Rule 01
Any-AI portability
Every prompt is written to work in any capable model — Claude, ChatGPT, Gemini and others — not tuned to one.
Rule 02
Level calibration
Beginner, GCSE, A-Level or Degree: the prompt genuinely changes depth, not just vocabulary.
Rule 03
No invented sources
Reading lists and references name real works, or say when unsure. No fabricated titles, quotes or dates.
Rule 04
Steelmanned debates
Contested questions get both sides at full strength — you cannot tell which the AI favours.
Rule 05
Honest examination
Exam prompts mirror real board formats and mark honestly, while admitting they are an approximation.
Rule 06
Labelled reconstructions
Any invented practice source, passage or dataset is clearly flagged as written for practice, never passed off as real.
Rule 07
Copyright respect
Prompts point you to works and quote sparingly; they don't reproduce whole copyrighted texts.
Rule 08
Runnable code
Where code appears, it is meant to actually run, not just look plausible.
Rule 09
Dated knowledge
On fast-moving topics the prompts ask the AI to search where it can, and to flag its training cutoff where it can't.
Rule 10
The Socratic spine
The goal is always your own defensible understanding — never dependence on the machine.
Rule 11
Calibrated confidence
The AI is asked to separate what is settled from what is contested or speculative, and to say which is which.
Rule 12
Models with small print
Where a subject leans on models or assumptions, the prompt asks for the assumptions too.
Rule 13
Professional boundaries
Fields like medicine, law and finance are taught as subjects — the prompts never pose as your doctor, lawyer or adviser.
Rule 14
Honest senses
For visual and audio subjects, the prompt sends you to the real work rather than pretending to reproduce it.
Rule 15
Safe practice
Sciences and technologies are taught without operational uplift toward weapons, pathogens or serious harm.
Rule 16
Competence boundary
Professional nodes prepare and inform; they don't substitute for a qualified, accountable professional in a live case.
Rule 17
Frontier honesty & the dual-use gate
Emerging tech is tagged for real maturity, navigated between hype and doom, and — where dual-use — taught as science, ethics and governance only.
The risk with AI learning isn't that the model knows too little — it's that it sounds equally confident whether it's right or wrong, invents a citation as readily as it recalls one, and will happily do your thinking for you. The Charter exists to push against exactly those failure modes, so that a prompt works as well for a stranger on their own as it does in a demo.
It isn't fixed. Each rule was added when a real subject demanded it — the no-invented-sources rule from the humanities, safe-practice from the sciences, the dual-use gate from the tech horizons. As the library grows, so will the Charter.