Uber Engineers Transition from Writing Code to Directing AI Systems
Uber's engineering teams are undergoing a profound transformation, spending significantly less time writing traditional code and increasingly focusing on instructing machines to generate it instead. According to the company's Chief Technology Officer, this strategic shift is not a cause for concern but rather the central objective. The ride-hailing pioneer, which revolutionized urban mobility, is now positioning itself at the forefront of another disruption—this time within its own engineering departments. The company believes proactive adaptation is far superior to being caught unprepared by this technological evolution.
Massive Adoption of AI Tools Among Uber Engineers
Praveen Neppalli Naga, Uber's CTO, disclosed in a recent LinkedIn post that an astonishing 95% of Uber's engineers now utilize artificial intelligence tools every single month. Furthermore, the company's proprietary internal coding agent is autonomously producing approximately 1,800 code modifications each week, completely without direct human authorship. While engineers still meticulously review and approve these AI-generated changes, they are no longer the primary authors of the code itself. The agent's contribution to Uber's total code changes has skyrocketed from under 1% to a substantial 8% in just a few short months.
"There is zero human authoring," Naga emphasized in his statement, highlighting the fundamental nature of this change.
From Autocomplete to Autonomous Agents: A Fundamental Workflow Shift
The most significant aspect of this transformation is not merely the volume of AI-generated code but the fundamental alteration in engineering workflows. Naga reports that 84% of AI users at Uber are now employing agent-style workflows. In this model, engineers delegate entire coding tasks to AI systems rather than simply accepting line-by-line suggestions from traditional autocomplete tools.
Adoption of Claude Code, a prominent AI coding assistant, nearly doubled at Uber within two months, jumping from 32% to 63%. Meanwhile, usage of traditional Integrated Development Environment (IDE) based tools has largely stagnated. Even within those conventional IDEs, approximately 70% of committed code is now AI-generated—meaning it is fully written by the AI, not just suggested. In essence, the engineering profession is being redefined. Engineers are not merely accelerating their existing work; they are transitioning to an entirely different category of work altogether.
Industry-Wide Challenges and Quality Concerns
Uber's seemingly smooth transition contrasts with a more turbulent industry-wide debate regarding the true costs and risks associated with AI-generated code. Amazon recently convened an all-hands engineering meeting following a series of service outages, some of which were directly linked to AI coding tools. In one notable incident, Amazon's internal Kiro agent attempted to resolve a problem by deleting and completely rebuilding an entire environment.
A comprehensive study by CodeRabbit analyzed 470 pull requests and found that AI-written code contained 1.7 times more issues and defects compared to code authored by humans. Anthropic, the company behind the Claude Code tool used by Uber and Spotify, introduced a specialized review feature this month specifically designed to identify errors in AI-generated output—with costs reaching up to $25 per pull request. Boris Cherny, the creator of Claude Code, has publicly acknowledged that current AI models are still "not great at coding" and that the emerging practice of "vibe coding" often falls short when long-term code maintainability and robustness are critical considerations.
Redefining the Engineer's Role in the AI Era
Uber's CTO frames this industry shift as a professional upgrade rather than a downgrade. "The role of the engineer is shifting—from writing every line to architecting systems and reviewing AI-generated code," Naga explained. He also noted that Uber's widespread AI adoption was not a top-down mandate; the most enthusiastic uptake originated from engineers independently and quietly experimenting with the tools themselves.
Andrej Karpathy, who coined the term "vibe coding," recently offered a different perspective, stating, "I've never felt this much behind. The profession is being dramatically refactored." Meanwhile, Anthropic CEO Dario Amodei has suggested that software engineering could become "fully automatable" within the next 12 months. Whether this statement is a genuine prediction or a strategic sales pitch is subject to interpretation, but Uber's leadership appears to be betting heavily that this automated future arrives precisely on schedule.
