Uber Embraces AI: 1,800 Code Changes per Week

Uber, the transportation giant, is making a decisive shift in its technological strategy by massively adopting artificial intelligence for the development of its software. Praveen Neppalli Naga, the company's Chief Technology Officer, revealed that Uber's AI agent now makes 1,800 code changes each week. This advancement is part of a broader trend where AI is redefining software production methods.

In a message posted on LinkedIn, Naga emphasized that 95% of Uber's engineers use AI on a monthly basis, describing this period as a "true reset moment for engineering." The company has ramped up its investments in AI coding, resulting in what he describes as phenomenal outcomes.

The most notable change lies in the use of agentic AI, where the software performs tasks autonomously, requiring little to no human supervision. At Uber, this translates to an increased delegation of tasks to AI agents, which write the code entirely. Engineers review and approve this code, but it is fully written by AI agents.

This transformation has allowed Uber's coding agent to increase its share of all code changes from less than 1% to 8%. This phenomenon is not isolated: a study conducted by Jellyfish on 700 companies found that 63% of them now use AI tools for the majority of their coding.

However, the adoption of AI in coding is not without challenges. Amazon, for instance, had to revise its rules after its AI coding tool caused outages, resulting in the loss of nearly 120,000 orders. These incidents highlight the potential risks associated with increased reliance on AI.

The rapid evolution of AI coding is profoundly transforming the role of software engineers, which was once seen as a stable and lucrative career. Some industry professionals report an increase in workload, sometimes leading to burnout. Naga concluded by stating that the role of the engineer is evolving, shifting from writing code to system architecture and reviewing code generated by AI.

Finally, Naga clarified that the best adoption of AI at Uber comes from engineers who experiment quietly, rather than from top-down imposition by management.