Uber CTO says AI coding is yielding “astounding” results

AI For Business


Uber uses software to change the way people ride. We are now using AI to change the way we create software.

Uber CTO Praveen Neppalli Naga said in a LinkedIn post this week that 95% of the company’s engineers use AI every month, calling it a “true reset moment for engineering.”

He said the company has been “working hard” on AI coding over the past few months and “the results have been phenomenal.”

Naga said the biggest change will come from agent AI, where the software will complete tasks autonomously with little or no supervision.

At Uber, he said, this means engineers are increasingly delegating tasks to AI, not just accepting AI suggestions. As a result, 1,800 code changes per week were created “all” by Uber’s in-house coding agents. He added that the coding agent’s percentage of total coding changes went from less than 1% to 8%.

“There is no human authoring whatsoever. Engineers review and approve it, but the code is written entirely by AI agents,” Naga said.

It’s not just Uber. A recent survey of 700 companies by engineering intelligence platform Jellyfish found that 63% of companies surveyed now use AI tools for most of their coding.

Some companies are encouraging employees to use AI tools, such as incorporating them into performance reviews. But Naga said the “strongest hires” at Uber come from engineers who are “quietly experimenting” rather than being “imposed top-down from leadership.”

AI coding hasn’t always been smooth sailing for companies. Amazon installed new guardrails in response to series of power outagesSome of that was primarily caused by the company’s AI coding tools, which led to about 120,000 lost orders, Business Insider reported earlier this month.

Rapid advances in AI coding are disrupting software engineering, once considered a safe and well-paying career option. Some software engineers say that using AI tools actually makes them work harder, leading to burnout.

“The role of engineers is changing from writing every line to designing systems and reviewing code generated by AI,” Naga added.





Source link