Field notes from the classroom.

A short predict, check, adjust loop you run yourself. No pitch. Guess first, look, compare, and watch the passes stack up as rings.

Today EducateWright launched publicly, with $25 million in pledged funding for anti-violence education and active-inference training. Here is what that funds, and how to step forward through any of three open doors (Class E).

A short, honest glossary teachers can post on the wall: five active-inference terms with one-line classroom definitions (Class E).

A five-day journal for a teacher new to active inference: what to read (Class E), what to try on the workbench (Class C), and where confusion is normal.

A short dialog showing how a teacher and a student revise beliefs together as new evidence arrives, with the shape of Bayes' rule visible (Class E).

A short guide for teachers on citing the standard active inference reference (Class E) without overclaiming validation, plus the classroom conventions we use (Class C).

A teacher-facing walkthrough of the generative-model idea, kept precise (Class E) and grounded in observable classroom demonstrations (Class C).

A plain-language read of KL divergence as a distance between beliefs, with honest notes on what a fuller treatment requires (Class E, Class C).

A calm, non-clinical way to introduce the Markov blanket to students as a boundary between self and environment (Class E).

A short, ordered reading list for teachers new to UNI: what to read first, second, and third, with Parr, Pezzulo, and Friston (2022) at the technical end (Class E).

One lesson, one gate, one written prediction: a low-risk way for a curious teacher to try the active-inference posture without changing a whole unit (Class E).

A teacher-facing summary of the Themesis 2026 resource map (Class E), recommended as an ecosystem starting point for colleagues who want to orient before choosing a pathway (Class C).

For math-hungry teachers preparing for UNI: Themesis T3 gives statistical mechanics grounding (Class E) with real effort required.

A plain-language walkthrough of the core claim in active inference: agents minimize free energy by updating a generative model, with classroom analogies and the technical terms kept visible.

A working checklist for building a UNI unit whose outcomes are stated as gates, whose evidence is classed, and whose tools are declared (Class E, Class C).

A classroom rubric that gives explicit points for stating what would prove a claim wrong, so falsifier writing becomes a habit (Class C).

How to map a UNI unit onto NGSS, Common Core, or a state framework without distorting either, and how to be honest about the loose seams (Class E, C).

A gate-oriented approach to classroom assessment that can coexist with, and in some cases reduce reliance on, standardized testing (Class E, Class C).

A, B, C, E, F, U as tags students attach to their own claims, with worked examples across subjects (Class E, Class C).

A worked semester in high-school science built on prediction, revision, and citation gates, with student-written falsifiers required as a graded artifact (Class E, C).

A worked example of three gates in a two-week science unit, each with an observable checkpoint, so grading rewards reasoning students can inspect (Class E, C).

A classroom walkthrough for reading a research preprint responsibly, covering authorship, evidence classes, falsifiers, and what a peer review would still need to add (Class E, Class C).

A calm, honest script for a teacher proposing a UNI pilot to a department head, with claims tagged by evidence class.

Ten questions a school leader can bring to any AI vendor meeting, covering student data, opt-out, authorship, audit rights, and exit clauses (Class E, C).

For leaders considering a small pilot: what one classroom, one teacher, and one term actually look like on the ground, with honest expected outcomes and continuation gates (Class E, Class C).

An honest, evidence-classed reflection on what shifted, and what did not, after a full academic year working with one partner school.

Straight talk on the price of a teacher cohort: what is included, what is not, and why the number is the number (Class C, Class E).

The measurement plan for a classroom pilot: what we count, what we refuse to count, and which classes of evidence apply to each finding (Class E, Class C).

The recurring questions from workshop follow-up office hours, with honest answers, source citations (Class E), and configuration notes (Class C).

A concrete walkthrough of the paid workshop day for teachers and school leaders: workbench tour, gate design lab, honesty policy drafting, and the follow-up cadence (Class E, Class C).

The exact follow-up cadence after a SolutionWright workshop: office hours, a shared gate library, a review cycle, and what does not happen without teacher commitment (Class C, Class E).

Practical guidance on staying trauma-informed in classroom practice while holding the boundary between educator work and licensed clinical care (Class E, Class C).

