Google’s CEO says that the most important metric they’re measuring right now regarding AI code generation is velocity.

He estimates that AI generated code is making the company 10% faster.

But, he still plans to hire more engineers next year because faster software development opens up more opportunities.

But does this mean product teams should necessarily build more stuff just because they can?

In this week’s Knowledge Series, I explore how AI coding agents work including SWE-bench tests, real world examples and the potential risks of AI feature slop: