Notes: The Design Process Is Dead - What's Replacing It

This post summarizes Jenny Wen’s conversation on Lenny’s Podcast about how AI is fundamentally reshaping the design profession. Jenny was head of design for Claude and now leads design for Claude Co-work. She previously served as director of design at Figma (leading FigJam and Slides), and has worked at Dropbox, Square, and Shopify.

The Traditional Design Process Is Dead

Jenny gave a talk in Berlin titled “Don’t Trust the Design Process,” arguing that the traditional design workflow — research, discovery, diverge, converge, repeat — that designers treated as gospel is essentially dead. It was already dying before AI, but now that engineers can spin up multiple AI coding agents simultaneously, designers must let go of the old process entirely.

Two Types of Design Work Emerging

Design work is becoming stratified into two categories:

1. Supporting execution and implementation — Engineers are using AI agents to rapidly build features. Anyone can pitch an idea and an engineer will quickly create a scrappy version. Designers no longer have time to produce polished mocks upfront. Instead, the role shifts to consulting: jamming with engineers, giving real-time feedback, and polishing implementations collaboratively.

2. Creating vision and direction — This work remains critical but has changed shape. The old approach of creating 2-year or 5-year vision decks is gone. Visions are now 3 to 6 months out and might just be a prototype that points everyone in the right direction. In a world where anyone can build anything with AI agents, someone needs to ensure the team is working toward a coherent goal.

This Is Not Just an AI-Company Phenomenon

Jenny’s talk resonated widely across the industry, not just at frontier AI companies. Designers at all types of companies are starting to feel this shift. However, there has also been backlash from those who have invested their careers in the traditional process and resist the idea that discovery and established rituals are no longer essential.

A Day in the Life at Anthropic

Jenny described her typical day as having several distinct components:

• Keeping up with internal developments — Anthropic’s Slack is a “gold mine” of model developments, prototypes with codenames, internal philosophical debates, and insights from researchers. She finds staying current more compelling here than at any previous company.
• Short-horizon vision work — Setting aside time to think about where the product goes in the next 3 months, given where models and the market are headed.
• Jamming with engineers — Conversations, whiteboarding, reviewing what they have built, and providing design consulting on the fly.
• Writing code — Polishing and implementing frontend details directly using Claude Code, especially in VS Code where she can see the code and talk to Claude simultaneously.

The Changing Pie Chart of a Designer’s Time

A few years ago, Jenny’s time split roughly:

• 60-70%: Mocking and prototyping
• 20%: Jamming and consulting with engineers
• 10%: Coordination and meetings

Today it looks more like:

• 30-40%: Mocking and prototyping
• 30-40%: Jamming and pairing directly with engineers
• A meaningful slice: Implementation and coding

Tools and the Figma Question

Jenny’s AI stack is fully Claude: Claude chat, Claude Co-work (which she has migrated most of her chat use cases to for longer-running tasks), and Claude Code in VS Code for frontend polish. She also uses Claude Code remotely through Slack — someone mentions an off icon, she @-mentions Claude, and the PR is done.

Figma Is Still Important

Despite the “code is the future of design” debate, Jenny still uses Figma and finds it valuable for:

• Exploring many options at once — Design benefits from throwing 8-10 different directions at the wall. Coding tools are too linear for this; you invest deeply in one direction and iterate on it.
• Fine visual and interaction details — Trying different typography, styles, and micro-interactions on a canvas is still more efficient than going directly to code for those decisions.

Interestingly, while engineers are moving away from IDEs toward command-line agents, the IDE may become more relevant for designers who need to tweak CSS and small visual details directly.

Building Trust Through Speed

When asked how to maintain design quality when products ship constantly, Jenny shared Anthropic’s approach of “research previews.” They ship things early with clear caveats, knowing it is the worst the product will ever be.

The key is the promise to users: we will iterate, we will take your feedback, and we will visibly improve. What destroys trust is shipping something early and then nothing happening. What builds trust is responding to feedback publicly and shipping fixes rapidly — which Claude Code enables.

Where Will Human Brains Still Be Valuable?

Jenny believes AI will get better at taste and judgment, and that designers may be “holding on a little too much” to the idea that aesthetic sense is uniquely human. However, she sees enduring value in:

• Decision-making accountability — Someone still has to decide what gets built and what matters. Just as engineers are still accountable for AI-generated code working correctly, someone must be accountable for product decisions.
• Resolving disagreements — The hardest parts of building software are often not the building itself, but the debates about what should go into a feature. AI can weigh in, but it cannot resolve human disputes.
• Choosing what to build — Even as AI helps generate ideas and even comes up with suggestions for what to build next, a human must still own the final decision.

The IC vs. Manager Question

Jenny joined Anthropic as an IC after being a director of design at Figma (managing 12-15 designers). She chose to go back to IC work because she wanted to be close to the work during this pivotal time, and had doubts about whether middle management would persist.

What She Gained From Going Back to IC

Being an IC through this transformative year gave her hard skills she would not have gained as a manager. The design process changed so dramatically that she believes managers who are not hands-on will struggle to empathize with or guide their teams effectively.

