The Engineer's Anxiety at the Penalty Kick

Picture a goalkeeper facing ninety penalties in ninety minutes. One a minute, the entire match.

1. 

   Penalty 1 — Minute 01

   The first one, he is everything a keeper is supposed to be. Full horizontal stretch, fingertips grazing the ball, beaten by centimeters.

   The miss doesn't matter. The effort is total. Every cell of him is in the moment of that kick.

2. 

   Penalty 4 — Minute 04

   By the fourth, the heroics have settled into craft. He reads the run-up, picks the corner, gets there, and the ball sticks in his gloves.

   This is competence working as designed. Attention, skill, and a decision that matters because there aren't many of them.

3. 

   Penalty 48 — Minute 48

   But there are many of them. By the forty-eighth, he stands on his line and doesn't move. His body is in the goal. The rest of him is somewhere else. The kick happens near him, like weather.

4. 

   Penalty 73 — Minute 73

   By the seventy-third he has left the goal entirely. He leans against the post from the outside, smoking, watching the balls go in the way you'd watch a neighbor's sprinkler.

5. 

   Penalty 89 — Minute 89

   And the eighty-ninth finds him at home, asleep in bed, the goal standing empty behind the kicker.

Nobody saves the 89th penalty. The kicks never got harder. Each one is exactly as saveable as the first. Nobody is built to mean the eighty-ninth decision of the hour the way they meant the first.

This is my working day now. Agentic coding multiplied how many decisions cross my desk in a day, and that volume is what wears comprehension down. A migration here, a refactor there, a dependency bump, a schema change, each diff arriving minutes after the last, each waiting for the same verdict. Approve or reject. I used to make a handful of consequential calls in a day and could feel the weight of each one. Now the calls arrive on a conveyor, and the conveyor doesn’t care which number I’m on.

What concerns me is when I find myself approving whatever without realising the decision nor the stakes at hand.

If the model can do everything I can do, where’s the line between us?

It’s a question of position. The model can generate a decision. It can’t be the place the decision lands - it can’t bear the consequence, hold the liability, be the person an organization trusts, or want one outcome over another for reasons of its own.

Then what’s my job, if not to produce the work?

To own it. And ownership without comprehension is just a signature. Hands-on practice is the only way I know to build the comprehension that gives my approval any weight. The easy version of this fear is being out-skilled. Mine is quieter - that I keep signing while I stop understanding what I sign.

Suppose the agent ships a migration that corrupts content in production. What happens to the agent? Nothing. Its context window closes and it ceases to exist in any meaningful sense. There’s no Tuesday morning for it. What happens to you? You’re in the incident channel, your name is on the merge, your credibility with the leadership you just won over takes the hit, your one-month deadline track record gets an asterisk. The consequences of the decision have somewhere to land - a continuous person with a reputation, a salary, a future at Wise, anxiety at 2am. The model generates the decision; you absorb its results.

So what do I do about it?

The trap is that disengagement doesn’t feel like disengagement. You steer the agent all day, your hands never leave the keyboard, and the motion feels like work. You can be busy approving and understand none of what you approve. Nothing in the loop stops to ask whether you still follow what you’re signing.

So the practice has to be deliberate - the comprehension won’t rebuild itself by accident. What that looks like for engineering is the question moo.md is for.

The real risk is to the practice that keeps my accountability honest.