When One Engineer Does the Work of a Team: New Rules for Technical Leadership

The author, who worked in hyper-growth companies from 2014 to 2020, reflects on how AI tools have changed the pace of work. He proposes five revised rules: 1) Migrations can be done by one specialist or lead team, completing 95% of the work in 10% of the time, but quality becomes more critical. 2) First-version code is nearly free, but working code still requires investment in dev infrastructure like tests, CI/CD, and preview environments. 3) Optimize the typical process scenario for AI agents; most steps of most processes can be fully automated, but risky areas need human judgment. 4) Stable teams with deep domain knowledge and full ownership are even more important now. 5) Companies must make high-quality decisions quickly to leverage AI. Practical results include increasing deployments from 6 per week to 200-400 per week, migrating to a monorepo, fully typing frontend code, and adopting Claude Code and Cursor for all engineers by end of February. The author also notes that managers who directly modify code see good results only if they also check their work and monitor dashboards. The company moved from Jira to Linear for better MCP and Slack integration, and automated client operations triage and initial code review using AI agents.