Can your team still own the contract when AI writes the middle?

Generation got cheap. Review and contract ownership did not. This is the Monday scorecard from Skin in the Runtime. Run it in 30 minutes with whoever owns delivery.

Score 1–5 on each line. 1 = we don’t know. 5 = we can prove it this week.

1. Named owner of the ask (the contract), not just the PR
2. Named owner of the outcome after merge
3. Review capacity vs. generation volume this month
4. Contract review vs. style review (we can tell them apart)
5. Time-to-owned-outcome, not lines / PRs / acceptance rate
6. Named owner for agent / tool sprawl
7. Named owner for token / inference spend
8. Last AI-heavy change: the team can explain intent, boundaries, invariants, and failure modes

Scoring

• 32–40: you have a system
• 20–31: generation is ahead of ownership
• Below 20: you are accumulating cognitive debt

The 30-minute facilitation

• 5 min: each person scores silently
• 15 min: pick the two lowest rows
• 10 min: name an owner and one gate for this week

This page is the argument. The pack below is the paper you mark up: this scorecard with score boxes, the runtime judgment test, and one gate per row, the single check to install for each claim this week. It is emailed after one confirm click.

If you want the diagnosis behind a row, start here: the review bottleneck, cognitive debt, the token bill, agent sprawl.

Skin in the Runtime is the diagnosis: where the constraint moved when generation got cheap. Contract Over Code is the operating model: two contracts, staffed hours, gates that recover senior time, a ninety-day install. Reading either with a team — bulk copies or a working session — is on the book pages. A decision where a second set of eyes would change the outcome: get in touch.