General Intelligence, an 8-person startup with 5 engineers, runs over 4,000 active Git branches and 100 parallel app versions on Vercel at any given moment. Each engineer ships 10 PRs and 70-plus commits per day on a $5,000 monthly token budget. That output is not human. It comes from Cofounder, their own AI agent platform, which uses a CTO agent to drive deployments, manage configs, change DNS, and handle billing entirely through Vercel's CLI and API.
The reason this case study is worth reading in full is not the headline numbers. It is the infrastructure logic behind them. General Intelligence left Render because provisioning preview environments for a full Python stack was too manual, and coding agents need programmatic access to everything, not 50 percent of operations. Vercel gave them one CLI, one API, and one observability layer across frontend and backend. When the CTO agent breaks something, both the agent and the human developers see the full picture in one place. Fluid compute usage grew 6.5x month-over-month earlier this year, almost entirely from internal agent activity.
The product ships this same architecture to customers. Every founder who signs up for Cofounder gets a provisioned GitHub repo, a managed Vercel deployment, a custom domain, and SSL, all spun up automatically through Vercel for Platforms. The customer's CTO agent runs the same branch-test-deploy loop General Intelligence uses internally. The company's next step is expanding the agent roster across more business departments, built on the same stack they sell.
[READ ORIGINAL →]