All podcasts / Lenny's Podcast / Summary

The design process is dead. Here's what's replacing it. | Jenny Wen (head of design at Claude)

2026-03-01 - 77 min - source - Read full transcript
Lenny Rachitsky (host)Jenny Wen

Key insights

The traditional discovery-diverge-converge design process is now largely dead because designers no longer have time to make polished mocks before engineers ship a working version.
Jenny says a talk she gave on this just months earlier already feels outdated given how fast engineers can now spin up prototypes with multiple AI coding agents running in parallel. Where mocking and prototyping used to take 60-70% of a designer's time a few years ago, it is now closer to 30-40%, with the rest split between pairing directly with engineers and hands-on implementation.
design-process-transformation
Design work has split into two distinct modes: supporting fast engineering execution day to day, and creating short-range (three to six month) directional prototypes instead of multi-year vision decks.
Because AI model capability shifts too fast to plan two, five, or ten years out, the 'vision' work that used to produce beautifully story-told decks now more often takes the form of a rough prototype that points teams in a shared direction, which still matters so that many people spinning up features independently converge on something coherent rather than efficient chaos.
design-process-transformation
Figma remains valuable specifically for divergent exploration and fine-grained visual/interaction detail, because coding tools are inherently linear and lock you into one direction too early.
Jenny still uses Figma daily despite being 'a fully clod house' otherwise. She argues the best design comes from throwing 8-10 different directions at the wall, which coding tools discourage since you get invested in one path as soon as you start iterating with an AI agent on it; Figma's canvas is where that breadth still happens.
design-process-transformation
AI's taste and judgment will keep improving, but accountability for what actually gets built stays with humans, because many hard product decisions are really disputes between people that AI can inform but not resolve.
Jenny pushes back on the idea that designers will always be needed for 'good taste,' agreeing AI will get better at it, but argues the harder part of building software is the human disagreement over what should or shouldn't go into a feature - AI can present options and data, but someone still has to own the decision and be accountable if it's wrong, similar to how a radiologist remains the accountable party even once AI reads the scan.
ai-and-future-of-design-role
Chatbots and terminals are not going away, but an increasing share of UI will be dynamically generated by models rather than hand-coded, producing a hybrid interaction model.
Jenny points to Claude's interactive widgets (weather, stocks, elicited multiple-choice questions) as an early example of tactile UI layered on top of chat, arguing that chat's real value is the infinite flexibility it opened up, while purpose-built UI still wins for specific, well-understood tasks - and that UI will increasingly be model-generated per situation rather than pre-built.
ai-and-future-of-design-role
Evan Tana's legibility framework - a founder x idea 2x2 of legible/illegible - is Jenny's method for spotting undervalued internal prototypes and ideas.
When both the founder/creator and the idea are highly legible, the idea is probably not novel since it's easy to understand and others will build it too. The valuable zone is illegible ideas that generate real internal energy despite being hard to explain; Jenny cites an internal prototype called 'Cloud Studio' that she initially didn't understand but sensed excitement around, which ultimately fed into both Anthropic's skills framework and the design of Claude Cowork.
ai-and-future-of-design-role
When hiring designers now, Jenny targets three specific archetypes rather than generic experience: strong 'block-shaped' generalists, deep T-shaped specialists, and 'cracked new grads.'
Block-shaped generalists are 80th-percentile across several core skills rather than merely competent at many things, making them rare but able to flex as the design role stretches toward PM and engineering shape. Deep T-shaped specialists are top-10%-in-industry in one area, often bordering on software engineering. 'Cracked new grads' are early-career people who build constantly, are unburdened by legacy process, and learn fast - a profile Jenny thinks most companies overlook in favor of hiring only senior, experienced talent.
talent-and-hiring
Rotating out of management back into an individual-contributor role can rebuild the empathy and hard skills a design leader needs during a period of fast process change.
Jenny spent this past year as an IC at Anthropic (after briefly managing a team), and says it gave her skills and firsthand understanding of how much the design process has changed that she would not have picked up purely managing. She compares it to engineering orgs that rotate new EMs through hands-on work before they manage full time, and argues design should adopt the same practice.
talent-and-hiring
Trust in an early or rough product launch comes from visibly, continuously shipping fixes based on user feedback, not from launching something polished.
Anthropic labeled Claude Cowork a 'research preview' on launch, explicitly signaling it was the worst version it would ever be. Jenny says what actually destroys brand trust is releasing something rough and then going silent afterward; committing publicly to rapid iteration - and following through - is what she calls 'building trust through speed.'
design-craft-and-quality
Craft and quality oversight doesn't disappear when shipping velocity increases; it shifts toward managing where a feature sits in its adoption lifecycle rather than reviewing every release.
Jenny distinguishes early 'research preview' releases, where flaws are expected and caveated, from mature features that get the traditional level of design polish, arguing this staged approach lets one designer's attention scale across many more shipped features than a uniform review process would.
design-craft-and-quality
The most effective 'low-leverage' tasks for senior leaders are often actually high-leverage, because it being that specific person doing the work is the signal.
Jenny rejects the common manager advice to delegate away all 'low-leverage' work. She points to senior leaders who personally dogfood the product, file detailed bug reports, or (citing Mike Krieger) submit their own PRs, arguing this builds product familiarity and visibly signals that no task is beneath the leader - the same logic applies to something as small as personally making a team member's work-anniversary card instead of delegating it.
management-philosophy
Letting a team comfortably 'roast' its manager is a healthy signal of psychological safety, but only works paired with unmistakably high standards.
Jenny describes her past team affectionately mimicking her verbal tics in critique sessions as evidence they trust her enough not to fear her. She frames the balance as being like a 'tough parent': team members know she won't fire them on a whim, but also know she has high standards and is directly invested in their best work, echoing the 'radical candor' idea of caring personally while challenging directly.
management-philosophy

