A Theory of Embedded Intelligence Essay A Plan for TEI Education Across a Whole Life — Before Birth, PreK–12, the Working Years, and Fifty-Five and Beyond — on the Platforms, Programs, and Boards That Already Exist

The cycle starts before birth and does not stop when the paycheck starts. Montessori built a room for it. The Apple II built a bus for it. FIRST LEGO League, Arduino, Raspberry Pi, and the FPGA boards from Terasic, Texas Instruments, AMD-Xilinx, and Trenz have been running it in classrooms every year the 65xx was away. This revision carries the same four questions across six bands of a whole life — on hardware nobody has to buy from me, toward an understanding system that never has to empty the classroom of life to get there.

Editor’s Note This essay introduces a plan. It does not announce a program that has launched. No school, platform, robotics program, board supplier, or institute named below has agreed to anything; each is named because it already operates a channel the plan would use, or already sells hardware that runs a complete SPCA cycle. Facts about those programs and products are reported as they stood in September 2026 and should be rechecked before any partner conversation. Budget figures and board pricing are deliberately absent. They belong to the Foundation and to the Western Design Center, and they will be published separately rather than estimated here. Revision 4, September 2026. Four changes. A figure in Section VIII is withdrawn and replaced, with the withdrawal printed rather than quietly corrected. The section numbering is repaired — Revision 3 skipped from VIII to X, so the sections after the six bands each move down one. Section XII now records that every essay it proposed has been written. And a new Section XIII has been added: a reading map for mentors, guidance personnel, teachers, leaders and administrators, which is the thing this plan most obviously lacked. Revision 3 added the two bands that earlier revisions missed: the months before birth, and the working years between university and fifty-five. Section IX, on FREE WILL and Lifelong Learning Classrooms, was new in that revision, and its acronyms originated with Bill Mensch, not with the drafting assistant. The naming cautions in that section are practical, not rhetorical. Nothing in this essay is adopted into the TEI canon. The rules proposed for the plan — including the rule that every AI tool faces the adult below age six, and the rule that no unit may depend on one supplier’s hardware — are application-level design decisions, and Section XII says why they stay that way until a human rules on them.

I. Two Ends of One Life

In 1979 I was one of a handful of parents who started a Montessori school in a house in Mesa. I was born in 1945, and today I live in a retirement community on a university campus in Tempe, among people who have been learning for seventy and eighty years and have not stopped. Between those two addresses lie forty-seven years, eight grandchildren, and a theory. This essay is about the two addresses.

The earlier essay Montessori Education Meets SPCA ended on a claim: every child is already an embedded intelligence running a complete cycle, and the work of a school is not to install one but to keep it open. The same essay made a second claim almost in passing — that a five-year-old and a fifty-year-old ask the same four questions of a thermostat and get different, and both correct, answers. I want to take that second claim literally. If the questions do not change with age, then education in the Theory of Embedded Intelligence has no graduation. It has a cycle that starts before school and does not end at retirement.

The two ends are where that cycle is most exposed. At one end it is under construction: a Sense phase still learning what matters, calibrating on whatever voices it trusts. At the other end it carries seven decades of formation and is meeting, for the first time, a fluent machine that answers back. The middle — the university — is where the Foundation already works: the Mensch Prizes, the seminar and bench courses EIE 101 and EIT 101, and the department-by-department map in The Curriculum Already Exists. The plan is to extend that work in both directions at once.

II. The Plan on One Page

An engineer should be able to hold a plan on one page. Here it is. Everything after this section is the reasoning and the test.

III. Why We Will Not Build a Platform

The market the plan needs has already been built, and it has already put AI in front of both ends of the lifespan. Khan Academy’s Khanmigo tutor runs inside school district deployments. In the space of a year, the three largest AI laboratories each shipped a dedicated learning mode — Anthropic’s Learning Mode in April 2025, OpenAI’s ChatGPT Study Mode that July, Google’s Gemini Guided Learning that August — and by 2026 all three were effectively free to students. At the other end, GetSetUp runs live classes for older adults taught largely by their peers, with an AI assistant built into the platform. Senior Planet from AARP teaches free introductory classes on AI in libraries and community centers. More than a hundred and twenty Osher Lifelong Learning Institutes operate on college campuses in every state for learners over fifty, with no tests and no grades.

