The personal one. The technical version, with the definition and its consequences, is on daxfoundation.org.
Human delta value is not a new idea for me. It is my name for an old question, one I have been asking for a long time, long before Meta DAX, the companion or anything else this year: what is it to be human? What does a person bring that nothing else brings? It is the whole reason I am interested in anthropology, which exists to answer exactly that: what it means to be human. Human delta value is the part of the answer that can be kept in view and measured.
What is new is watching it show up in Meta DAX. On 27 September Meta DAX ran end to end for the first time. A model played the teacher and a simulated adult played the learner, so nothing about people was proven that day, and I have said so everywhere the results are published. What caught me when I read the report was not the physics.
It was the teacher’s workflow. Twelve stages: know the learner, pick a yardstick, set the destination, draw the map, fence the scope, choose the voice, plan the misconceptions, design the page, build the labs, generate and check, hand over, review and loop. The yardstick was an outcomes rubric, picked from a catalogue, and the catalogue grows every time a teacher publishes one. The lesson plan arrived defaulted. Every one of the teacher’s decisions had become a micro-adjustment to something that was already there.
I read that and recognised it: the old question, sitting in the first report, uninvited. For the record, and dated: the term was on the Foundation’s definitions page on 21 September, six days before that run. The report did not give me the idea. It showed me where the idea lands in teaching.
What it is
Human delta value is what a human being brings: the part of an outcome that exists because that person was involved. It has two halves.
The first is what only a human can currently do. The word currently is doing real work. That boundary moves, and anyone who anchors their sense of worth to it is choosing to retreat for the rest of their life, one capability at a time.
The second is what a person and their cognitive companion reach together that neither would reach alone. A person and their companion are one unit. The second half grows as the companion grows, provided the person brings their best, and it is the durable one. The better the companion gets, the more your best is worth. Not less.
Reclaimed cognition
Here is what the report showed me about the first half, and why the usual way of measuring it is wrong.
When the lesson plan arrives defaulted and the rubric comes from a catalogue, a great deal of cognitive labour has been taken off the teacher. The lazy measure of that is time: hours saved. I refuse to measure it by time, because time saved says nothing about what happened next, and what happened next is the whole point.
What matters is what the teacher spends the cognition on now that it is theirs again. I call that reclaimed cognition, and it goes on the things that intelligence, in the Foundation’s narrow sense of navigating a constrained possibility space, is not. Noticing that one learner has gone quiet, and deciding what to do about it. Reading the dynamic of a group. Knowing that this particular person needs to be pushed and that one needs to be left alone for a week. Bringing out the best in someone. No model in the room does any of that; the teacher does, if they have the capacity left to do it. The system takes the busywork so that they do.
And it has to be reported by the person, in their own words, and recorded as they said it. Never inferred, never turned into minutes. “The sequencing and the quiz were off me; I spent the evening watching him, not the page.” That sentence is the measurement.
Inverting the ratio
Meta DAX is built to invert the ratio of students to teachers. Today it is one teacher to thirty learners, or one to three hundred, and the teacher’s attention is the scarcest thing in the building. Meta DAX turns that over: one learner, and behind them every teacher’s move that was ever captured in the record, every misconception ever mapped, every dead end anyone ever found first, offered as one way among many and never as the way. The learner gets more teaching than any teacher could give. And the teacher gets their cognition back.
I want to be explicit about the second part, because it would be easy to read it as a side effect. It is not. We are doing it on purpose. We are deliberately giving teachers tools, and the opportunities that come with them, so they can put their cognition on problems they could previously only dream about: the learner who was going to be lost, the subject nobody has taught yet, the move that works for one person and then, through the record, for everyone after them. Human delta value is not a one-off measurement we take at the end of an experiment. It is the thread through all of them, and it is close to being the point.
Everybody needs to step up
This is where it stops being comfortable.
The unit is only as strong as the human’s best in the moment. However capable the companion becomes, a person who cannot bring their best gets less from it, not more. The companion does not fill the gap. It amplifies whatever is there, and if what is there is tired, distracted or absent, that is what gets amplified.
So this is a demand, not a reassurance. The teacher whose plan arrives defaulted is not being let off. They are being handed the hardest part of the job, undiluted, with no busywork left to hide in. That is true of every person working as one unit with a companion, and it is true of me. The honest reading of “the better the companion, the more your best is worth” is that everybody needs to step up, and that the step is now visible, because the record will show what you brought.
Why it is in the constraint
The DAX constraint is preservation and expansion of life, humanity, and consciousness. Humanity is named in it, and every human being is part of it. Human delta value is how that word reaches one person at a time: it is the measure that says whether a system is expanding what a person can do or quietly replacing them, and a system that cannot say what the human brought has stopped extending people and started replacing them. It has to be kept in view and measured continuously, in every process, implementation, automation and optimisation a person takes part in. Not once, at the end. Continuously.
That is the whole post. Twelve stages, a defaulted plan, a rubric from a catalogue that grows, and the one thing that was always going to be the point: what the teacher does with the cognition they got back, and whether they bring their best to it.
Versioned like everything else here; the History is public. The definition, with its consequences and its failure modes, is the technical piece. First published 1 October 2026.