Books referenced

Media referenced

Companies

Techniques and frameworks

Summary

Jenny Wen, who leads IC design work on Claude Cowork at Anthropic and previously ran design for Claude and was Director of Design at Figma (FigJam, Slides), joins Lenny to argue that the design process most designers were trained on is effectively dead. The traditional discovery-diverge-converge cycle, which Jenny herself argued against in a talk just months earlier that she says already feels outdated, has been forced to change because engineering changed first: with AI coding agents letting engineers spin up working prototypes faster than designers can produce polished mocks, the mocking-and-prototyping share of Jenny's time has fallen from roughly 60-70% a few years ago to 30-40% today, with the rest split between pairing directly with engineers and hands-on code implementation.

That shift has stratified design work into two modes: supporting fast day-to-day engineering execution, and creating short, three-to-six-month directional prototypes rather than multi-year vision decks, since AI model capability moves too fast to plan far out. Figma still earns its place in this world specifically for divergent exploration and fine visual detail, because coding tools are linear and lock you into one direction as soon as you start iterating with an agent - Figma's canvas is still where Jenny throws 8-10 different directions at the wall before committing.

On where human judgment remains valuable, Jenny agrees AI's sense of taste will keep improving but argues accountability for what actually ships stays with people, since the hardest disputes in building software are between humans (what belongs in a feature, what doesn't) rather than something AI resolves by presenting options. She uses Evan Tana's "legibility framework" - a founder x idea 2x2 - as her personal method for finding undervalued opportunities inside Anthropic: illegible ideas that nonetheless generate real internal energy, like the "Cloud Studio" prototype she initially didn't understand but that ultimately fed both Anthropic's skills framework and Claude Cowork's design.

On hiring, Jenny looks for three specific archetypes rather than generic seniority: "block-shaped" strong generalists (80th-percentile across several skills), deep T-shaped specialists (top 10% in one area, often bordering on engineering), and "cracked new grads" who build constantly and are unburdened by legacy process - a group she thinks most companies overlook. She credits her own decision to rotate from management back into an IC role this past year with giving her skills and empathy for how much the process has changed that she would not have gained purely managing, and argues design should copy engineering's practice of rotating new managers through hands-on work first.

The conversation closes on management philosophy: Jenny explains how Claude Cowork's "research preview" label and Anthropic's public commitment to rapid, visible iteration builds what she calls "trust through speed" rather than waiting to launch something polished, and she reframes conventional advice to delegate away "low-leverage" work, arguing that senior leaders personally dogfooding, filing bugs, or fixing small things is often exactly what builds outsized trust because it's them doing it. She pairs a culture where her team comfortably "roasts" her with deliberately high standards, describing the balance as being like a "tough parent" - safety and warmth alongside direct, demanding feedback.

Notable Quotes

"This design process that designers have been taught, we sort of treat it as gospel. That's basically dead." - Jenny Wen

"You as a designer actually, like, do not have the time to make these beautiful mocks anymore." - Jenny Wen

"The way that you really lose trust around quality and releasing something early is if you release it early and then nothing ever happens." - Jenny Wen

"Someone still needs to be accountable for the decision." - Jenny Wen

"It is what it is." - Jenny Wen, on her go-to work and life motto