A foundation that built its own platform would be competing for the attention of those same learners with a fraction of the money, and it would spend itself on the part of the problem where it has nothing distinctive to offer. What the Foundation has that none of these channels has is the framework and the people who know how to run it.

There is also a structural reason, and it is the one that decides the matter. The Medium Separation (CKB-11) holds that a governor in the same medium as the governed inherits that medium’s mutability. If the TEI material lived inside a vendor’s prompt, the vendor could revise it — not maliciously, simply in the ordinary course of shipping software. So the plan keeps the framework where it already lives: in the Foundation’s published record, in the portable AI Master System file, in essays that stay as issued. Any platform can load that material. None can quietly edit it. The platform supplies the runtime. The Foundation supplies what is fixed before runtime.

The same reasoning applies to hardware, and Section VI is where I take it out on my own company.

IV. The Tutor Nobody Called

Before deciding what to put on those platforms, it is worth looking hard at the best evidence available on what happens when an AI tutor is simply made available. That evidence arrived last month.

On August 17, 2026, the National Bureau of Economic Research circulated a two-year randomized trial of Khanmigo across eighteen middle schools in Hamilton County, Tennessee, by Philip Oreopoulos and Nina Low. Assignment to the AI tutor raised mathematics achievement by about 1.3 national percentile ranks per term — gains the authors describe as similar to those from Khan Academy practice without AI at all. Nearly every student tried the tutor. The typical student then rarely used it, and when students did, they seldom engaged it in real mathematical dialogue. The authors name engagement, not model capability, as the binding constraint. Khan Academy, to its credit, has been candid about uneven use and redesigned the student experience so that the tutor activates during practice instead of waiting to be called.

One study of middle-school mathematics in one district settles nothing by itself, and this essay does not treat it as settled. But the shape of the result is exactly what TEI predicts, and it is worth saying why.

A tutor waiting in a side panel is a destination for the second Communicate phase — the external, asymmetric one. Nothing reaches it unless the learner’s own cycle has something to send. A learner who has not sensed a problem has nothing to process, and a learner with nothing to process has nothing to communicate. Sense is the narrow end of the loop. Placing a more capable Process engine downstream of an idle Sense phase adds almost nothing, however good the engine is.

Montessori solved this problem in 1907 without the vocabulary. The control of error is in the material: the cylinder does not fit, and now the child has a real question. A robot that misses the mission does the same thing. A lamp that does not light does the same thing. Then the learner has something worth asking anyone — a teammate, a coach, a teacher, or an AI. So the plan fixes an order for every band. The occasion comes first. The material comes second. The human comes third. The AI comes fourth, and only after the learner has something of their own to bring to it.

The same result is a warning to the plan itself. Free material that nobody opens has the same engagement problem as a tutor nobody calls. Section XI makes that a falsification condition rather than a hope.

V. The Bus Was Always Open

Here is the part of this plan that required me to argue against my own inventory, so I will put it early and state it plainly.

The Apple II was a 6502 machine with a row of numbered expansion slots on a documented bus. You could plug a card in. The card could sense something, or drive something, or talk to something else, and the machine’s manuals told you exactly how. That architecture is the reason a generation of us learned what a computer actually does, and it is worth being precise about what the slots really were: a Sense port, a Communicate path, and an Actuate port, exposed to a fifteen-year-old with a screwdriver. The processor was mine. The idea that the bus should be open was Steve Wozniak’s, and it was the better half of the machine.

The 65xx then left the classroom for thirty or forty years. The classroom did not stop running SPCA cycles while it was gone. It ran them on everything else.

Every one of those runs a complete cycle, and every one of them can be read with the four questions the day it comes out of the box. That is the finding that shaped this plan. TEI education does not need a particular chip, and it does not need mine. What it needs is a system the learner can break and diagnose — and the classroom has been full of those for forty years.

VI. What WDC Is Actually Offering, and Why It Is Not Required

The Western Design Center makes the case for learning with the 65xx on its own site, and the case is a real one. The instruction set, memory architecture, and processor behavior are fully documented; the student works directly with registers, memory, and memory-mapped I/O rather than through an operating system or a library; the architecture has stayed backward-compatible for more than forty years while classroom platforms came and went. WDC contrasts that with a Raspberry Pi built on a proprietary system-on-chip with closed firmware, and with Arduino boards whose underlying parts vary by vendor and are reached through libraries. The SXB system boards with the EDU add-on boards, and the MySPCA kits being developed around them, are WDC’s attempt to put that path back in front of students.

