Jenny Wen on Designing Claude, the Legibility Framework, and the Three Designer Archetypes

Speaker:
Jenny Wen
Source:
Jenny Wen on Designing Claude, the Legibility Framework, and the Three Designer Archetypes
Episode:
Listen →
transcriptdesignaianthropicclaudefigmalenny

Jenny Wen on Designing Claude, the Legibility Framework, and the Three Designer Archetypes

Jenny Wen is Head of Design for Claude at Anthropic, previously Director of Design at Figma. She spoke at Config with a talk titled ‘Don’t Trust the Design Process’, arguing that process optimised for predictability is wrong for the current moment, when the right answer is genuinely unknown and the design surface changes faster than any process can track.

Key ideas

  • Design is stratifying. The role is splitting into two distinct activities: execution support (implementation, component consistency, shipping fast with AI tooling) and vision creation (deciding what should exist, what the interaction model should be, what the product means). AI compresses execution time; the premium shifts toward the second activity.
  • Legibility framework. Not every good idea is legible when first encountered. Borrowed from VC Evan Tana (a partner at SPC) and re-applied to design, the two-by-two crosses legibility (can observers immediately grasp this?) against energy (does it generate excitement and curiosity?). Illegible ideas with high energy deserve the most investment — they are where genuine novelty lives and where conviction without full comprehension is the right posture. The internal Claude Studio prototype, which Wen found incomprehensible but visibly energising, seeded the skills framework and much of Cowork’s interface.
  • Three designer archetypes. Block-shaped generalists are 80th-percentile across several core skills and flex into PM- and engineer-shaped work. Deep-T specialists run top-10% deep in one domain. Craft new-grads arrive early-career, humble and fast-learning, unburdened by ingrained process. Each serves a different function; teams that understand which they are hiring avoid systematic mismatches.
  • “Low leverage” tasks are often the highest leverage. The tasks that look like design distraction — dogfooding, reproducing bugs, filing a PR, vibe-coding an anniversary card — are often where familiarity and influence are built. A senior leader doing them signals that nothing is beneath them.
  • Roasting as psychological safety signal. Teams that can roast each other’s work — and the manager’s crit phrases — are signalling real safety: honest enough to be uncomfortable, secure enough to survive it. Paired with high standards, this is Radical Candor in practice.

Topics covered

  • ‘Don’t Trust the Design Process’ talk and its argument
  • AI’s compression of implementation time and what designers do with recovered capacity
  • Implementation slice: Jenny’s team’s process for working directly in code
  • Claude’s design stack: Cowork for collaborative work, Claude Code for implementation
  • Figma’s continued value for multi-direction exploration
  • Cowork’s origin story: 10 days from internal prototype to external product
  • Legibility framework: mapping ideas by legibility and energy
  • Three designer archetypes: block-shaped generalists, deep-T specialists, craft new-grads
  • Roasting as psychological safety signal in critique
  • Design’s invisible leverage in engineering and sales settings

Notes

For source-grounded literature notes (Adler frame, glossary, framework detail with section citations), see Jenny Wen on Designing Claude, the Legibility Framework, and the Three Designer Archetypes under wiki/notes/.