Eddie Kim, Gusto's CTO, took an unusual approach by diving back into coding to launch an AI product in just 10 weeks. With a lean team and a bold methodology, he challenges traditional software development norms.
At first glance, Eddie Kim's return to writing code might seem like a detour for a CTO. But under his leadership, Gusto has taken a bold step. Over ten weeks, Kim and a small team of three engineers and one designer built Gusto Cofounder, an AI-driven product, from scratch. This sprint wasn't about just slapping a model on a GPU rental. It was a testament to the power of focused, agile development.
The Trash-Can Method #
Forget the bureaucracy of endless planning documents. Kim's team adopted what he calls the 'trash-can method.' Instead of drafting lengthy product requirements, they wrote and reviewed full pull requests, discarding them as needed. It's agile development stripped to its core, bypassing the usual red tape. This approach allowed rapid iteration, letting product decisions emerge organically from the code itself.
Reimagining Team Dynamics #
The project also challenged conventional team setups. With no project managers or typical standups, they relied on a 'perma-Zoom' setup. This kept communication smooth, replacing what would typically be a labyrinth of Slack threads and status meetings. And then there's the designer who, in this unconventional setting, hit the 94th percentile for shipping code. If a non-engineer can reach this level, what does it say about traditional engineering hierarchies?
From Prototype to Production #
Launching to production from day one, the team didn't just prototype, they committed to real-world application. The use of Claude Code to solve customer issues demonstrated a commitment to practical results over theoretical perfection. Kim's methodology poses a provocative question: if a lean team can achieve this in ten weeks, why can't larger teams do better?
Gusto Cofounder may signal a shift in how AI products get built. It's about more than the technology itself. It's about rethinking the processes that have long defined the software industry. Show me the inference costs, and then we might just have a new blueprint for innovation.
Get AI news in your inbox
Daily digest of what matters in AI.