Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab

Individual gains from coding agents can shift work onto teammates who must review the extra code. In the publisher's account of Aditya Khandelwal's talk, his experience leading a ten-person team at Amazon AGI Lab supports treating adoption as a leadership responsibility: shared repository structure and working practices require team-level decisions.

His team's example is a reusable skill, a packaged set of instructions an agent can load, called ship it. It is described as carrying a change from finished code to a review-ready pull request, handling the description, review comments and failed checks. This is a team case account, not a controlled productivity comparison.

“Leave agent adoption to individuals and the engineer shipping two PRs a day ends up reviewing the ten that the early adopter ships.”

“adoption is therefore a leadership problem and not an IC one, because the changes that actually work, restructuring a codebase for progressive disclosure and converging on a shared setup, are not changes one engineer can make alone.”

Aditya Khandelwal

“Someone says the model got dumb today, when the model did not change and the harness did.”

“Their fix centered on one high value skill, called ship it, that carries a change from code done to PR ready, handling the description, the review comments, and CI failures, and often runs for over an hour.”

“It was not clean: agents filing against each other took the repo to roughly 4,500 open issues in a couple of weeks.”