Legibility Framework
A two-by-two for judging early ideas by how readily others grasp them. The framework was originated by the venture investor Evan Tana (a partner at SPC) to sort founders and ideas, and re-applied to product ideas by Jenny Wen, who leads design for Claude Cowork at Anthropic.
The two axes
An idea is legible when an observer immediately understands what it is, and illegible when it sits on the frontier — genuinely novel, or simply told in a way that does not yet land. Tana’s original grid crossed idea-legibility with founder-legibility:
- Legible idea, legible founder. Probably not novel. If everyone can see it, someone is already building it.
- Illegible idea. The interesting territory. Either nobody has done it, or the person describing it cannot yet make it land.
Wen’s reapplication adds the second thing she watches for — energy. An illegible idea that nonetheless generates excitement and curiosity, often among people who cannot articulate why it excites them, is the one worth the most investment.
Legibility measures familiarity, not quality. The fact that an idea is hard to read says nothing about whether it is good.
Why illegible-plus-energy is the signal
A legible idea is, by definition, one the market has already priced. Novelty lives where comprehension lags — so the ideas most worth backing are the ones that look confusing but pull attention anyway. The investor’s job, and Wen argues the frontier designer’s job, is to sit with that confusion rather than dismiss it, and then do the work of making the idea legible: through storytelling, through interaction design, through the form factor.
The designer as internal VC
Wen scans Anthropic’s internal channels the way a VC scans a deal flow — looking for the prototypes that carry energy she does not yet understand, then diving in to understand them. Her worked example: an internal prototype called Claude Studio looked incomprehensible to her (‘I don’t know what’s going on’), but researchers were visibly energised by it. Out of it came the skills framework and the instruction-markdown files that tell Claude how to do a task; its display patterns — Claude’s plan, its to-dos, its context and files — were later pulled into Claude Cowork.
Relation to early-stage pattern-matching
Lenny notes a parallel finding from a study of people who joined Palantir, Stripe, Linear, Notion, and OpenAI very early: one recurring marker was that the idea sounded so crazy outsiders laughed at it, paired with visible excitement from those close to it, and a top-1% founder. The legibility framework names the same instinct — back the idea others cannot yet read, when the energy around it is real.
Source
See Jenny Wen on Designing Claude, the Legibility Framework, and the Three Designer Archetypes and the notes for the worked Claude Studio example. Related: Three Designer Archetypes, Product Taste, Inflections and Pattern Breakers.