THE 90-DAY PLAN // FOUR WEEKLY FIGURES · THREE MILESTONES · FIVE RULES

Don't Take The Plan On Faith.

The book tells you to spend 90 days tracking your own AI mistakes, build a way to catch them, and show your boss the result. Here is that plan in full, before you spend a dollar: what gets logged, what you have on paper at each milestone, and the five rules every entry is written under.

THE FULL PROTOCOL · FOUR FIGURES A WEEK · A NULL RESULT COUNTS THE SAME AS A WIN

PLAN / AT A GLANCE

90 DAYS OF YOUR OWN DATA.

~25 HOURS OF WORK THREE MILESTONES · DAYS 30 / 60 / 90
4FIGURES LOGGED EVERY WEEK
3MILESTONES · DAYS 30 / 60 / 90
5RULES EVERY ENTRY IS LOGGED UNDER
01

What Gets Logged

FOUR FIGURES A WEEK · LOGGED AT THE TIME RATHER THAN RECONSTRUCTED

#Weekly measureCadenceWhere it lives
M1AI uses logged. Every AI-assisted output that enters the record.EVERY SUNDAYAI ERROR LOG
M2Share actually checked. The fraction of those outputs that got a real verification pass.EVERY SUNDAYAI ERROR LOG
M3Errors caught. What the checking found before it shipped.EVERY SUNDAYAI ERROR LOG
M4Minutes spent checking. The cost side of the ledger, in time.EVERY SUNDAYAI ERROR LOG
02

The Three Milestones

WHAT YOU HAVE ON PAPER AT EACH ONE

DAY 30
The Baseline

Your own logged error rate for AI-assisted work across thirty days.

DAY 60
The Control, Measured

One control designed, run, broken, and revised, with its time cost measured.

DAY 90
The Observed Multiple

A Judgment Multiple built from your observed inputs instead of modeled assumptions, with the full assumptions block visible.

03

The Five Rules

THE TERMS EVERY ENTRY IS LOGGED UNDER

RULE 01 · LOG THE MISSES

"A method that only reports wins is marketing."

The day-60 milestone is explicitly about what broke.

  1. Observed stays separate from modeled

    Anything measured gets OBSERVED. Anything derived stays MODELED. They never merge into one number.

  2. Redact by default

    Error classes, not employers. Nothing identifying any organization appears in the log.

  3. No retroactive filling

    Numbers reconstructed from memory stop being observed. Logged at the moment or not at all.

  4. If it fails, it counts anyway

    A null result is a finding, logged on the same schedule as a win. It is still a number you can defend.

04

The Tools In Use

NOT FREE DOWNLOADS · PRINTED IN THE BOOK OR EDITABLE IN THE $19 OPERATOR TOOLKIT

TOOL 01 AI Error Log

Where each AI use and each caught error gets recorded.

PRINTED IN THE BOOK · EDITABLE IN THE TOOLKIT
TOOL 02 Pre-Ship Verification Checklist

The check run before work ships with your name on it.

PRINTED IN THE BOOK · EDITABLE IN THE TOOLKIT
TOOL 03 90-Day Tracker

The week-by-week record of all ninety days.

EDITABLE IN THE $19 TOOLKIT

Two of the eleven operator tools are printed in full in the book; all eleven exist as working files in the $19 Operator Toolkit.

05

Run The Same 90 Days

THIS PAGE EXISTS SO YOU DON'T HAVE TO TAKE THE PLAN ON FAITH

This is the plan in the book, unchanged. Start your own baseline this week and in ninety days you will have your own data, measured against the same five rules, in the form a budget conversation can actually use.

INSTANT DOWNLOAD · EVERY UPDATED EDITION FREE TO EVERY BUYER

Or Just Follow The Evidence

One email when there's a verified result worth your time: an updated edition, a new control template, a tool you can use at work that week.

DOUBLE OPT-IN. CONFIRM VIA THE EMAIL YOU’LL RECEIVE.