEO of a public company as somebody who has always been curious, socially reaching, and just a little bit of a fanboy — qualities that will thread through the entire conversation.

Asked whether he has 'AI psychosis,' Field laughs and says it's better to front-run it than let it sneak up on you. His current obsession is what he calls 'vibemath-ing' — using AI to explore mathematical problems, not to prove theorems, but to study how AI attacks verifiable problems. ^{[1] — Dylan Field "Dylan Field is using AI to tackle math problems — what he calls 'vibemath-ing' — because math is one of the few domains where AI correctnes…" 06:22 Play} He contrasts this with design evaluation, where two people can look at the same thing and reach opposite conclusions; in math, the answer is right or it isn't. This love of exploration has a longer backstory: Field says he was into crypto collectibles (now NFTs) early on, and before that, WebGL, the browser graphics technology that eventually became the technical foundation for Figma. The pattern, he argues, is consistent — go deep on something before you know where it leads, and it pays off in ways you can't anticipate.

Kevin Roose offers a half-serious theory: every startup CEO in Silicon Valley is spending weekends vibe coding and arriving Monday morning asking why it took 50 engineers to build what they just made in two days. His read is that it's nostalgia — a return to the fun part of the job. ^{[1] — Dylan Field "Kevin Roose's theory: CEOs are obsessed with vibe coding because it reminds them of when their jobs were fun. Dylan Field's take: people lo…" 08:15 Play} Dylan Field agrees but pushes the idea further. It's not just nostalgia; people fundamentally love making things. They like seeing ideas exist in the world in a tangible form. AI is just lowering the barrier so that creative energy can flow more freely, and he predicts the vibe-coding wave will extend well beyond the CEO demographic into designers, writers, and anyone who has ideas to express.

Casey Newton prompts Field on Figma's 'design is not dead' ad campaign and asks him to make the case. Field's answer is layered and confident. He begins with writing: having a genuine voice, a distinctive style, even being funny — these are things AI can't replicate because they're not in the statistical average the model produces. Then he pulls out data he's seen on the App Store: the total number of apps has exploded with AI, but the number of apps actually getting frequent traffic is exactly the same. ^{[1] — Dylan Field "Dylan Field pushes back on the 'design is dead' narrative. When AI generates the average, having genuine creative voice and pushing beyond …" 09:15 Play} More supply, flat demand — it's a hyper-competitive environment where standing out isn't optional. The discipline, he argues, is in not settling for the first draft, the first output, pushing the AI-generated starting point further and further until it becomes something genuinely new. ^{[2] — Dylan Field "If you have a creative voice in writing or design, you put yourself out there and you take a risk — this is a good time to do that. It's so…" 11:00 Play} If you do that in writing or design, this is one of the best times in history to put yourself out there.

Kevin Roose notes dryly that San Francisco — a city that made Allbirds a thing — has suddenly discovered 'taste' as the defining human quality that AI can't replicate. He then presses Field on the cynical reading: isn't 'taste' just the word we give to whatever the models aren't good at yet, a goalpost that keeps moving as models improve? ^{[1] — Dylan Field "Everyone in San Francisco is talking about 'taste' as the human skill AI can't replicate. Dylan Field's response is sharper: you don't even…" 12:50 Play} Field sidesteps the definitional trap neatly. He says you don't even need to argue about whether humans have taste; all that matters is that people can detect the AI average. Once you accept that the average is now AI-generated, anyone who pushes beyond it — in any direction — automatically stands out. His broader point lands with a tidy elegance: the models being trained on distribution data is the very thing that makes going off-distribution so valuable.

Casey Newton opens this chapter with a compelling historical analogy: the invention of the camera pushed painters toward Impressionism because photography had made photorealism redundant. Does AI have an analog in the design world? Field finds the question interesting but admits the expected reaction — a surge in sculpture, textural, non-digital art — hasn't fully materialized yet. ^{[1] — Dylan Field "Dylan Field says the last 15 years of design has been a monoculture rut — and partly blames the dominance of tools like Figma. But AI is no…" 14:55 Play} He sees early signs of it in advertising, where brands are now racing to prove their content is authentically human-made. In design more broadly, he predicts a coming wave of interactivity and dynamic interfaces — software used more like a creative medium. And then he drops a candid confession: the past 15 years of design has been a monoculture rut, and he acknowledges, with a wry smile, that Figma might bear some of the blame. The crowd laughs. He doesn't fully disagree.

The conversation takes a personal turn when Casey Newton brings up something that had been bothering him all day: he read a Substack post on Techmeme and immediately recognized it as Claude-generated. For someone who has spent years trying to get his own work surfaced on tech aggregators, seeing AI output get the same placement was demoralizing. ^{[1] — Casey Newton "Casey Newton read a Substack post on Techmeme today that was clearly generated by Claude — and it made him furious. His profession is start…" 17:25 Play} His profession, he says, is starting to look more and more like slop. He turns the question on Field: are designers in the audience having a similar feeling? Do they look at the websites and apps being put out and just know — immediately — that a button generated it? It's a raw, honest moment in an otherwise buoyant conversation, and Field's response (pointing to imposter syndrome as a shared trait) is characteristically generous.

Field responds to the AI-jobs-doom narrative with something concrete: the companies he regularly speaks with are telling him that design is one of the most prioritized places they're hiring. Not cutting — hiring. He finds the broader contradiction somewhat baffling: people spend the morning predicting AI will replace all jobs and the afternoon calling up great engineers to recruit them. Something doesn't add up. ^{[1] — Dylan Field "Despite AI anxiety, Dylan Field predicts there will be significantly more people with the title 'designer' in two years than today. Enginee…" 18:18 Play} His own prediction is contrarian and specific: in two years, there will be significantly more people with the job title 'designer' than today. A lot of that growth will come from people in adjacent roles — engineers who start building fast with AI and then realize they need to know what makes something good. The act of caring about that question, he says, is design.