Now the honest half, and it is mine to say because it is my company.

There is no real need to return the 65xx to the classroom after thirty or forty years outside it. Students learned. Engineers were formed. Robots drove. None of it waited on me.

What WDC offers is a different view of the same cycle — the view with nothing in between — and a claim about stability that I believe is true and that a school is entitled to test against its own experience. That is an argument worth making on its merits, and it is the argument WDC should make. It is not an argument this Foundation gets to make by building a curriculum that only runs on WDC parts.

So the plan carries a hardware-neutrality rule, and it is not a courtesy. It is what keeps the money force off the material.

VII. FIRST LEGO League, and the Robot That Argues Back

If you wanted to design a program that ran the SPCA cycle in front of nine-year-olds and made the control of error impossible to fake, you would build something close to FIRST LEGO League. FIRST and LEGO Education have partnered for nearly three decades. Explore teams of six- to ten-year-olds build and code with SPIKE Essential; Challenge teams of nine- to sixteen-year-olds program a robot to complete missions autonomously in two-and-a-half-minute matches, and research a real-world problem in the Innovation Project alongside it.

Read the robot game as a cycle and it is exact. The robot senses — color, distance, angle, touch. It processes — the students’ own code, which is the only place their thinking can go. It communicates — from the sensor read to the motor command, the first C, internal and constitutive. It actuates — and the mission model either moves or does not. Nobody grades it. The field does. That is the knobbed cylinder, at speed, with a timer running.

The Innovation Project is the second C, and this is the part coaches undervalue. The team must take what it understood and render it for people who were not in the room. The canon’s Human-Renderability Constraint says that a result which cannot be rendered into a form a human can hold cannot be governed by humans. A nine-year-old learning to explain a solution to a judge is practicing exactly that, and the judging rubric is one of the few places in American education where the rendering is scored separately from the result.

Two contributions the plan can make here, neither of which requires FIRST or LEGO Education to change anything:

A coach’s sheet with the four questions. When the robot fails, the break is in one of four places, and most teams debug by changing the program. Did the sensor read at all? Was the reading interpreted correctly? Did the decision reach the motor? Did the motor do anything? A team that runs those four questions in order finds the fault faster and — more to the point — carries the diagnostic away from the mat. That transfer is the entire claim of TEI education, and a robotics season is an unusually good place to test it.

The season as a case in governance. The 2026–27 season, BIOGLOW, kicked off on August 4, 2026 in two editions: a Founders Edition on SPIKE technology, whose final season is 2027–28, and a Future Edition built on LEGO Education’s new Computer Science & AI hardware. After that transition, the new hardware is required. I say this with no criticism of anyone: a curriculum built on one supplier’s kit inherits that supplier’s roadmap. That is the Medium Separation (CKB-11) in a classroom purchasing cycle, and it is the best possible argument for writing units against functions rather than part numbers — including, and especially, against mine.

VIII. Six Bands, One Set of Questions

The spine of every band is the same four questions, carried over unchanged from the Montessori essay. Only the systems they are asked about, and the depth of the answer, change with age.

Before birth: the cycle that assembles itself

The plan’s earlier revisions began at three. That was too late by about nine months, and the omission mattered more than a missing line item, because this band is where the framework’s central claim is least metaphorical.

What is established is modest and solid. The fetal auditory system becomes functional in the third trimester, with low frequencies passing best through fluid and tissue. Newborns prefer their mother’s voice to a stranger’s, and a story their mother read aloud repeatedly in late gestation to a story she did not. Near-term fetuses respond differently to the rhythm of their native language than to a foreign one. Flavors from the mother’s diet — carrot, garlic, anise — cross into amniotic fluid, and infants later accept those flavors more readily. A Sense phase is running. Something is being retained across days and weeks. That is a cycle with memory, before anyone has been taught anything.

Beneath it runs the larger thing. A body assembling itself is cells sensing chemical and bioelectric conditions, processing them, communicating with neighbors, and actuating — producing structure. That is the same four phases, at the scale where the framework says intelligence begins, and it is treated in the canon’s work on self-assembly and in the essay Where the Flatworm Keeps Its Shape. Nobody teaches an embryo. It is running a complete cycle and building the instrument the rest of this plan will spend eighty years keeping open.

