Three forfeits in three days. The contract was right. The week was wrong.

Yusuf had been tracking his phone pickups since the previous quarter, when his iPhone's Screen Time summary landed on his home screen during a meeting and showed 143 unlocks by 3pm. He had not been aware of the number before he saw it. He was aware of it afterward in the way you are aware of a word after someone defines it — it named something he had been doing without naming it himself. The pickups were distributed throughout the day: a check after sending a Slack message, a check before a call, a check at the elevator, a check at the coffee machine, a short check that led to a slightly longer check when something in the notification shade was interesting enough to open. None of the individual checks were long. The Screen Time data showed that most pickups lasted under thirty seconds. What they added up to was a texture of fragmentation that ran across the entire day and into the evening — a persistent low-level monitoring posture that never fully released. He could sit through a conversation and be present for it while also being dimly aware that he had not checked his phone in eleven minutes, and that eleven minutes was longer than usual. The awareness itself was the problem. He was not addicted to the content. He was addicted to the checking. The distinction mattered because content addiction responds to content restriction, and he had already tried turning off most notifications. The pickups had not changed. The checking was the behavior, not the notifications, and he had been feeding it for two years without a contract that named it.

Sunday, 11:04pm. A Stripe integration that had been tested in staging all week broke in the production environment two hours before the launch window. The fintech's API team was in a thread on Slack. Yusuf was coordinating between product, engineering, and the external partner. His phone was in his hand continuously from 10:45pm until past midnight. He did not count pickups. He was not thinking about pickups. The Screen Time counter was running regardless. Monday morning, he opened the Lockin dashboard before his standup. Sunday had logged 94 pickups. Technically not a contract day — his contract ran Monday through Sunday. He noted the number and moved on. The Stripe issue had been resolved at 1:17am. Monday proper: customer escalation from the partnership team, a support queue that had caught three edge cases in the new flow, and a regulatory inquiry that landed in his inbox at 4pm from a compliance officer he had not spoken to in two months. Fifty pickups by noon. He saw the count at 11:52am and understood what it meant. The Lockin dashboard showed 48 when he checked it. Two more and the forfeit locked. He put the phone on his desk face-down and did not pick it up for eleven minutes, which was the longest stretch of the day. Then a Slack notification came through on his laptop and he picked up the phone to respond faster. Fifty-one. Tuesday, 9:50am. He opened the dashboard at his desk with a coffee. The counter showed 50. He had not started his second meeting of the day. He had an all-hands at 10, a partner call at 11:30, an afternoon of review sessions for the post-launch metrics deck. The forfeit was already locked before he had done any of the day's work. He kept checking the phone. The count went to 187 by 10pm. A second forfeit. Wednesday, the same. A third forfeit — 142 pickups, the counter climbing steadily from the first Slack thread of the morning. He had stopped watching the dashboard by midday. Watching the dashboard required picking up the phone. On Thursday morning, before his first meeting, he opened the app and ended the contract early.

Yusuf did not close the Lockin account. He started a new contract the following Monday with a revised structure. The new limit was 70 pickups per day — up from 50 — with a higher daily stake of $7.50 instead of $5. The higher stake was the cost of the higher ceiling: the limit was less aggressive, so the per-day pressure had to be heavier. The more significant change was outside the app, in his calendar. He kept a running list of launch weeks for the rest of the year — feature releases with downstream dependencies, regulatory deadlines, partner integrations going live. For each one he set a recurring reminder to end the active contract on the Friday before launch week and start a fresh contract the Monday after. The contract did not have a launch-week mode. He had to be the mode himself. The new contract ran cleanly through its term, and he renewed it back-to-back across the sprints that followed. His daily counts settled between 55 and 68. He took one launch week off the contract entirely — a feature release that had downstream dependencies with two external partners — and rejoined the rhythm the Monday after. The contract had become something the original one had not been: durable across the full range of the work cycle rather than calibrated only to the portion of it that looked like a steady state. Launch weeks were part of the calendar. A commitment device that pretended otherwise was a habit tracker for ordinary time, not a contract.