The Future of Design Management

Jenny believes management will persist as long as there are teams of people, but the role must evolve. Pure people management — career coaching, one-on-ones, making sure people feel good — is no longer sufficient on its own. The future manager must also provide direction on the work itself, engaging deeply with what the team is building.

Her advice for design managers: take an IC rotation to truly understand how much has changed before managing through it.

Hiring: Three Archetypes

Jenny looks for resilience and adaptability as baseline qualities, then identifies three archetypes she finds especially valuable:

1. The Strong Generalist (Block-Shaped)

Not a regular generalist who is okay at many things, but someone who is 80th-percentile good at multiple core skills. In the T-shaped framework, they are more like a block — strong across the board. This matters because the design role is stretching into PM territory and engineering territory simultaneously.

2. The Deep Specialist (Deep T-Shaped)

Someone whose T goes deeper than most — top 10% in their specialization. This could be a designer who is essentially 50% software engineer, or someone with extraordinary visual design skills. When anyone can make anything with AI, deep specialization helps differentiate.

3. The Crack New Grad

The archetype Jenny believes most companies are overlooking. An early-career person who seems wise beyond their years, humble, and eager to learn. The advantage: they come with a blank slate, no baked-in processes or rituals, and can adapt quickly to this rapidly changing environment.

Advice for Aspiring Designers

Build actual things. Use the technology. Do not feel limited by lack of experience — that lack of baggage can be an advantage. Find a community of builders (Jenny cites Socratica as an example) and share what you make.

For senior designers: you do not necessarily need to learn to code from scratch, but you should bring coding tools into your toolkit and know how to use them. The abstraction layers will keep moving up.

The Legibility Framework

Jenny is a big fan of a framework from Evan Tana (partner at SPC) that maps founders and ideas on a 2x2 of legible vs. illegible:

• Legible founder + legible idea = probably not novel; someone is already doing it
• Illegible idea = on the frontier; people do not get it yet; potentially very valuable

Jenny applies this internally at Anthropic, scanning prototypes and internal projects for illegible ideas that have energy around them but are not yet understood. She acts like an internal VC, spotting what is worth pulling into product work.

The Co-work Origin Story

A concrete example: an internal prototype called “Claude Studio” — a dense, powerful interface built on an agentic harness — confused Jenny as a designer when she first saw it. But she noticed significant energy from researchers and builders around it. Over time, concepts from that illegible prototype (like the skills framework, Claude’s plan and to-do displays, and context visualization) were extracted and refined into what became Claude Co-work.

Co-work itself took about 10 days to go from internal prototype to shipped product, but the journey of exploration, prototyping, and iteration leading up to that moment was much longer. The “10 days” narrative went viral, but the real story involved many people, many prototypes, and many ideas converging at the right moment.

Management Philosophy

Low-Leverage Tasks Are Actually High-Leverage

Jenny challenges the common management advice to only do “high-leverage” work. She has observed that the leaders she respects most choose seemingly low-leverage tasks — dog-fooding the product obsessively, reproducing bugs with engineers, fixing bugs themselves, handmaking anniversary cards for team members. When a senior leader does this, it creates familiarity with the product, signals that nothing is beneath them, and builds deep team trust.

Psychological Safety Through Roasting

Jenny’s teams are known for being both happy and high-performing. Her approach: create psychological safety where team members feel comfortable poking fun at each other and even at her. This is not forced roasting — it is a natural sign that people feel safe, trust each other, and are not afraid of their manager.

The balance: pair that safety with genuinely high standards. These two things are not in tension — psychological safety actually makes it easier to hold people to high standards because feedback can be given without fear. Jenny describes it as being like a “tough parent” — always there for the team, but always pushing for the best work.

Lightning Round Highlights

• Books: The Power Broker by Robert Caro (for long-arc thinking about how someone gets things done across decades) and Insomniac City by Bill Hayes (a memoir about love, mortality, and Oliver Sacks’s final days)
• Movie/TV: A Sentimental Value (Norwegian film by the director of The Worst Person in the World) and The Pit season 2
• Product: Retro, a community photo-sharing app with no social media trappings — after 2 years of use, it has become a way to revisit each week of her life
• Life motto: “It is what it is” — not defeatist, but a reminder that you cannot control everything, and sometimes you need that levity to move forward
• Favorite Co-work use case: Introspection — pointing Co-work at a folder of years of personal notes (one-on-ones, random thoughts, memos, interview notes) and having it extract insights, like generating a rubric for evaluating design craft based on implicit patterns in her own writing

Key Takeaways

1. The design process must evolve — The traditional research-diverge-converge cycle cannot keep up with AI-powered engineering velocity
2. Designers are becoming hybrid roles — Part consultant, part coder, part vision-setter, with less time spent on polished mockups
3. Ship early, iterate visibly — Trust is built through speed and responsiveness, not through waiting for perfection
4. AI taste will improve — But human accountability for decisions remains essential
5. Managers should go back to IC work — To understand how much the craft has changed before managing through this era
6. Look for block-shaped generalists, deep specialists, and eager new grads — The three archetypes most valuable in this new landscape
7. Spot illegible ideas with energy — Act like an internal VC, finding prototypes that people are excited about even when nobody can fully articulate why