Summary
Eddie Kim, CTO and co-founder of Gusto, joins the How I AI podcast to recount how a five-person team — four engineers (including Kim) and one designer — built “Co-founder,” a new AI-powered product, in just ten weeks from zero code to a tier-one launch at Gusto’s 1,000-person R&D organization. The technical stack was deliberately minimal: Cloudflare Workers for the agent loop and Vercel AI SDK for model-switching flexibility.
Kim describes a development culture that eliminated nearly all traditional process overhead — no Figma handoffs, no Jira, no sprint planning. Instead, the team used what he calls the “trash can method”: ship features fast, evaluate them in real time, and be willing to delete entire branches and rebuild from scratch because the cost of code with AI assistance has dropped to near zero. A standout workflow was the “prototype-to-production” pipeline, where a fully faked front end with canned responses was shipped behind a feature flag and then slowly had real AI functionality breathed into it — allowing designer and backend engineers to work in parallel without blocking each other.
Perhaps the most striking finding is the performance of designer Katie, who — with no formal software engineering background — landed in the 94th percentile for PR throughput across Gusto’s entire R&D organization, outperforming classically trained engineers. Kim explains her edge as a combination of technical curiosity and working closely with a small team of engineers who coached her through the gap. He frames the whole Co-founder build as a proof of concept for how Gusto’s broader 1,000-person R&D organization might work — and the challenge of scaling that culture without losing what made the small team so effective.
📺 Source: How I AI · Published June 29, 2026
🏷️ Format: Interview
