You're a software engineer at Cogwheel, it's a Tuesday, and the company you work for is losing.
Thirteen years ago it was three people and a clever idea about how teams track their work. Today it's a few hundred engineers, thousands of paying customers, and a codebase old enough to have its own legends. Most of them are about the Core, the original monolith at the center of everything, that nobody fully understands anymore.
For most of those thirteen years, Cogwheel was the fast player in the market. That's the part that's gone...
Feature velocity has slid four quarters straight. The things your competitors ship in three weeks take Cogwheel a full quarter, if they ship at all. Customer sentiment is down two quarters running. And there's Northwind, the flagship account, with thousands of seats and a large slice of the revenue that keeps the lights on. Its VP of Engineering has just put four words in an email that everyone upstairs has now read: actively evaluating our alternatives.
This morning there's an all-hands, and the CTO says the thing nobody wanted said out loud. "A year ago," they tell the room, "we bought every engineer an AI subscription. Our bill went up. Our velocity didn't."
Nobody argues, because he is right. You felt it. You have the tools. You've used them. You've pasted an error into a chat and gotten unstuck, vibe-coded a prototype over a weekend that you never touched again. Some weeks it felt like magic. And yet the team as a whole moves almost as slowly as it did before.
Sitting in that meeting, you feel the two things at once that everyone in your position feels. The first is fear. The ground is shifting under your profession, the people who figure this out will be fine and the people who don't will not, and you genuinely don't know which one you are. The second is a quieter suspicion: that the tools aren't the problem. That the magic is real but nobody around you is holding it right. That there's an actual discipline, not a trick, that turns all this raw capability into something a team can ship, safely, under pressure, into a codebase as messy as yours. You just don't know what it is yet.
The all-hands ends with a mandate that is equal parts hope and threat: become an AI-native engineering organization, for real this time, and prove it by finally shipping the bet the company has been too slow to ship for two years. It has a name: it's called Pulse, and somehow it's landed on your team.
You walk back to your desk with no idea how you're going to do any of it.
This is where you are. It's probably close enough to where a lot of readers are that you didn't need the story to recognize it.
So before you go back to that Tuesday (and you will, in the last chapter of this book, where you'll watch this exact team dig all the way out), you do the one thing that turns out to matter.
You open this book, and you start to read.
— end of foreword —