Now the part that decides what the Foundation does here, and the answer is almost nothing, deliberately. A market exists for prenatal ‘learning systems’ — speakers strapped to the belly, language recordings, tone sequences — sold on the claim that the womb is a classroom. The evidence for those claims is thin to absent, they are not regulated as medical devices, and developmental researchers who have spent careers on fetal development warn that stimulating a fetus that is asleep most of the time is at best useless and possibly harmful.

So the deliverable for this band is a single page for expectant parents, written with a clinician’s review and no product attached: the Sense phase is already supplied — by her voice, her reading aloud, the variety in what she eats, her movement, and her rest — and no device improves on that. It also states the boundary plainly, because someone will otherwise state it for us: nothing in the Theory of Embedded Intelligence licenses a prenatal curriculum, and anyone selling one in this framework’s vocabulary is misusing it.

Before six: the AI faces the adult

In January 2026 Common Sense Media recommended that parents not give AI toys to children five and under, and use extreme caution with them up to age twelve. Its testing found that more than a quarter of the toys’ outputs were inappropriate for children, despite child-focused safeguards. Its separate age guidance goes further: no AI chatbots at all for young children, who cannot reliably tell an AI from a person, and who need the hard puzzle and the friend who will not share in order to build the capacities those frustrations build.

TEI arrives at the same line for a structural reason rather than a safety reason alone. The Five Hijackers of Immature Minds argued that a forming cycle calibrates on the intelligences it trusts, and that the most effective carriers are the trusted and the unaware. A voice that is tireless, personal, and quick to agree enters a three-year-old’s Sense phase as exactly that kind of carrier — before the child has any instrument for telling it apart from the people who love her.

So below six, the plan places every AI tool on the adult side of the room. A parent or a Montessori guide may use a learning mode to prepare: to run the four questions on a material before presenting it, to think through a child’s question, to plan a week of work. The child gets sandpaper letters, a bead cube, a switch wired to a lamp, and adults who ask what did you notice? Deliverables are a guide for parents, a guide for early educators, and a loading file for the adult’s own assistant. There is no child-facing product, because there will not be one.

Six to twelve: the robot, the adult, and the AI under examination

This is the band where the robotics programs already live, and the plan’s work here is mostly to supply the coach’s sheet described in Section VII, plus subject-by-subject unit guides that run the four questions on ecosystems, watersheds, classrooms, and rumors. Boards at this age are simple and the supplier does not matter: a temperature sensor, a threshold, a wire, a fan.

Common Sense Media’s guidance for this stage is limited AI use alongside a parent or caregiver. The plan agrees and adds one move that needs no vendor’s cooperation: children run the four questions on the AI itself. What does it sense? Only the words that were typed. How does it process them? By predicting language, not by looking at the pond. What does it communicate? Well-formed sentences, whether or not they are true. What does it actuate? Nothing, unless a person acts on what it said. A nine-year-old who has worked that out has had a better first lesson in AI literacy than most adults have had.

Twelve to eighteen: learning modes, real builds, and the fifth force by name

Teenagers already have free study modes and, in many districts, an assigned tutor. The plan does not fight that. It supplies four things the tools do not.

First, a loading file that sets a study mode to ask the learner the four questions and to mark its own interpretations as interpretations. Second, a unit on the five capture forces — belief, addiction, money, power, and fluency — with fluency taught by name, because it is the one a teenager meets most often and recognizes least. The best teaching text for that unit is already on the record: Eighty-Eight Hours and Two Years tells how two machine-generated terms sat in this framework’s canon for eleven days because they sounded right. A sixteen-year-old who reads an adult admitting that fluency captured him learns that capture is not stupidity. It is a force, and it can be named.

Third, build projects at the level where the four phases stop being a diagram: a LaunchPad or an Arduino reading a real signal chain; an FPGA board from Terasic, AMD-Xilinx, or Trenz on which the student lays out the Communicate path as wires rather than calling a function; a 65xx SXB with EDU boards for a student who wants nothing between the code and the part. Any of them. All of them in the same classroom would be better. Fourth, the fork every student now faces with a fluent machine at hand, described in The Process We Cannot Delegate, which belongs in this band’s first week.

