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

Episode Overview

• This episode explores the fundamental collapse of the traditional "Double Diamond" design process due to the explosion of engineering velocity enabled by AI coding assistants.
• It frames the evolving role of the designer, shifting away from pixel-perfect mockups toward "system equipping," code-level collaboration, and high-level directional guidance.
• The discussion provides critical advice for hiring in the AI era, specifically identifying the value of "cracked new grads" and generalist "block-shaped" talent over rigid specialists.
• Readers will learn practical frameworks for innovation, including how to spot "illegible" engineering prototypes and transform them into valuable product features.

Key Concepts

• The Death of the Traditional Design Process: The rigid "discover, diverge, converge, build" flow is obsolete. Engineering speed (boosted by AI) now outpaces the time it takes to create polished mockups. Designers who cling to long phases of static specification risk becoming bottlenecks.
• The Designer's Bifurcated Role: The role is splitting into two extremes: Execution Support (polishing code engineers have already built) and Directional Vision (setting a rough 3-6 month "North Star" rather than a 5-year roadmap).
• Designing for Non-Deterministic Outcomes: Because AI models are probabilistic (outputs vary), static tools like Figma cannot capture the core user experience. Designers must prototype in code or use the actual models to understand edge cases and "feel," rather than theorizing about states that may not exist.
• The "Last Mile" Implementation Shift: Designers are increasingly acting as "gardeners" of code, using AI agents to implement UI polish directly in the repo. This removes the friction of filing tickets for minor visual changes and blurs the line between design and front-end engineering.
• Divergent vs. Linear Workflows: While AI coding agents are excellent at linear execution (finishing a task), they lack the ability to explore laterally. Figma remains essential for the "divergent" phase—visualizing 8-10 different approaches simultaneously—before committing to a code path.
• Trust Through Speed (The Research Preview): In the AI era, trust is built through responsiveness rather than initial perfection. Launching "Research Previews" allows teams to fix issues rapidly based on feedback, which often generates more user trust than a slow, polished launch.
• The "Legibility" Framework: Designers in R&D environments act as "Internal VCs." Their job is to spot "illegible" engineering prototypes—confusing concepts that generate internal energy—and use design to make them "legible" (understandable and usable) for the broader market.
• Hiring Archetypes for the AI Era: The ideal talent profile is shifting toward "Block-Shaped" generalists (80th percentile in multiple skills) and "Cracked New Grads" (blank slates with no rigid processes to unlearn).

Quotes

• At 0:05:33 - "This design process that designers have been taught... we sort of treat it as gospel... that is basically dead." - Explaining why the traditional security blanket of the 'design process' is now a liability.
• At 0:07:50 - "We used to go off and make this two-year, five-year, ten-year vision even... Now, it becomes a vision that's three to six months out." - Illustrating the collapse of strategic timelines due to the pace of AI advancement.
• At 0:09:02 - "You're better off not blocking that [engineering speed], letting them cook... and then helping engineers and teams execute... not just telling them 'here's the design.'" - Redefining the designer's relationship with engineers from 'specifier' to 'collaborator/polisher'.
• At 0:12:15 - "With these new developing AI models that are non-deterministic, you just can't mock up all the states... you can't even make a clickable prototype... you sort of have to use the actual models underneath." - Explaining why Figma/Sketch are insufficient for core AI product design.
• At 0:18:05 - "A few years ago, 60-70% of it was mocking and prototyping... Now I feel the mocking up part of it is 30-40%... and there's that other 30-40% there that is now jamming and pairing directly with engineers." - Quantifying exactly how the day-to-day workload of a modern designer is shifting.
• At 0:20:53 - "The best design happens when you're able to just throw a bunch of ideas at the wall and curate... Right now working with some of these coding tools doesn't lend itself super well to that because it's super linear." - Explaining why visual design tools are still necessary despite AI coding capabilities.
• At 0:26:58 - "It's building trust through speed... making people feel like they've been heard and that we're fixing things based on what they're trying to use it for." - On how rapid iteration replaces perfection as the primary trust-builder.
• At 0:31:09 - "Even though Claude can write all this code for you today, it is still an engineer who's accountable for... does that code actually work? Does this actually make sense in the product?" - Defining the shift from execution to accountability.
• At 0:47:16 - "People that are like block-shaped... they're really good at a few core skills. Like 80th percentile good... The design role... is stretching and spanning." - Explaining why versatility beats specialization in adapting to AI workflows.
• At 0:50:48 - "Someone who is early career... feels wise and experienced beyond their years... having somebody who almost has a blank slate... that's super valuable." - Identifying a high-value hiring archetype in the AI era.
• At 0:55:50 - "Some of their best traits is that they choose low leverage tasks that they take on themselves, and that actually ends up being a very high leverage thing because it's them who's doing it." - Challenging the standard management advice of constant delegation.
• At 1:04:18 - "Part of the role of the designer... is kind of spotting the ideas that are illegible and trying to understand what's there and how to take that and transform it." - Defining the strategic role of designers in R&D environments.

Takeaways

• Stop prioritizing static "pixel-perfect" mockups; prioritize prototypes that use live models to test non-deterministic behaviors early.
• Adopt the "Research Preview" mindset: Release features earlier than feels comfortable, then build trust by rapidly fixing issues based on user feedback.
• Pivot from being an "Architect" (handing off blueprints) to a "Gardener" (pruning and polishing code directly in the repo using AI agents).
• When hiring, look for "Cracked New Grads"—talent without rigid process baggage who can adapt to new AI workflows faster than senior specialists.
• Managers must rotate back into Individual Contributor (IC) work occasionally; without understanding the new AI toolchain, you cannot lead effectively.
• Use the "Legibility Framework" to evaluate projects: Look for confusing engineering prototypes ("illegible") that have high internal energy, and apply design to make them understandable ("legible").
• Keep a running log of raw thoughts and meeting notes, then use AI to analyze your own history to uncover your implicit decision-making rubrics.