This edition reads one thing: what your organization can now attempt, carry and finish. It is for a company already using AI that can do the work your people were trained to do.
It is called AGI POSITIVE. It reads what you have already attempted, carried and finished.
AGI READY™ is the standard underneath it: six areas, a set of weights, four bands and one rule about evidence. This edition points it at one question: what could you attempt now that you could not eighteen months ago?
In each of the six areas the check asks the same thing. How much can one of your people now carry from beginning to end? And what had to be there for them to carry it? So it counts attempts made, work one person carried end to end, and a record a machine can work from.
The six areas are on The Six, the reading on Your Index.
What to do this week. Take the check yourself rather than delegate it: twenty minutes for all thirty-six questions. Print it and write the date at the top. You end up with your first dated reading, and every later one is measured against it.
That belief is what the six areas are for.
An amplifier makes a signal louder. It does not decide whether the signal was worth hearing. These machines do the same to an organization: good work gets louder, and so does everything nobody wrote down.
An organization gets more out of these machines in proportion to what it has written down and who will sign for it.
Readiness here is how that record and that signature get built. It means one person choosing the work, a second checking it, a third signing. It means writing down what you will not build at any price. And it means the nerve to attempt what you would once have declined.
What to do this week. Ask each of your leaders for one project their area declined in the last two years, and why it was declined. Thirty minutes, in a meeting you already hold. You end up with the first page of an attempt list. It is finished when every area has named one and the page carries a date.
Four rules hold this honest. You can hold us to all four.
The first rule carries the others, because self-report here is measurably unreliable. In a randomized trial, sixteen experienced open-source developers were given real tasks. The tools were allowed on some of those tasks, at random, and not on others. They forecast a 24 percent speedup. The clock recorded them 19 percent slower. Afterwards they still believed the tools had made them 20 percent faster. That is a thirty-nine point gap between belief and measurement, among people who do this work for a living. Evidence
A reading built on completed acts is worth an afternoon.
The trial in full: sixteen experienced open-source developers, real tasks, randomized task by task, timed rather than surveyed. The same researchers announced on 24 February 2026 that they were redesigning the study, because developers now insist on keeping the tools. This page carries the trial with its design, its numbers and its dates. Every shorter mention of it elsewhere on this site is this same trial.
What to do this week. Take one claim your organization makes about AI saving time. Ask the person who owns the slide three things: who counted it, on what work, and on what date. Twenty minutes. You end up either with a measured number or with a known gap, and you know which one you have.
Three kinds of note, printed where the fact is. Two mark other people’s work; the third marks ours.
What to do this week. Give half an hour to your most skeptical reader: pick any three facts here and find the note behind each. Send us any one that has no note. You end up knowing what the rest of the page is worth.
One rule keeps them apart, for two years after any help we gave you.
The rule has a price and we pay it. Where we have helped build inside a part of your business, we cannot read that part for two years. The fee for reading it goes to somebody outside our team.
What to do this week. Ask us in one email for the panel and the date of the last change to the weights. A written reply comes back within two working days. You end up with both dates on a page you can keep and hold us to.
Which one you need depends on who is asking.
| What you get | Who decides it | What it carries | Who may rely on it |
|---|---|---|---|
| Your own dated reading | You | A reading from 0 to 100, six area scores, three moves | You alone |
| AGI READY VERIFIED | Our readers, against your documents | What was checked, in writing, on a stated date | You, and anyone you show it to |
| AGI READY CERTIFIED | A panel, none of whom advised you | A seal with a number, a date and the band reached | Anyone. It is the only one published |
The panel invents no level. It states which band the evidence reaches, and grants nothing below the third.
We sell the report, the reviews, the ninety-day reading, the certificate and help keeping the record current, all priced on the pricing page. We sell no software and grade nobody on your staff.
What to do this week. Ask whoever keeps sending you questions which of the three they want, and ask for the answer in writing. One email. Their written answer picks the door on the services page.
Many instruments score an organization on how much of the work would still get done if these tools were removed. Each of the three is met here by an instrument built for it.
What that older reading cannot do is tell those three apart. It also cannot tell a person freed from drudgery from a person who never had to struggle. It returns one number for both, and its highest score belongs to whoever changed least.
So the reading here counts the other direction. What did you attempt this quarter that you would have declined eighteen months ago? Who carried it, how many people did that work need before, what did it cost, and what made the next attempt cheaper? Each answer is a fact about something that already happened.
What to do this week. Find the last paper that scored you on what would survive if these tools were removed. Write the second question above in its margin and hand it to whoever owns that paper. Ten minutes. You end up with one page asking what you attempted, and it is done when an answer is written under it.