Dylan Field tells Sourcery that design is far more than visuals, pushing back on the claim that design is dead as engineers, PMs, and designers each argue they are not needed.
Field argues that AI-generated design defaults to the average, and points to Figma Weave, which lets users steer model outputs, as the path to bolder work.
Field says brand trust comes from consistency over time and cites Elon Musk's view that a special product needs no logo, using Tesla as the example.
Figma's Dylan Field Says AI Raises the Value of Taste
Figma's Dylan Field argues that AI does not kill design, because it widens who can build while leaving taste and judgment as the scarce skills. Left to defaults, AI gives the average, and design is largely non-verifiable, so humans must steer. That makes bolder bets, better interfaces than prompting and trust built over time the real work ahead.
Takeaways
Design is not dead; AI widens who can build. Field says engineers, product managers and designers each tell the others they are unneeded. He sees a temporary phase of excitement before limits show. Design covers user experience and understanding users and culture, and engineers building fast now want to make things great.
Default AI output is average, so taste matters. People who ask AI to design everything get the average result. Figma Weave passes model outputs through a workflow so users can steer them. Field predicts bolder bets will unlock an explosion of creativity, and he sees early signs people are internalizing this.
Prompting is a weak interface for AI. Field is still shocked that AI interaction is prompt-based, since other ways to explore latent space exist. He urges the industry to revisit 1960s to 1980s computer interaction work and to explore touch and VR, noting pull-to-refresh is probably the last big paradigm.
Judgment cannot be fixed by adding model IQ. AI suits verifiable domains like pure math, which Field uses it for. Design is largely non-verifiable and needs a human judge. Judgment does not simply rise with IQ points, so humans in the loop steer AI to the best outcomes.
Trust comes from consistency and shared value. Field says brand trust is built through consistency over time. He cites Tim O'Reilly's advice to create more value than you capture, and Microsoft's aim for the Windows economy to exceed Windows revenue. Doomerism about AI, he argues, has to stop.
Agents are unclear and safety is understudied. Field says the unit of an agent is unclear, which complicates whether agents replace employees. He calls jailbreaking understudied and suggests incentive programs like HackerOne's, since long conversations with models can reach states that amount to jailbreaks.
Software value rests on network effects and data. Field says network effects, customer distribution and data liquidity make software valuable to customers. He is bullish on SpaceX for its missionary talent, though he cannot read its AI and cloud business. He agrees a special product, like a Tesla, needs no logo.
Field's early path
His first professional internship was at O'Reilly Media, where he took in Tim O'Reilly's advice on creating value.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
At 20 he interned at Flipboard, working alongside colleagues in their 30s.
In 2012 he joined the Thiel Fellowship, a two-year program with $100,000 of funding.
He says the fellowship's community mattered more than the money.
He credits Shao, Figma's first manager hire, with teaching him management.