Why not just ask for “simpler language”?

Because simplification without a frozen-fact rule compresses the facts too. The constraint is what keeps “simpler” from becoming “less true”.

Does this work for translation as well?

Yes, and the meaning-change list matters even more there. Ask for the list to name any idiom it had to substitute, so you can check the substitution yourself.

Writing & content

Tone shift, same facts

Rewrite anything for a different audience while locking the facts in place — with an explicit check that nothing was added, softened, or quietly dropped.

Use this when the same information has to reach two very different readers — a teacher and a parent, a recruiter and an engineer.

The pattern

Copy this and replace everything in [SQUARE BRACKETS] with your own specifics.

Prompt template

Rewrite the text below for a new reader.

Original: [PASTE THE TEXT]

New reader: [WHO THEY ARE, what they already know, what they will do with this, and what they are worried about.]

Constraints:
- The facts are frozen. Do not add a fact that is not in the original, do not remove one, and do not soften a number.
- Change only vocabulary, sentence length, ordering and emphasis.
- [ANY HARD LIMIT — word count, reading level, no jargon, etc.]

After the rewrite, list every change you made that could be read as a change in meaning, and why you judged it safe.

Why each part is there

What they will do with this: Audience descriptions like “non-technical” are too vague to act on. What the reader has to DO decides what goes first and what can be cut.
The facts are frozen: Rewriting for a friendlier tone is where numbers quietly get rounded and caveats quietly disappear. Freezing the facts turns tone into the only variable.
Change only vocabulary, length, ordering, emphasis: Names the four legitimate levers. Anything outside that list is a content change wearing a tone change's clothes.
The meaning-change list: A self-audit you can read in 20 seconds. It surfaces the judgement calls — “I wrote 'roughly a third' for 32%” — instead of hiding them.

Before and after

The lazy version

Make this sound friendlier and easier to understand.

What goes wrong: Friendlier reliably means vaguer. “Attendance fell to 61%” becomes “attendance has dipped a little recently”, and the one number the reader needed is gone.

The pattern, filled in

Worked example

Rewrite the text below for a new reader.

Original: [the technical write-up of my ML project — 400 words of model architecture and validation metrics]

New reader: my mother. She has no technical background, will read this to understand what I spent two months on, and is quietly worried that this is not "real" studying that helps my exams.

Constraints:
- The facts are frozen. Do not add a fact that is not in the original, do not remove one, and do not soften a number.
- Change only vocabulary, sentence length, ordering and emphasis.
- Under 200 words, no acronym without a plain-English gloss on first use.

After the rewrite, list every change you made that could be read as a change in meaning, and why you judged it safe.

What that buys you: The accuracy figure and the dataset size both survive, in plain words. The audit list flags one real judgement call — that “overfitting” became “it had memorised the practice questions” — so you get to approve the metaphor rather than discover it later.

Now do it on your own work

Take something factual you wrote for one audience and rewrite it for a genuinely different one. Then run the ORIGINAL and the REWRITE past a third person and ask them to list the facts in each. The lists should match.

Check yourself — every box should be true

Every number in the original appears in the rewrite, unrounded.
No fact appears in the rewrite that is absent from the original.
The meaning-change list has at least one honest entry (if it is empty, the rewrite probably changed nothing).
Your third reader's two fact lists match.

Then explain it back

Tone and content are supposed to be separable, but one of your edits sat right on that line — which edit was it, and how did you decide it was still the same fact?

Write two or three sentences in your own words. If you cannot, you copied a prompt rather than learning a pattern — which is the whole difference this library is trying to make.

Does this depend on which AI tool I use?

Tool-independent. The one practical difference: assistants with a persistent style or custom-instruction setting will drag their house voice into every rewrite, so if the output keeps sounding like the same person regardless of audience, check your global instructions before blaming the prompt.

Why it actually works

Same facts, different surface form. That is the intuition behind embeddings: meaning as a position that can survive a change of wording.

See meaning separated from wording →

Prompting well is a skill you can pick up in a week. Understanding what is happening underneath is the 18-month curriculum — and it is what stays useful when the tools change.

FAQ

Why not just ask for “simpler language”?: Because simplification without a frozen-fact rule compresses the facts too. The constraint is what keeps “simpler” from becoming “less true”.
Does this work for translation as well?: Yes, and the meaning-change list matters even more there. Ask for the list to name any idiom it had to substitute, so you can check the substitution yourself.

Next patterns

The Prompt Patterns Library is a separate, optional track — it does not change the 78-week curriculum or your roadmap. See what the curriculum actually covers →