Summary
DoorDash co-founder Tony Xu sits down with Anthropic to walk through how the company rolled out Claude Code to every employee — including non-engineers — as part of what they call a “co-work” initiative aimed at raising AI fluency across the org. The conversation covers Xu’s personal journey: he stopped writing code as DoorDash scaled, then made a comeback shipping production code in five languages using Claude Code agents running in parallel sessions via Git worktrees.
The video details concrete organizational outcomes: a single engineer completed a large-scale code migration in three weeks that historically would have taken four engineers a full quarter, and multiple product launch timelines were cut by more than half. DoorDash built internal infrastructure called Flux — a platform of security-blessed VMs in the DoorDash cloud — to run Claude Code sessions at scale, including powering an internal AI code review agent built on the Claude Agent SDK.
Perhaps the most interesting thread is how DoorDash is extending AI coding tools beyond engineers: designers and product managers are now embedded directly in the software development cycle and shipping to production. Xu describes the challenge of getting teams to unlearn attachment to past implementations as models improve rapidly, and frames leadership’s job as setting aggressive time-reduction goals — 3 to 5x — and then studying the pockets where teams actually achieve them. The video is a candid, unusually detailed look at what company-wide agentic coding adoption looks like in a large consumer tech company.
📺 Source: Claude · Published July 07, 2026
🏷️ Format: Workflow Case Study