A short teacher-facing read on Jay Kumar Chimata's JobFirst.ai framing of the actual labor market (Class E), useful when advising older students about tools, roles, and consent.

When a chatbot writes the paragraph, the student skips the prediction-error cycle that would have updated their model (Class E).

A ten-minute classroom routine grounded in active inference (Class E): students commit to a prediction, gather evidence, and log the update.

A practical frame for teaching students how to choose tools, disclose their use, and opt out without penalty (Class E, Class C).

How a written prediction-then-check routine surfaces thinking from students who rarely speak up, grounded in active inference (Class E) and a small classroom pilot (Class C).

A teacher's guide to spotting the gap between polished writing and internal generative modeling, with quick classroom probes (Class E, C).

Companion notes for a short video tour of the classroom workbench, with pointers to Themesis video resources (Class E) and configuration notes for teachers (Class C).

How two teachers can share a UNI session at the workbench, splitting conceptual framing from hands-on inspection so students see both the model and the receipts.

A short walkthrough of live gate inspection during the workshop: what teachers see on screen, what they do not, and how the exercise teaches epistemic humility (Class C, Class E).

A calm walkthrough of the inspectable workbench teachers use in the workshop: what a running generative model looks like on screen, where the gates are, and how to read them (Class C, Class E).

A plain statement of data posture (Class C): in the workshop configuration, student data does not flow into the teachers workbench, and the workbench itself is inspectable.

A short reference on the authorship fence for classrooms: what it says, when it applies, and a copy-ready block teachers can adapt (Class E, Class C).

A brief explainer for teachers on why we say General Natural Intelligence (GNI), not AI, for our own work (Class E, Class C).

A practical checklist for teachers and school leaders: what the tool sees, what it stores, what it claims, and where its gates live (Class E, Class C).

Concrete practices for citing, tracing, and evaluating sources when students may have used generative tools, tied to evidence classes.

A one-page declaration students attach to their work naming which tools helped, how, and what they did with the output (Class E, Class C).

A plain list of claims SolutionWright will not make about UNI in schools, with the reasoning behind each fence (Class E, Class C).

A one-page teacher script for the parent conversation: the honesty fence, the working-hypothesis posture, and what students actually do (Class E, Class C).

A calm, non-punitive classroom protocol (declaration, gate check, re-prediction) that keeps the learning intact when a student has used a chatbot (Class E, Class C).

A working practice for the moments a gate breaks, a claim wobbles in class, or a public statement needs to be retracted (Class C, Class E).

A short, honest explainer for teachers, colleagues, and parents on why our program refuses the AI label for its own work and uses a working-hypothesis frame instead.

A four-move stance a teacher can hold in front of any class: state the claim, state the evidence class, state the falsifier, invite challenge (Class E, Class C).

A pillar for curriculum designers: gate-oriented assessment that observes prediction, revision, and citation (Class B, C, E, F) instead of scoring fluent output.

A realistic readiness path into Universal Natural Intelligence for teachers without a coding background, with honest scope, evidence-classed claims, and a workbench you can inspect.

A pillar policy for classrooms: every artifact carries an authorship fence declaring tool assistance, protecting learners from extraction and inflated vendor claims (Class C, E, F).

A pillar on protecting student agency by treating the learner as a prediction-making mind (Class E), not a consumer of chatbot output.

A pillar page for school leaders: the scope, cadence, and price of a SolutionWright partnership, and the lines we will not cross (Class C, Class F).

The pillar case for shifting classroom time from prompt tricks to active inference literacy, with observable gates students can inspect (Class B, Class C, Class E, Class F, Class U).

A teacher-facing summary of the SeedIQ ARC result (Class E): what it showed, what it did not settle, and why non-transformer approaches belong in the conversation.

A teacher-facing explainer of the Stratified Palimpsest benchmark, what it tests (Class E), what it does not settle (Class C), and how to bring it into your classroom without hype.

A careful naming of UNI as a working-hypothesis build program on an attainable path toward General Natural Intelligence, with the boundary of what it does not claim (Class E, Class C).

The honest answer for teachers weighing a free chatbot against a gated, evidence-classed pedagogy (Class E, Class C).