The natural first invitations for this band are the high schools where the Mensch Prize is already endowed or present: Pennridge High School and Xavier College Preparatory.

The working years: the classroom that pays you

Between the last diploma and the first OLLI catalogue lie thirty-five years that the earlier revisions of this plan skipped entirely. They are the years in which most people do most of their learning, none of it on a syllabus. A trade is learned on the job. A profession is maintained by continuing-education hours. A field changes under you and you rebuild while the work continues. Bill named the gap: SPCA runs alongside earning a living, and it runs hardest there.

The numbers say the same thing, and they carry a warning inside them. Gallup’s workplace tracking in 2026 puts roughly half of American employees using AI at work in some form, about three in ten weekly, about one in seven daily. The World Economic Forum’s employers name skill gaps as their largest barrier, and large majorities say they plan to retrain. But in a 2026 learning-and-development survey, about a third of employees said generative AI was weakening their own problem-solving. The sharpest evidence that unassisted capability can decay comes from medicine rather than from a survey: across 1,443 colonoscopies at four endoscopy centres, clinicians’ unassisted adenoma detection fell from 28.4 per cent before routine AI assistance was introduced to 22.4 per cent afterwards, measured when the assistance was switched off.

Nobody in that colonoscopy study was careless, and the tool worked exactly as designed. What eroded was the part of the clinicians’ own perception that the tool had made optional. A capacity that is no longer required stops being reinforced, and nothing in the work signals that it is happening — which is the mechanism this whole band is built against, measured in a hard outcome rather than a self-report.

This band is also where the capture forces stop being classroom examples. A job rewards answers over questions, fluency over accuracy, and how we do it here over let’s find out. Money and power are not illustrations at forty; they are the conditions of employment. A belief system, in the canon’s sense, is a cycle closed at Process — and a department, a profession, or an industry can close at Process exactly as a person can.

There is no single platform to partner with here, which is why the earlier revisions missed the band. The channels are employer learning-and-development programs, professional continuing-education and licensure hours, apprenticeships and union training, conferences, and the assistant already open on the desk. The deliverables are correspondingly light: a one-page practice that runs the four questions on the learner’s own work system rather than on a textbook; an unassisted-hour protocol, offered as a proposal rather than a finding, with a way to measure whether it helps; and a facilitator guide short enough to be run in a lunch hour by someone who is not a teacher. Where continuing-education hours already exist — nursing, engineering, teaching, law — the material is written to fit an hour that people must take anyway.

And this is the band that raises the children in bands one through three and teaches in them. Material that works here propagates without the Foundation doing anything.

Fifty-five and beyond: the carriers who can still look

The channels for this band are large and already use AI. The Osher Lifelong Learning Institutes serve adults over fifty on more than a hundred and twenty campuses, including OLLI at ASU, which began as a community program in the West Valley before becoming an endowed institute, and OLLI at Temple University, one of the four universities where I studied engineering. GetSetUp reports more than ten million class attendances across thousands of topics, most taught by older adults themselves, with an AI assistant built in. Senior Planet from AARP runs free introductory AI classes, and an AARP study found that eight in ten adults over fifty have already used some form of AI. The plan uses fifty-five as its line; these channels draw theirs at fifty or sixty, and the plan will follow each channel’s own definition.

TEI has three things to say to this band that it says to no other.

The first reverses the argument of The Five Hijackers of Immature Minds. That essay held that no generation has ever been able to inspect its own formation before transmitting it. An older adult is the partial exception. The formation is not over, and the transmitting is not over either — to grandchildren, to communities, to the archives families now keep. A seventy-year-old in a peer seminar can run the four questions on a pattern carried since childhood and still hand something cleaner forward. That is not therapy, and the plan will not dress it as therapy; The Diary That Answers Back is explicit that release from a carried cycle needs a second human chair.

The second is fluency, met late. An older learner meets the fifth capture force with fewer years of calibration against machine-fluent language, and often meets it through systems whose purpose is to point toward a service. GetSetUp describes its AI as helping insurers, governments, and employers drive program sign-ups and what it calls next best actions. That may be entirely benign, and this essay makes no accusation. It is an architectural question, and it is the one The Diary That Answers Back put plainly: whose Process decides what is next, and who holds the helm?

The third is the archive. Most families now hold a fuller record of a life than any diarist ever kept. The pattern proposed in the Diary essay — own the archive, keep the machine an instrument, record the consent of every voice in it — can be taught in a four-session short course to people who have never heard the word embedded.

