The Napkin Test

A test the author has to pass afterward

Micky Malka describes how Ribbit Capital checks its own thinking: "they start by being probably 20 pages long, but I can actually get a napkin and explain to you each one on a napkin. And if anybody can explain anything complex to you on a napkin, it means that they have studied enough that they know how to articulate it."¹ The twenty pages still exist; the napkin does not replace the essay, it is a test the essay's author has to pass after writing it. Compression is evidence of understanding rather than a substitute for the work that produces understanding in the first place.

The test gets applied at two very different scales in the same conversation. The token thesis, the idea that every company is essentially a token factory producing machine-readable information whose delivery requires identity, value, and intelligence, opens on a napkin before expanding into labs, application layers, money, and blockchains. And Ribbit's founder screen, known internally as Eye of the Tiger, a five-part framework that took roughly ten years to develop, was compressed onto a napkin about five years in, a moment Malka treats as confirmation rather than simplification: "if we wrote it on a napkin, it means that we've thought about it long enough that we have it right." A decade of pattern recognition, in his account, is not finished until it fits.

Why it matters

The test supplies something normally unavailable: a cheap, immediate, objective check on whether someone actually understands a claim rather than just sounding fluent about it. Fluency is not evidence, length is not evidence, and the ability to answer follow-up questions is not evidence either, since all three can be produced by Chauffeur Knowledge, someone who has memorized an explanation without originating it. Reconstructing the whole structure from nothing, at napkin scale, in front of another person, is far harder to fake. The test also inverts the usual relationship between effort and length: the natural output of hard thinking in progress is more words, while the natural output of thinking that has actually finished is fewer.

Open question

A napkin can hold a slogan as easily as it can hold a real model, and the test screens for articulacy, which correlates with understanding without being identical to it. Some ideas may also simply not compress: a ten-year pattern about how founders behave might fit on a napkin, while a mechanism with genuinely interacting parts might not, which means treating napkin-fit as the finish line risks selecting for whichever ideas happen to be compressible rather than whichever ones are actually true. Ribbit's own founder screen has been on the napkin for roughly five years with no recorded revision since, which is either a sign the framework is genuinely finished or a sign that a finished napkin is also a frozen one.