Notes to Prose

About the Notes to Prose

Everyone has the notes. Six lines in a text file, or scrawled during a call, and every one of them is true and useful and none of them is a sentence. Turning that into something you can send is a specific and disproportionately tedious task: ordering the points, joining them so each follows from the last, and giving the thing an opening and an ending that were never in the notes at all.

This page does that joining. You paste the fragments, it returns finished writing. The connective tissue is the work — the ordering, the transitions, the framing — and the facts stay exactly as you left them.

Four forms are on offer because rough notes turn into different documents. Plain paragraphs suit a passage inside something longer. A short report adds a one-line summary and two or three headed sections, which is the shape people actually read at a desk. An email body drops straight into the message with no subject line and no signature. A status update follows the order that makes an update useful: what happened, what it means, what is next.

The switch marked "Add nothing I did not say" is on by default and is the reason to trust the output. Expanding notes is precisely the situation where a language model will helpfully invent — a plausible figure to round out a sentence, a second example beside your one, a reason that fits. With the switch on, every fact in the result has to trace back to a note, and where a note is too terse to build a whole sentence from it writes the short sentence the note supports rather than filling the space. Anything genuinely unclear is listed at the end instead of being guessed at, which turns an ambiguous note into a question you can answer in five seconds.

The depth slider controls how much connective writing goes around the points. It never controls how many facts appear, because that number is fixed by what you pasted.

How to use it

1. 1Paste your notes. Bullets, fragments, half-sentences and shorthand are all fine.
2. 2Pick the form: paragraphs, a short report, an email body or a status update.
3. 3Leave “Add nothing I did not say” on unless you want it filling gaps.
4. 4Set the depth slider for how much connective writing you want around your points.
5. 5Read the UNCLEAR list at the end, if there is one, and answer those questions in your notes before running it again.

Questions

Will it invent facts to pad the writing?

Not with the default switch on — it is instructed to write the short sentence a terse note supports rather than the long one it cannot. Turn the switch off and it may add ordinary connective context, though even then it is barred from introducing a figure, date, name or claim that is not in your notes.

How rough can the notes be?

Very. Fragments, arrows, abbreviations and single words all work. What limits the result is not the tidiness of the input but its completeness: a note reading “pricing issue” gives it nothing to build on, whereas “pricing issue — enterprise tier undercuts mid tier at 40 seats” gives it a sentence.

What happens to the order of my points?

It is free to reorder them, and usually should. Notes come out in the order things were said, which is rarely the order that reads well. If a particular sequence matters, number your notes and say so in the first line.

Is this the opposite of the bullet point page?

In effect, yes — that one goes from prose to points, this one from points to prose. Running one after the other on the same material is not a useful loop, though. You would get a document twice removed from the thinking that produced it.