Uber CTO: AI Writes Code
Uber’s CTO said AI agents are now writing code end‑to‑end in many cases, calling the moment a 'real reset' for engineering as teams adopt agentic workflows. The quote underscores why interviewing and hiring are shifting toward system‑level reasoning, review patterns, and human oversight of AI outputs. (businessinsider.com)
Praveen Neppalli Naga, Uber’s CTO for Mobility & Delivery, announced in a LinkedIn post that roughly 95% of Uber engineers use AI tools at least once a month. (businessinsider.com) Naga said Uber’s in‑house coding agent is producing about 1,800 code changes per week and that the agent’s share of all code changes rose from under 1% to about 8%. (businessinsider.com) An engineering-intelligence analysis of 700 companies by Jellyfish found 63% of firms now use AI tools for most coding, with top adopters seeing nearly double the number of weekly pull requests compared with lower adopters. ( ) Naga emphasized that Uber’s adoption has been grassroots — engineers “quietly experimenting” and delegating tasks to agents rather than the company enforcing mandatory AI use from the top. (businessinsider.com) The push at Uber comes amid high‑profile production mishaps elsewhere: internal reviews tied Amazon’s AI coding assistant to a March 2 outage that generated about 1.6 million website errors and ~120,000 lost orders, prompting a company-wide 90‑day code‑safety reset. ( ) Uber’s public materials describe agentic-plus-generative workflows that orchestrate multiple agents and include red‑teaming and evaluation frameworks for reliability, while conference reports show engineers at Uber assigning discrete coding tasks to autonomous agents like managers assign work. ( )
