Curriculum / Course Design — by an AI council
Design a course that causes real learning, not just covers topics. Four rival AIs debate it and hand back: A course outline with objectives, sequence and assessments that prove learning.
A course outline with objectives, sequence and assessments that prove learning.
Signed off by an independent Final QA audit — with a “verify before you rely on this” list, never a black box.
Why this team
Courses default to coverage — everything the expert knows, in order — and coverage is not learning. The Educator brings the learning-science toolkit (retrieval, spacing, worked examples), the First-Principles Thinker re-derives what the learner must actually be able to do at the end, and the End-User Advocate sits in the seat of the struggling learner in week three, where the dropout happens. The Minimalist deletes the content that feels important but serves no objective, and the Pragmatist fits the design to the real constraints of format and time.
Each seat runs on a different frontier model — GPT, Claude, Gemini, Grok — so it’s a genuine cross-model review, not one AI in costumes. How a run works →
What the debate sounds like
An illustrative excerpt — A data analyst is turning her internal SQL training notes into a six-week evening course for career-changers with no programming background.
Start from the terminal behaviour, not the syllabus: at the end of six weeks, a career-changer should be able to take a messy business question and produce a defensible query against an unfamiliar database. Every module now has a test — does it move a novice toward that? Half of the internal notes fail it, because they were written for colleagues who already think in tables.
The week-three problem in this design: JOINs. That is where “this makes sense when the instructor does it” collides with “I cannot do it alone”, and where evening-course attrition concentrates — these are tired adults with jobs, and one lost week compounds. Week three needs to be half the pace of the plan: worked examples, then faded scaffolding, then solo — not one demo and a homework cliff.
The single highest-impact structural change: every session ends by writing queries against a realistic messy dataset, and every session opens with five minutes of retrieval from previous weeks — ungraded, low stakes, brutal for retention in the best way. Passive re-watching is what these learners will default to, and it is the method with the worst evidence base in the literature.
Questions
What should I bring — I only have rough notes, not a curriculum?
Rough notes are the normal starting point: describe the subject, who the learners are, their starting level, the format and hours, and what they should be able to do at the end. The council builds the outline from the objectives — arguably better than from polished materials, which tend to smuggle in the coverage instinct.
Does this work for corporate training and workshops, not just courses?
Yes — the method is format-agnostic: objectives stated as behaviours, sequence built on prerequisites, activities that cause learning, assessment that proves it. A two-day workshop gets the same treatment as a semester; the constraints change the answer, which is why the brief asks for them.
How does the design prove that learning actually happened?
Each module carries an assessment matched to its objective — performance tasks rather than recognition quizzes where the objective is a skill. If learners can only answer multiple-choice questions about JOINs but cannot write one against a new dataset, the course taught familiarity, not capability, and the design treats that as failure.
Your material is used only to run your review — never to train public models. Encrypted in transit and at rest. Security & privacy →
Want full control — pick your own minds, set the depth? Open the full council →