1,400 Hours

Sometime last week, a counter I keep crossed 1,400 hours. That is my running estimate of time saved using AI tools since early March, tracked weekly, one line per task, on a public page anyone can read. Seventeen weeks. 1,420 hours. If you believe the number, that is thirty-five standard work weeks, roughly eight months of full-time labor, compressed into four months of actual calendar.

Here is the thing: I don’t fully believe the number either. And the reason I don’t believe it is more interesting than the number itself.

How the number gets made

Every week I publish an AI This Week log. Each entry is one task: what was done, which tool did it, what came out the other end, and a conservative estimate of what it would have cost me to do by hand. A quick lookup counts for ten or twenty minutes. Drafting a document counts for an hour or two. A deep research report with dozens of verified sources counts for eight to twelve hours, which is honest, because I have written those the old way and eight hours is generous.

The estimates are deliberately boring. No task gets credit for “strategic value” or “improved decision quality.” If I couldn’t picture the hours on a timesheet, they don’t go in.

Seventeen weeks of that discipline still produced 1,420 hours. The weekly average is about 84 hours. The slowest week logged 28. The biggest logged 146.

So the honest reading of 1,400 hours is not “Jason got faster.” It is this: most of that work would simply never have existed.

Nobody was ever going to spend eight hours writing a 78-reference literature review to strengthen a grant application. The grant application would have gone out with a paragraph of citations and crossed fingers. Nobody was going to hand-verify the state’s regional teacher salary rankings against every district’s actual published schedule. We would have taken the state’s number at face value, which would have been a mistake, because when I did check it last week, the state had understated Greeneville’s average by nearly $2,000. The rankings were wrong and no human was ever going to find out.

That salary check took maybe an hour of my attention. I logged it as three hours saved. The real value is not three hours. The real value is that the check happened at all.

After seventeen weeks I can sort everything on the log into three piles.

The first pile is genuine time savings, the timesheet kind. Filing monthly finance reports. Reconciling nutrition program figures. Processing insurance renewals. Grading with a rubric. Building travel expense forms. This is real work that had to happen either way, and it now takes minutes instead of afternoons. This pile is maybe a third of the log, and it is the pile that gave me my evenings back.

The second pile is work that got dramatically better because the cost of thoroughness collapsed. My dissertation committee reviews used to be a read-through and a page of notes. Now a student gets 82 margin comments and a prioritized memo. The document is better. The student is better served. Same task, ten times the depth.

The third pile is the counterfactual pile, and it is the biggest: 68 published research reports, a public operations dashboard, a multi-year accountability dashboard a colleague can now deploy without me, automations that watch fifteen district websites for salary schedule postings every Monday at 7 a.m. None of that was on any to-do list in February. There was no version of my life where it happened by hand.

Somewhere around week eight, I stopped thinking of the counter as time saved. It is closer to a ledger of work that used to be impossible at my scale, one line at a time, 1,420 hours and counting. The counter is public. Watch it, or better, start your own.