Proximity is calibration

What the paper argues

Staying close to the work is a control, in the sense a regulator would recognise, not nostalgia, a productivity choice or a personality trait. As AI moved the work from producing to directing, every layer moved up one: the engineer from writing the change to reviewing a generated one, the lead to reviewing the review, the manager to approving a summary, the director to reading a dashboard. Each move is defensible on its own. Stacked, they put every approval one layer further from its artefact, and because nobody decided it, nobody assessed what the distance costs.

The mechanism is cognitive. Verification is a different task from creation: writing something builds understanding as a by-product, inheriting it does not. Proximity was how people stayed qualified to judge output, and without it an organisation holds a chain of approvals in which no single link is independently calibrated. Every signature is real and given in good faith. The aggregate assurance is still theatre.

Key points

• Rule 01: your approval is worth the distance between you and the thing you approved.
• The paper weighs its own evidence. Sonar’s 2026 survey puts AI at 42% of committed code, and finds 96% of developers do not fully trust AI output while only 48% always check it; the paper treats these vendor figures as directional. DORA’s qualitative study is used because its mechanism is structural.
• Measured versus governing: the review happened, against whether the reviewer could have caught it; the gate was passed, against whether the gate has refused something.
• Article 26(2) of the EU AI Act requires deployers of high-risk systems to assign oversight to people with the necessary competence, training and authority. The AI Omnibus moved enforcement dates; Article 26 itself is unchanged. Anchor to the obligation, not the deadline.
• A five-question proximity test, runnable without instrumentation, opening with when you last read an artefact rather than a report about one, and asking who is being trained to judge.
• The hardest objection is stated plainly: a hands-on senior person can become the bottleneck and take the work that would have grown their team. The only honest test is directional, whether you are taking work off your team or sitting alongside it.