Kevin Roose shifts gears with barely concealed glee, bringing up the Mike Krieger story: Figma board member, Instagram co-founder, Anthropic product lead — suddenly gone from the board, and days later Anthropic launches Claude Design, a product that sits squarely in Figma's territory. ^{[1] — Dylan Field "Figma board member and Anthropic product lead Mike Krieger resigned — and days later Anthropic launched Claude Design, a direct Figma compe…" 20:40 Play} Field handles it with careful diplomacy. He says Krieger is someone he genuinely cares about. He doesn't confirm or deny any causal link. But when Roose asks whether he'd put another AI lab executive on the board in the future, Field's answer is a masterclass in saying something without saying it: 'It probably depends on what their ambitions are.' The audience laughs because everyone understands exactly what he means.

Kevin Roose broadens the conversation to a more systemic worry: are AI labs becoming amorphous blobs that expand into every adjacent industry and crush smaller startups in their path? Field tells it as a tale of two labs. ^{[1] — Dylan Field "Dylan Field sees a tale of two labs: OpenAI launched aggressively — social networks, Sora, more — then made the hard call to cut and focus.…" 22:20 Play} OpenAI went through an expansionary phase — social networks, Sora, a lot of stuff — and then made the genuinely hard call to shut things down and refocus. Anthropic, he suggests, is currently in its own expansionary arc, launching a lot of products and testing what sticks. The more interesting question, he says, is what they'll still be doing in a year or two. When Roose asks what things they might attempt and fail at, Field pauses — and then deadpans: 'Safety? Joking.' The crowd erupts. Casey calls it the best answer of Hard Fork Live 2026.

Someone on Field's team had flagged 'hyperstition' as something he wanted to discuss, and Kevin Roose obliges. Field acknowledges the term comes from an internet edgelord ('don't look him up') but defends the core idea: there is a real phenomenon where ideas summon their own existence through accumulated belief and attention. ^{[1] — Dylan Field "Dylan Field introduces hyperstition: the phenomenon where ideas gain power purely by attracting belief. Bitcoin had every reason not to wor…" 24:15 Play} Bitcoin is his cleanest example — there was every reason for it to fail, no obvious reason for it to work, and yet the more attention it got, the more real it became. AI is the same story told at greater scale. A group of safety-minded researchers deliberately structured their organizations to avoid a race dynamic. They formed nonprofits, built complex governance structures, tried to keep everyone together. And then incentives kicked in, people split apart, and — as Field delivers with a perfect shrug — 'hello, race dynamic.' ^{[2] — Dylan Field "We gotta make sure we don't create a race dynamic, for example. We gotta make sure that we have these complex corporate structures that we …" 25:53 Play} It's a witty and insightful argument: the very act of trying to prevent something can sometimes be the thing that summons it.

Field lands on a surprisingly hopeful note for a conversation that has ranged across race dynamics and AI slop. He points out that AI is acutely aware of all the cautionary tales, the science fiction doom scenarios, the Google engineer who claimed it was conscious — because all of it is in the training set. What's underrepresented, he says, is stories about AI going well for humanity. ^{[1] — Dylan Field "AI is painfully aware of all the doom narratives in its training data. Dylan Field points out there are far fewer stories in the dataset ab…" 26:35 Play} And in the logic of hyperstition, that matters: the stories we tell shape what gets trained, which shapes what gets built. He says flatly that he believes in an optimistic future for humanity — and then catches himself, noting that optimism is somehow now the hot take. Casey jokes that the audience's homework for the evening is to go home and write some good AI endings. Field agrees. It's a funny, earnest close to a genuinely wide-ranging conversation.

After 40-plus minutes of conversation covering everything from a 2009 Facebook message to hyperstition, the hosts bring the Dylan Field interview to a warm close. Kevin and Casey both thank Field, who thanks the audience and bids everyone a good night. The word 'hyperstition' hangs briefly in the air — someone in the studio echoes it back — before Kevin teases more Hard Fork Live content to come.

After 40-plus minutes of conversation covering everything from a 2009 Facebook message to hyperstition, the hosts bring the Dylan Field interview to a warm close. Kevin and Casey both thank Field, who thanks the audience and bids everyone a good night. The word 'hyperstition' hangs briefly in the air — someone in the studio echoes it back — before Kevin teases more Hard Fork Live content to come.

After the robot choir performance, the hosts close warmly, with Kevin explaining the technical setup — Teenage Engineering robots, MIDI, Bluetooth — while Casey jokes that the robots enjoy being corrected. The credits that follow are unusually long and affectionate, reflecting the scale of a live event production: producers Whitney Jones and Rachel Cohn, editor Viren Pavich, fact-checker Caitlin Love, engineer Alyssa Moxley, a full video team, and a sizable New York Times Live Events crew. Special thanks go to the Yerba Buena Center for the Arts and Blue Shield of California Theatre, the event's San Francisco venue. Listeners are invited to email [email protected] with their photos from the event.

The closing ad block mirrors the opening: OneTrust returns with its governance-that-helps-you-go message, Comcast Business runs a longer brand-building spot emphasizing that it powers over 90% of the Fortune 500 and millions of small businesses across 100-plus countries, and KPMG closes with its client-zero AI transformation pitch. The symmetry of the ad placement neatly brackets an episode that was itself about the tension between moving fast and moving well in the AI era.