“And I think the angle with Shepherd ultimately is we need to keep evergreen code. We want to keep our code healthy, … we want to give people the primitive to be able to do that.”
Context. Ramakrishnan is answering a question about Shepherd, which the host describes as Uber's platform for large-scale code changes such as migrations and upgrades. The recording is captioned 'Shepard' on the session page; both passes heard 'Shepherd'.
Nikhil Ramakrishnan, AI Foundations & Developer Experience. Works on the developer platform and builds the tools he describes (Minion, DevPod), so he knows how they are used; an insider speaking at a vendor-run summit about his employer's own work, with no figures beyond a stated target.
How this was checked
Checked against the source by a first reader, then confirmed independently by a second reader who fetched the source without seeing the first reader's notes (Independent review agent (Claude Sonnet), blind second pass from passage text and citation only, no repository access, 2026-11-07).
Exact match in whisper.srt around 10:08 ('we need to keep evergreen code. We want to keep our code healthy'), with the ellipsis covering the omitted words. The captions differ, so I confirmed on whisper only.
How to cite this passage
Nikhil Ramakrishnan (Uber), “Uber Minion: Agent-Ready Dev Platform Behind 11% of Merged PRs”, Ona (YouTube), Background Agents Summit, 2026-05-07, at 10:08. https://www.youtube.com/watch?v=o7G5EHDqQ1g Archived at https://the-ai-record.com/passages/uber-shepherd-evergreen-code-primitive-large-scale-change-2026
Context. Ramakrishnan is answering a host question on whether Uber's builder-oriented engineering culture drives how fast it adopts AI. The sentence before it, on mandating AI use, is left out because the two passes hear it differently.
Context. Ramakrishnan is describing Minion, Uber's background-agent platform, and the engineers who get toil tickets, such as a vulnerability fix with a long validation loop, in the middle of product work.
Context. Ramakrishnan is answering a host question on how Uber picks use cases; just before, he says Uber goes strictly by developer-satisfaction data when developers cannot choose.
Context. Ramakrishnan is describing how Uber's platform team gets engineers to keep using a tool, after saying that adoption comes from letting people fail in a safe way.
Uber