The 3-Person Unicorn Startup

NFX essay, July 2023. By James Currier, General Partner at NFX.

Key Takeaways

• The thesis: generative AI lets three exceptionally talented people build software businesses reaching $100m+ in revenue, by automating the labour-intensive functions that used to require headcount.
• The mechanism is allometric. “Fewer people means less of everything that isn’t building and selling to customers” — the overhead that scales with team size compounds against you, so removing people removes it superlinearly.
• The precedent is pre-AI: Instagram at 12 employees when sold for $1bn; WhatsApp at ~50 for $21bn. What was possible only in social is now claimed across industries.
• The uncomfortable corollary: “AI is going to beat most people’s average performance” on routine work, so the binding requirement shifts from technical execution to vision and judgment — which makes the model outlier-dependent, not merely lean.
• The three roles: an aggressive generalist CEO, a “words person” (marketing and sales messaging), and a “numbers person” (metrics and data). All three have to be outliers and versatile.

Connections

• Saved against No One Wants A Boss Anymore under the note “Single person unicorn.” The tension worth drafting: Currier’s argument is about leverage, but Kyle’s note title says single person — and the essay explicitly says three, with the third-person requirement doing real work. The gap between “a company of one” and “a company of three outliers” is where the interesting argument lives.
• The outlier-dependency point cuts against the optimistic reading: if AI beats average performance, the model only works for people who were already exceptional, which makes it a story about inequality of talent returns rather than democratization.