KorvaJoin the waitlist
Startups

A Small Group of Engineers Ships Far More Code Than the Rest

The highest-output engineers were not necessarily spending more on AI.

AI image
All-In with Chamath, Jason, Sacks & Friedberg · Databricks CEO: AGI Is Already Here, Why Companies Are Behind, and the Superintelligence Debate

At Databricks, code output among about 2,000 engineers forms an L shape, with a small top group shipping far more code into production than everyone else.

Databricks CEO Ali Ghodsi said the company investigated the pattern using its internal AI tool. The top performers were running automated factories, meaning AI tools drafted, tested, and prepared code with little hand typing while a person still checked and approved it.

The measure was pull requests, batches of proposed code submitted for review before going live, counted on a roughly weekly basis. So the chart showed who was getting the most finished work through the door.

The surprise was cost. Some of the highest-output engineers were not spending more on AI. Ghodsi said the trick was spending smartly on tokens, the units AI models charge by, including using open-source models and routing simpler questions to simpler, cheaper models.

Questions people ask

Does shipping more code mean someone is a better engineer?

Not necessarily. More shipped code can mean faster progress, but it can also mean larger, sloppier, or duplicative changes. Managers usually pair output measures with review results, bugs, and whether the code actually solves a customer problem.

How can higher AI output cost less?

Because AI models charge by use, often priced in tokens, or chunks of text processed. Sending easy tasks to smaller, cheaper models, including open-source ones, can produce a lot of work for little money, while sending everything to the most powerful model gets expensive fast.

More in Startups