We're not meant to build alone

By 6 PM last Tuesday I was more drained than I'd been in months.

Five hours into debugging a ranking regression that I was sure was going to sink the whole release. Model checked out. Features checked out. Training data checked out. And yet the offline eval kept telling me the new model was worse.

I was alone on it because I'd chosen to be. I had the whole problem in my head. I didn't want to explain it to anyone. I didn't want to hear someone else's guess. I wanted to solve it by myself, because solving it alone is what felt like real work.

This is a version of myself I know well. I can spot her from the other side of the room. She usually gets me nowhere.

So I walked down the hall, pulled a chair up next to my manager, and said: "I don't think I can get this before Friday."

He asked four questions. Two I'd already answered. One I didn't know. The last one was the one.

Ten minutes. Five hours wasted on the wrong hypothesis because I was protecting the illusion of individual genius.

The lesson, for the 400th time: we're not meant to build alone.

Ask earlier. You'll still be the one who ships it.

“Turns out the real breakthrough was human.”