The cost of saying yes has changed

The GitHub Blog reports on a fundamental shift in software engineering: the expensive part of a small feature request used to be building it. Now it's often the argument about whether to build it at all. AI changes this equation. An agent can produce a working patch while the team is still debating scope. But here's where teams stumble: generated code is a starting point, not a solution. The patch becomes evidence—you can run it, test it, examine whether it touches five packages or four lines. A decision that would've taken two days of debate happens in thirty minutes when you're staring at actual code. The critical mistake is treating fast code as final code. Authorization logic, data-retention rules, privacy decisions remain expensive to own no matter how quickly they're written. Real cheapness requires both easy writing and easy ownership. The shift means moving the scope question from planning to review: ask for a constrained attempt, and let the patch tell you whether the request is truly small or just sounded small. The skill that separates good decisions from bad ones becomes pricing uncertainty—understanding when to try something cheap and when to say no. The economics of saying yes have changed, and the economics of saying no should change accordingly.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton