The method
Something happens live. Then you do it.
Every session is 90 minutes in two halves. The taught hour is a mirror or a build-along; the last 30 minutes demonstrate the technique on a realistic student task, for students to repeat on their own.
Every session
One rhythm, two shapes
- 01SOMETHING HAPPENS LIVE
An experiment on the class, or a build-along
The room commits an answer on their phones and sees itself fooled, or builds the prompt, the workspace or the pipeline live with the presenter.
- 02THE MECHANISM
Why the machine does that
The effect explained from how a language model works, so it can be predicted next time rather than memorised.
- 03ON A REAL TOOL
Shown on a real tool
Proved live on a free-tier account, with designated moments where the same task runs on a stronger model beside it.
- 04A NAMED SKILL
One thing to do on Monday
A practical technique, demonstrated in the lab and repeated by students on their own work.
The mirror · 4 sessions
An experiment on the class, then on the machine.
For sessions where the room has to feel the effect first: the anchoring split, the model that folds when you push back, the real-or-AI confidence gap, the authority-check reveal, the audit. You commit an answer on your phone, the room sees itself fooled, and a real model is then caught doing the same thing on screen.
The build-along · 4 sessions
Build it with the room. Leave holding it.
For sessions that teach craft. The class builds the thing with the presenter, on a free-tier account: the specification and its output contract, the judgment in the window, the research workspace over a bundle of authorities, the drafter–critic–verifier pipeline. The mechanism is explained as it is built.
The lab · 30 minutes
A demonstration, its reveal, and the assignment
Each lab is scripted well enough that someone who did not write the session could run it, on a free account.
- 01≈25 min
The demonstration
The instructor runs the technique end to end on one realistic student task (a moot issue, a judgment, a bundle), with the exact prompts on screen and the room committing predictions by phone before each result.
- 02≈5 min
The assignment
The same technique on a task of your own, done after the session and submitted on the course platform before the next one. The steps, tools and checklist are on each session's page.
- 03in the demo
The reveal
The room's predictions against what actually happened, tallied on screen as the demonstration runs.
Every lab completes on a free account. How the free tiers are handled
The designated comparisons
Model's ceiling, or your prompt's ceiling?
Three times, the same task runs on a free tier and on a stronger model or higher reasoning-effort tier, side by side. At least once, paying changes nothing. What transfers is the diagnostic.
The clause rewrite — where paying changes nothing
Rewrite a dense indemnity clause in plain English for a client, to a tight specification with two exemplar rewrites supplied: a pure reshaping task with a full brief.
Paying changes nothing
The two-hundred-page judgment — where the window is real
Build an issues-and-holdings table, with paragraph pinpoints, for a very long multi-opinion constitution-bench judgment — the kind with concurrences that decide everything.
Partly the model's ceiling
The limitation computation — where paying is the answer
A s.34 Arbitration Act challenge: signed award received on a stated date, court closed for vacation over the deadline, filing on a stated later date. In time, out of time, or condonable? Work every limb: three months, the thirty-day proviso, 'but not thereafter', and the s.4 Limitation Act interaction.
Genuinely the model's ceiling
On your phone
The clicker in your pocket
No app, no account, no setup. You vote from your phone's browser, one tap to join and one to commit. Answers stay private until the reveal, and your laptop stays free for the lab.
On your laptop
Real tools, free accounts
A free ChatGPT, Claude or Gemini account, Google NotebookLM, and Indian Kanoon. The presenter drives the same tools on the shared screen, so the machine's behaviour happens live in front of everyone. Where a paid tier appears, the presenter runs it and you watch.
Inside a mirror · commit, then reveal
How the class experiment is run
Answers are locked before anyone sees the spread, so the room cannot anchor on itself.
- 01
Commit
Everyone votes a private answer on their own phone: an estimate, a prediction of what the model will do, a confidence level. Nobody anchors on the loudest voice.
- 02
Lock
The poll closes and answers are sealed before anyone sees the spread. The room's confidence is captured along with its choices.
- 03
Reveal
The distribution goes up on screen. Confidence and correctness rarely line up, and the gap is the lesson.
- 04
Mirror it on the model
The presenter runs the same prompt live on a real tool. The machine's failure mode mirrors the room's, and one skill fixes both.
Joining a session
The poll link is unique to each session and shared on screen when class begins. There is nothing to set up in advance; keep your phone handy.
Link shared in class
How the live room worksWhere this lives
Every session is built on this rhythm.
See the eight sessionsEverything taught is then practised.
You feel the effect or you build the thing in class, then repeat it on your own task and submit it on the platform. Bring a phone; the polling runs on it.