Using AI well, and protecting the people it touches.
Twelve lessons on working with AI the way a careful professional works with any source: knowing what it is, asking it well, checking what it says, deciding who signs off, judging the screens it builds, keeping people at the center, recognizing manipulation, from people and from AI. Lessons 9–12 are the Dark Triad series. Essentials are shorter habits that fit in at any point: staying secure, keeping your devices and accounts safe, and talking with AI on the phone. There are three audience tracks: educators, professionals and students.
Designed by David Petry.
A course on checking AI output should show which of its own claims are checked.
Every factual claim this course makes about AI is a record in
attestation-ledger, the
same engine behind the Live Sound EQ SOP.
A claim written by a model shows as Proposed until a named person attests
it. An attestation states who signed, on what basis, and when. Attestations
lapse unless someone re-checks them: after two years by default, or sooner
or later when the attester sets a different shelf life for that claim. This matters here more
than most places, because claims about how AI tools behave go out of date fast.
A stale claim shows as Lapsed on the page instead of quietly going wrong.
A model can propose. Only a named human can attest.
Lesson 4 teaches that rule. The rest of the course follows it.
The objectives, assessments, diagrams and principles are the same for everyone. What changes by track is the hands-on example, why it matters, and what to practise afterwards. In Lesson 1, educators mark up an AI-drafted staff newsletter, professionals an AI-drafted email to a manager, and students an AI-written history essay paragraph. Each passage contains the same kinds of error.
| Framework | Its job here |
|---|---|
| Backward design (Wiggins & McTighe) | Objectives first, then the evidence each was met, then the activities |
| Revised Bloom's taxonomy | Every objective names a level and an observable verb |
| Concrete → Pictorial → Abstract | A real output in the learner's hands, then a picture, then the principle |
| ARCS motivation (Keller) | Why this matters, said differently for each track |
| Kirkpatrick evaluation | Exit ticket, pre/post checks per objective, a transfer task with two-week follow-up |
The test suite enforces all of it. A lesson can't be marked ready until every objective has a matched pre-check, an activity, and a post-check, and the stage timings add up to 45 minutes.
| # | Lesson | Status |
|---|---|---|
| 1 | What a Model Actually Does | Built |
| 2 | Asking Well: Prompt Structure | Built |
| 3 | Checking What It Says | Built |
| 4 | Who Signs Off: Human in the Loop | Built |
| 5 | How It Goes Wrong, and a Checklist That Holds | Built |
| 6 | UX/UI: Judging What AI Builds | Built |
| 7 | Accessible and Mindful UX | Built |
| 8 | Human-Centered AI | Built |
| 9 | What the Dark Triad Is, and Isn't | Built |
| 10 | Spot the Pattern, in Person and Online | Built |
| 11 | When AI Manipulates | Built |
| 12 | Protect Yourself and Your Gifts | Built |
| Essentials | Title | Status |
|---|---|---|
| 1 | Staying Secure in the Age of AI | Built |
| 2 | Keeping Your Devices and Accounts Safe | Built |
| 3 | Talking With AI on the Phone | Built |
Every lesson's claims start as model proposals. The hub shows how many have
been attested so far. The author checks and signs them at a private
attestation desk on his own computer (npm run attest; see
docs/attestation-guide.md).
npm install
npm run preview # http://localhost:8080
npm testNo build step. The pages are plain HTML and ES modules. They need to be served over http rather than opened from disk. The ledger engine and the fonts are hosted with the site, so pages make no requests to other sites.
Claims live in claims.js. After checking one, replace its
attestation block with your own:
attestation: {
source: "practitioner",
by: "Your Name",
role: "Your role",
basis: "What you checked it against",
verified: "2026-09-23",
ttlDays: 730 // optional shelf life; 730 days if left out
}docs/attestation-guide.md has worked examples
of attesting, rejecting and re-checking.
The model's original proposal stays in git history. If the wording is wrong, don't edit it while attesting. Reject it with a reason and propose a replacement.
- Grade 6 Singapore Math (CPA): the curriculum this one's method and styling come from.
- attestation-ledger: the claims engine.
- Live Sound EQ SOP: the same rule applied to console EQ values.
- Code (renderer, styles, pages, tests, scripts): MIT — see LICENSE.
- Course content (lessons, claims, course text, lesson designs): CC BY 4.0 — see LICENSE-CONTENT.md for exactly what's covered and how to credit it.
- Fonts (Lexend, Fraunces): SIL Open Font License — see
fonts/OFL-*.txt.