The video is short on purpose. These notes are what a teacher would want at the elbow while watching, so the tour lands as a working tool and not a demo.
What the video shows
The tour walks the workbench in the same order a class would meet it on a first day. You see the artifact panel (where a student's work lives), the gate list (the small, testable claims that must pass for the artifact to advance), the evidence column (Class A empirical-in-session, Class B code or inspection, Class C configuration, Class E expert citation, Class F falsifier present, Class U unverified), and the review pane where a teacher signs the AI-authorship fence. Nothing on the screen is decorative. Every field maps to a move a student or teacher will make out loud.
The workbench does not run a chat model against a student's paper. It records what the student did, what the student cited, what test would break the claim, and who reviewed it (Class C on the configuration; the fence is enforced at the file boundary, not by trust).
Teacher notes, panel by panel
Artifact panel. Treat this as the student's lab notebook page. When a student updates a claim, ask for the reason in one sentence: what changed in the evidence. That single sentence is the Bayesian update in plain speech. You do not need the math to teach the move.
Gate list. A gate is a small predicate that must be true for the artifact to move forward. Good gates are boring: "the cited page number matches the quote," "the falsifier is written before the conclusion," "the source is a primary source or is labeled as secondary." Boring gates teach thinking because they force the student to look again.
Evidence column. When a student tags a claim Class E, ask: which sentence in the cited source carries the claim. When they tag Class F, ask: what observation would make you drop the claim. Class U is fine as a working label; it just cannot leave the page as Class A.
Review pane. The AI-authorship fence is a checkbox with teeth. The student signs what a tool did and did not do. The teacher countersigns. This is a workflow rule, not a moral stance.
Where to go for deeper technical context
For teachers who want to sit longer with the underlying ideas, the Themesis YouTube channel is the resource we point to most often. Two starting points, with our one-line frames rather than a paraphrase of the video itself:
- Deep Learning Did It. Transformers Did It. Active Inference Just Did It Again (Part 1): a video explainer in the UNI cluster (Class E) that we recommend after a teacher has already seen the workbench once, so the vocabulary has something concrete to attach to.
- Themesis YouTube channel: a running library of deep-dive videos on active inference (Class E). We recommend it to learners who want the theory at a slower pace than a classroom period allows.
We link. We do not summarize her prose. If a video changes your teaching, cite it the same way you would ask a student to cite it.
What the video is careful not to say
The video does not claim the workbench heals anything, fixes anything, or replaces a teacher's judgment. It is a place to make thinking visible and checkable. UNI is a working hypothesis on an attainable path toward General Natural Intelligence: a natural, active-inference approach whose evidence is growing, evidence-classed, and tested in the open. Do not take the claim on faith. Test the build, inspect the gates, and help us find where it fails.
The tour also does not claim the tool authored anything a student turned in. The fence is there so that question has a documented answer every time.
Next steps
- Walk the same panels in your own room with The Teacher's Workbench Tour.
- See what changes when two teachers share the pane in Co-Teaching With the Workbench.
- For a longer reading and viewing list, see A Themesis Resource Map, a Teacher's Note.
- If you want the workbench brought into your building, start at /workshop.