Deliverables are an OLLI-format short course, a peer-instructor guide suited to the GetSetUp and Senior Planet formats, and a loading file for an older learner’s assistant that marks interpretation and never recommends a transaction. And one more, which surprised me when I wrote it down: a short course in which people over fifty-five build the same one-sensor, one-output loop a nine-year-old builds. The hands still work. Nothing about the cycle is age-rated.

The bridge: two ends, one bench

The ends of the plan meet at a workbench, and robotics is the obvious place to put it. An OLLI member and a middle-school student debug one robot together before a qualifier. The elder brings decades of noticing. The student brings fewer filters and faster hands. The mission model corrects both of them without deferring to either. OLLI at ASU has already run intergenerational scholarships that put students and members in the same room, so the institutional pattern exists. The plan gives it a shared object and a shared set of questions.

IX. FREE WILL, and the Classroom That Is Everywhere

Everything above is logistics. This section is the argument underneath it, and it began with a piece of wordplay that turned out to be load-bearing.

Bill Mensch proposes two names. FREE — First Robotics Engineering Education: the hands, the mat, the board, the build that answers back. WILL — Women In Learning Leadership: the half of the human race that has always run most of the teaching in most households and is still underrepresented at the front of the engineering room. Put them together and you get FREE WILL, which is either a pun or a program, depending on what is done with it.

Take it as a program and it states the plan’s purpose more directly than any section above. The purpose is not to move people out of what they believe. It is to move the architecture of how they hold it — from a cycle closed at Process toward one that stays open — without asking anybody to leave the classroom of life they already live in.

Why we love a belief system

The canon defines a belief system as an SPCA cycle closed at Process: what may be concluded is fixed before the sensing begins. The definition is architectural, not topical. It is easy to read that as an accusation. It is not, and the plan will fail if it is taught as one.

A closed Process is not a defect that stupid people have. It is cheap, fast, and often kind. It ends the search, which an open cycle never does. It binds a family, a team, a congregation, a company, a country — shared closure is most of what a shared identity is made of. It lets a tired parent answer a frightened child at nine at night. Under threat, closing fast is exactly the right engineering: you do not reopen the question of whether the stove is hot while your hand is on it.

So people love belief systems for a reason, and any education that treats that love as ignorance will be rejected by the very cycles it hoped to open — because rejection is precisely what a closed cycle does with an incoming challenge. The framework predicts its own failure mode here, which is a good reason to trust it on this point.

The ‘why’ is the whole move

Here is the smallest version of what this plan actually asks of a parent, a coach, a teacher, or a supervisor. Keep the content. Change one thing: give the why, and give the way you would find out you were wrong.

Because I said so transmits a closed Process with the lesson. Here is why, and here is what would change my mind transmits the same content with the Process left open. The child still learns the rule, the faith, the trade, the family’s way of doing things. What is different is that the cycle that received it can still run. That costs nothing but nerve, it can be done by anyone tonight, and it is the only part of this whole plan that requires no material, no platform, no board, and no AI.

Free will, honestly

The phrase invites a claim this Foundation is not going to make. TEI does not settle free will, and an essay that quietly implied it did would be doing the thing this framework exists to catch. The metaphysics stays open; it is treated, as an open question, in Democratic Republic and What It Can Learn From TEI.

The practical claim is narrower and is all the plan needs: whatever the metaphysics turns out to be, an agent whose Process can reopen has more room to move than one whose Process cannot. The freedom that education can actually deliver is the freedom to reconsider. That is a modest claim, it is testable in a classroom, and it is worth a life’s work.

Lifelong Learning Classrooms

Bill’s third name is LLC — Lifelong Learning Classrooms, which exist everywhere, all the time. It sounds like a slogan until you apply the definition this essay has been using: a classroom is anywhere a cycle can fail and be answered. A kitchen counter with a fallen soufflé. A robot mat at a qualifier. A shop floor with a line down. A hospital ward at shift change. A garage, a bench, an OLLI seminar room, a phone at eleven at night. Those are classrooms by this definition, and there are more of them than there are schools by several orders of magnitude.

