Writing · Pragmatism
A prototype ends the argument
James Dyson built 5,127 prototypes before his vacuum cleaner worked. Jeff Hawkins carried a block of wood in his pocket. A model on the table settles questions that a deck cannot.

In the late 1970s James Dyson noticed that his vacuum cleaner lost suction as its bag filled with dust. At his factory he had built a cyclone tower that pulled paint particles out of the air by spinning it. His first test was a cyclone of cardboard and tape, fixed to his own vacuum cleaner where the bag had been. It was crude, and it worked well enough to keep going. Making it work properly took, by his own count, 5,127 prototypes.
The big makers turned it down. Selling bags was good business for them, by Dyson’s account more than $500m a year in Europe alone, and his machine needed none. He licensed the design in Japan, where it sold as the G-Force, then started his own company. In 1993 the DC01 went on sale.
A block of wood
Around 1995 Jeff Hawkins had a different problem. Handheld computers had failed one after another, Apple’s Newton among them and Palm’s own Zoomer too. Before building anything, he measured his shirt pocket and cut a block of wood to fit it. A chopstick served as the pen. He took the block into meetings, scribbled imaginary notes on it and kept a tally of how many taps an everyday job would need. The block is now in the Computer History Museum in California.
The PalmPilot that followed cost $299. About a million were sold in its first 18 months.
What a prototype does that a deck cannot
A deck describes a product. A prototype makes it answer back. It shows you that the thing is too heavy or that the button sits where no thumb will find it. Sometimes it shows that customers want it for something else. An argument that would run for weeks in a meeting ends in a minute when the model is on the table.
This is no plea for endless tinkering. Each of Dyson’s prototypes was cheap and tested one change. Hawkins’s block cost almost nothing and settled the size of a product before anyone designed a circuit. The skill is to build the cheapest thing that answers the most expensive question.
What we ask founders
Industrial design taught us the habit, and we keep it in every company we start. Our first question to a founder is rarely about the market. It is: what can you show us? A cardboard model will do, or a recording of five customers trying a rough version. Either says more than 30 slides.
Investors like plans because plans are easy to compare. A product is easier to judge. When a founder puts something on the table, we spend the hour on it.
Sources
More writing
EnergyAI runs on electricityData centres will use about twice as much power by 2030. Microsoft has bought 20 years of power from one reactor at Three Mile Island, and Google and Amazon back small reactors. AI now waits on the plug.
RoboticsRobots learn from peopleLanguage models learned from trillions of words. Robots have no such library yet. In robotics the scarce thing is now data on how people work.
CapitalEfficiency is the new moatDeepSeek wiped $589bn off Nvidia in a day by doing more with less. The lesson for founders is older than AI: whoever needs the least money is hardest to catch.