The AI Record
So if you think about back in the good old days of 2022 and 2023, when GitHub co-pilot first came out, it was a pretty novel way of augmenting developments. … we saw about maybe a 10 to 15% bump in overall diff velocity. … but this by itself didn't push us in the direction that we've seen over the past, let's say, year, where the paradigm has shifted to peer programming, where you could hand off workloads that are running asynchronously.

Anshu ChadhaDirector of Engineering, Dev Platform, Uber · Leader, Uber

Context. Chadha is describing what Uber measured when GitHub Copilot arrived and how the work has shifted since.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 03:46.

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025

Filed under

Who is speaking

Anshu Chadha, Director of Engineering (Engineering Director on Uber's Dev Platform). Leads Uber's developer platform organization and describes its own tooling, adoption, measurement and cost first-hand in his own recorded talk; promotional for the employer, though he also states the drawbacks (slow adoption, a 6x cost rise).

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, 2026-11-04).

Around 3:46 in whisper. The ellipses cover 'This is a pretty phenomenal,' and an earlier sentence. The remaining wording matches. Speaker is Anshu.

How to cite this passage

Anshu Chadha (Uber), “Uber: Leading engineering through an agentic shift - The Pragmatic Summit”, The Pragmatic Engineer (YouTube), 2026-03-10, at 03:46. https://www.youtube.com/watch?v=i1tZN41VKcE Archived at https://the-ai-record.com/passages/uber-back-2022-2023-copilot-10-15-diff-velocity-peer-programming-2026

More from the same source

… the issue is that these are activity metrics … These are not necessarily business outcomes. … when we start talking about the costs of this technology, … our CFO has asked me, what is the impact of this … I can't point him to diffs. I need to show him what's the impact on revenue. I'm sure you folks are dealing with the same problem. This is not necessarily a solved problem for us.

Anshu ChadhaDirector of Engineering, Dev Platform, Uber · Leader, Uber

Context. Chadha is describing the gap between the activity metrics Uber tracks for AI use and the business outcomes its CFO asks about.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 35:07.

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

As soon as we made our agentic … available to developers, what we saw is 70 percent of the workloads that developers are pushing into the system were toil tasks. There's a couple of reasons for that. One is the accuracy of these tasks was much higher compared to the more ambiguous workloads. And it makes sense, right? Like the start and end state of some of these tasks, if you think about a library upgrade or a migration, is much more straightforward versus, say, building a brand new feature or an experience … that requires an experiment.

Anshu ChadhaDirector of Engineering, Dev Platform, Uber · Leader, Uber

Context. Chadha is describing what developers first sent to Uber's agentic system and why.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 06:47.

  • 70% toil workload share, after agentic workflows became available to developers. Level not stated

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

what we found is by really focusing on this project, building a custom agent on top of our internal … we were able to get a much higher quality type of unit test output. And so at this point, we're seeing about 5,000 tests generated and merged per month around the company from this. And almost a 3x rate of quality versus something that would be generated from your typical generic agent.

Ty SmithPrincipal Engineer, AI & Developer Productivity, Uber · Practitioner, Uber

Context. Smith is describing Auto Cover, Uber's custom unit-test generation agent, and why Uber built it instead of using a general coding agent.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 24:45.

  • 5,000 tests merged per month, per month, as of the talk. Level not stated
  • 3 x test quality vs generic agent, as of the talk

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

… speaking of costs, the cost of AI is too damn high. … since 2024, our costs have gone up at least 6x. … again, no question that it's had positive impact, but it's gone from something that I can self fund using my own budget to something that I need to ask permission from … the CFO.

Anshu ChadhaDirector of Engineering, Dev Platform, Uber · Leader, Uber

Context. Chadha is describing how the cost of AI tools at Uber has grown and what it changed about who approves the spending.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 35:56.

  • 6 x ai cost growth, since 2024, at least

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

I ran a demo session with some of my VPs, and in 24 minutes, I had four VPs land code for the first time in years. It was a pretty amazing experience. They were pretty satisfied by it, too. But our adoption for this technology has been relatively slow. It's been slower than I've expected, and part of it is because we're trying to have developers do something that they're so not used to. They're used to looking at code and generating from scratch, operating in their IDE, and we're telling them to take a risk by operating in a very different way.

Anshu ChadhaDirector of Engineering, Dev Platform, Uber · Leader, Uber

Context. Chadha is describing a demo of Uber's agentic tooling to VPs and why developer adoption has been slower than he expected.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 32:45.

  • 4 people vps landing code, in a 24-minute demo session. Level not stated

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

We've tried a top-downs approach, … directives from leaders to say, you must do X, Y, and Z, you must adopt. It's had some impact as we track, as you folks know. You track a metric, it's going to go up, or it's going to improve. The more successful technique that we've applied is actually just sharing wins. … as we share examples between different engineers, cool things that they've tried that have resulted in wins, adoption of that technology has erupted. … that's been the tactic that we're pushing on now is key promoters pushing techniques to their peers, because those promoters are typically engineers, and engineers trust other engineers as opposed to directors like me.

Anshu ChadhaDirector of Engineering, Dev Platform, Uber · Leader, Uber

Context. Chadha is comparing the ways Uber has tried to raise adoption of its agentic tools among engineers.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 33:25.

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

Right now, … the workflow for the developers has changed and is changing further. They're spending more and more time in planning and code review because there's so much more code being generated that they're being forced to do it. This probably isn't the favorite type of work that developers love doing, code review.

Ty SmithPrincipal Engineer, AI & Developer Productivity, Uber · Practitioner, Uber

Context. Smith is describing how developers' time at Uber is shifting as agents write more code.

Talk: Uber: Leading engineering through an agentic shift - The Pragmatic Summit, The Pragmatic Engineer (YouTube), March 10, 2026, at 18:38.

Checked against the source [ Grade B ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

Something wrong? Report a correction. JSON: /api/v1/passages/uber-back-2022-2023-copilot-10-15-diff-velocity-peer-programming-2026.json