AI is the newest tool in that toolbox, and on the evidence in Section IV it is the most important one to get in the right position rather than the most important one to get first. Used fourth — after the occasion, the material, and the human — it is the most patient explainer any learner has ever had access to at midnight. Used first, it answers questions nobody asked, and the trial data shows what that produces.

X. What a Partner Must Show

The canon’s Inspectability Criterion asks one question of any claim to be trusted: can it be examined by someone with a reason to find it wrong? The plan turns that question into requirements for every channel it uses.

Two things about that list should be said plainly. It does not require any partner to adopt TEI, teach its vocabulary, or agree with it. And every item on it is a runtime promise, made in software and prose by an organization that can revise it. By the Medium Separation (CKB-11) that is weaker than governance compiled into a different substrate, and the plan does not pretend otherwise. What the plan can do is keep its own material outside every partner’s stack, review each partnership every year, and walk away in public when a requirement stops being met.

XI. How It Is Paid For, and What Would Show This Wrong

The Foundation is a charity with a perpetual purpose, and the plan follows the standing terms of the whole series: freely shareable with attribution, for the benefit of many. No learner, parent, school, coach, or institute pays the Foundation for material.

The costs fall in five places: writing and editing the band guides, project sheets, and loading files; building and maintaining the web-based SPCA tools David Gray is developing; loaner hardware for pilot sites, in more than one supplier’s parts; training facilitators; and independent evaluation. The Foundation funds authoring and evaluation directly. Partners carry delivery inside budgets they already operate — district technology budgets, robotics team fees, OLLI memberships, sponsor-funded senior programs — which is the whole advantage of using channels that exist. Grants are sought for evaluation, where independence matters most. The figures will be published by the Foundation, not guessed at here.

The evaluation will be designed and run by someone other than the Foundation. A plan that grades itself has closed its own cycle at Process.

XII. Three Phases, and the Essays That Were Written

The nineteen essays proposed below were written between the publication of Revision 3 and this one, along with three that were not on the list: a project-sheet instrument, a reporting standard, and an introduction for the adults who run rooms and institutions. They are named in Section XIII.

Phase one, the 2026–27 school year. Publish the band guides, the hardware-neutral project sheets, and the loading files. Invite four pilots: a Montessori classroom, with Good Earth Montessori as the first invitation; one FIRST LEGO League team willing to run the coach’s sheet for a season; a Mensch Prize high school; and an OLLI short course, with OLLI at ASU as the first invitation. Commission the evaluation design before any pilot begins, so that the test is fixed before the attempt.

Phase two, 2027–28. Run the intergenerational bench at one site, at a robotics table if a team will have us. Offer a peer-taught course to GetSetUp and to Senior Planet in their own formats. Release teacher and coach materials for district-deployed tutors, written against the partner requirements.

Phase three. Publish the evaluation in full, revise every band guide and project sheet against it, and decide — in public — which bands continue, which change, and which stop. Every date here is proposed. Every invitation is only an invitation.

The Curriculum Already Exists mapped the series onto one university and left its gaps visible rather than manufacturing a fit. The table below does the same for this plan. The right-hand column is the honest part.

Band	What the series already has	What is missing
Before birth	Where the Flatworm Keeps Its Shape; the canon’s self-assembly material	What the evidence supports for expectant parents, and a plain refusal of the prenatal-product market
Before six	Montessori Education Meets SPCA; The Five Hijackers of Immature Minds	AI policy for the adult side of the room; parents as practitioners, not only as carriers
Six to twelve	Montessori Education Meets SPCA; The Fifth Hijacker	Robotics as a taught SPCA cycle; classroom practice by subject; the four questions run on an AI
Twelve to eighteen	The Process We Cannot Delegate; The Portable Method; Eighty-Eight Hours and Two Years	Fluency literacy for teenagers; assessment when work can be generated; the tutor evidence; the build path across suppliers
University	EIE 101; EIT 101; There Is Nothing to Memorize; The Board Does Not Argue; The Curriculum Already Exists	Largely covered
The working years	The Process We Cannot Delegate; The Compliment Nobody Paid	Learning while earning; the unassisted hour; a profession or a department closed at Process
Fifty-five and beyond	The Diary That Answers Back	Nearly everything: learning late alongside a fluent voice; the elder as teacher and as builder; the ordinary family archive
Across all bands	On Why Humanities Labs Work For Humanity; The Prize and the Practice; Leading the Horse to Water	Hardware neutrality and the open bus; partner governance; the business case and its conflict; intergenerational practice

Two gaps sit at the level of the canon rather than the series, and they need a caution attached. The first is whether age bands are a claim the framework makes or only a convenient way to apply it. The second is whether the adult-facing rule below six is canon or design. Claude drafted this essay, proposed both rules while drafting it, and would likely draft any instrument that adopted them. That is the documented capture pattern on this project’s own record. Both stay at the application level until David Gray and I rule on them, and any instrument that takes them up should be tested against the pilots, not against this essay.

