The Takeaway: One founder can run a company if agents handle the layers below.

The Takeaway: Andrew, founder of General Intelligence Company of New York and a second-time CEO who dropped out of college at 18, thinks the real breakthrough isn’t a smarter assistant — it’s a company that runs itself. His contrarian bet is that the future isn’t “one agent for everything,” but a management stack: human at the top, coordinator agents in the middle, and specialized agents underneath for code, support, sales, and operations. That philosophy came out of building Co-founder, which already helped more than 8,000 startups automate pieces of their work. But Andrew draws a hard line: an assistant that writes better emails is not a company. So his team rebuilt the problem around the actual founder workflow: create, sell, support, operate. The new product, Co-founder CTO, is his answer to the first step — an autonomous engineering department that plans work, writes code, spins up preview environments, runs browser tests, merges PRs, and monitors infrastructure without local development at all. The sharpest idea here is that the “middle management” layer is the product. Andrew isn’t trying to replace every task with one giant model; he’s orchestrating agents like a video game, borrowing the logic of Starcraft and Civilization. He even says, “we decided we didn’t want to do local development at all.” That’s the kind of move that sounds extreme until you realize it removes an entire class of friction. His internal numbers are the tell: 400% faster, about 25 PRs per engineer per day, and some days more spent on tokens than salaries. The bigger thesis is blunt: companies won’t just use AI — they’ll be reorganized around it.