Photo
Aaron Levie, co-founder and CEO of Box, told a16z's own show that building a startup in 2025 is the most different it has looked in his 27 years in software.
Most founders still think their speed is capped by how fast they can type or write code themselves. Aaron said that cap is gone: a three-to-ten-person team orchestrating AI agents can now produce what used to take a team of 50 to 200.
"When we started Box, our rate of progress was 100% correlated to how many hours we could be on a keyboard and type things."
Aaron co-founded Box in a garage in 2005, when the playbook looked almost identical to a 1995 startup's, and has run the company as a public enterprise-software business for over a decade.
The full interview is covered here so you can skip it. 33 minutes of audio, 8 minutes of reading.
Here are the 7 lessons that matter.
a16z · Oct 25, 2026 · 33m · 25m saved
With Aaron Levie, Co-Founder and CEO at Box
Key Takeaways
A 3-to-10 person team can now match a 50-to-200 person team's output, Aaron said, because AI agents remove the old limit of how fast one person can type or code
Being an editor of AI-written code is 10x the leverage of writing it yourself, since one person can run many agents in parallel but only operate one terminal
Agents are "a slot machine," not a panacea — they still make mistakes, and incumbents can't match the gains because their workflows are too wired into old ways of working
The premium has shifted from typing speed to taste and specs — what to build, and specialized sub-agents instead of one "super agent"
Legal review that costs $1,000-2,000 an hour today could run $5-10 an hour with an agent, Aaron said, unlocking demand that price was suppressing
Hiring a person costs $100,000-300,000; standing up an agent might cost $1,000
Aaron bets on enterprise over consumer by sheer market size, even though his own personal AI use cases already feel "fully solved"
Distribution still decides who wins — "if you build it, they will not come," Aaron said, and a great agent alone will be "drowned out by 10,000 other startups"
1. The most different playbook
Aaron opened by comparing startups across decades. A 1995 company and his own 2005 one, Box, ran on an almost identical playbook — same keyboards, same data centers, same link between hours typed and progress.
Comparing 2025 to 2005 or 2015 is the most different shift Aaron has seen in 27 years
"If I compare the 2025 startup to us in 2005, or let's say the 2015 startup — using 10-year increments, it is the most different that I've ever seen in the history of doing technology now for 27 years." — Aaron Levie
That shift, in his telling, comes from AI-first orientation: a three-to-ten-person team can now produce the output of a team ten to twenty times its size.