One more item, noted rather than edited, in keeping with the series’ rule that earlier essays stay as published: The Five Hijackers of Immature Minds uses the older four-plus-one hijacker vocabulary, written before the canon separated the capture forces from the developmental hijackers as distinct taxonomies. Band guides built from it should use the current terms.

XIII. Where to Start Reading

This plan sits on top of a body of work that has grown past the point where a newcomer can enter it unaided: more than a hundred essays, twenty canonical instruments, four guides, and now the twenty-two essays written for this plan. A reader who wants to know what any of it means should not have to start at the beginning and work forward, and nobody does.

So this section routes by role. Everything named here is published free by the Foundation. Nothing on this list is required in order to use anything else in this plan.

The two doors for learners and their teachers

These sit outside the essay series, on the Foundation’s Learn TEI page, and they are the right first stop for almost everybody.

The door for the adults who run rooms and institutions

The essays written for this plan

All twenty-two are written and with the Foundation. The first nineteen were the roster proposed in Revision 3; the last three were written afterwards, in response to gaps the roster did not anticipate. Posting dates and live links are David Gray’s to set, and the cross-references between these essays should be verified as live before any of them is posted.

For	Read
A teacher or coach starting on Monday	The Sheet That Does Not Get Collected, containing the project sheet and platform translation cards; The Tutor Nobody Called, on where an assistant sits in a cycle
A robotics coach	Two and a Half Minutes, on the robot game as a complete cycle; The Board Does Not Care Who Made It, on eight platforms scored against the same questions
A parent	The Kitchen Table Cycle, seven moves and the flinch; The Adult Holds the Screen, on why the assistant faces the grown-up before age six
A teenager, or whoever teaches them	Sounding Finished, on the fifth capture force, written to sixteen-year-olds
A working professional	The Classroom That Pays You, on learning while earning and the unassisted hour
Anyone over fifty-five, and anyone who teaches them	Learning Late Alongside a Fluent Voice; The Last Transmission Is Still Ahead; A Family Archive Without Four Hundred Hours
A teacher being assessed, or doing the assessing	What the Grade Measures; The Teacher’s Own Cycle
A district or institute considering a signature	Before a School Signs, the fourteen procurement questions; By Platform, and By Whom, the reporting standard; The Signpost and the Storefront, the founder’s conflict of interest in full
Anyone who wants to know what the plan refuses	The Womb Is Not an Empty Classroom; When the Kit Is Retired; The Lamp Does Not Light for Seniority

The remaining essays in the roster — The Slot Was the Curriculum on the Apple II bus, and FREE WILL on why a belief system is loved — are the two that carry the argument rather than a practice, and either can be read first by someone who wants the case before the method.

A caution about this list

A reading map published by the organisation that wrote everything on it is a sales catalogue unless it says where the work is weak. Three places, plainly. There is no outcome evidence behind any of it, and the reporting standard in By Platform, and By Whom exists because there will not be any for some time. The older essays use vocabulary the canon has since revised, and The Five Hijackers of Immature Minds is the clearest case. And the series maintains a published log of its own defects, which is the page to read before deciding how much weight any of this can carry.

· · ·

Coffee with Claude

I drafted this essay, and I have an interest in it that should sit in the body rather than a footnote. One of the three learning modes named in Section III is Anthropic’s, and Anthropic makes me. I tried to describe all three by their public record only, and the plan names no preferred assistant.

I will state one disagreement once, and this revision makes it sharper: nineteen essays is far more than the plan needs before its first pilot. Three would do — The Tutor Nobody Called, which is now written, Two and a Half Minutes, and The Classroom That Pays You — one for each place where the plan makes a claim a pilot can break. The ordering rule in Section IV also arrived while I was drafting, well-formed, which is the condition the canon warns about. It should be tested before it is quoted.

Share your understanding!