Skip to content

DERIVATION.md

Latest commit

 

History

History
140 lines (113 loc) · 10.3 KB

DERIVATION.md

File metadata and controls

140 lines (113 loc) · 10.3 KB
Raw
title Derivation: from notes to an essay
status draft
date 2026-10-14
related
SMART_NOTE_POLICY.md
notes/index.md
essays/index.md
.claude/skills/essay-montaigne/SKILL.md
AGENTS.md

Derivation: from notes to an essay

A note holds one idea (SMART_NOTE_POLICY.md). An essay is several notes tried against one question. This page is the short way from the first to the second. It was written while the first essay was derived, New vs. better, 2026, and the last section shows that derivation as it was done.

The rule that makes it short

Nothing new is written at the essay. Every fact, every quotation and every "I" is already in a note, with its status. The essay only chooses, orders and joins. If the essay needs something that no note holds, stop and write the note first, or leave a marker.

That is why it is quick: the slow work (to fetch a source, to check a number, to ask the owner) was done once, at the note.

Seven steps

  1. One question. Write it in one line. It comes from a literature note: somebody else's sentence that the essay will try. No question, no essay.
  2. Pull the notes. List every note that touches the question: follow the links of the first one, then read notes/index.md by section. Five to twelve notes is the right size.
  3. Sort them into three piles.
    • the case: one or two fleeting notes with something that happened to the writer. The essay opens with it.
    • the borrowing: the literature notes. They enter after the writer has spoken.
    • the trials: the permanent notes, each a position of the writer. One paragraph each.
  4. Find the gaps. Set the brief beside the piles. A topic of the brief with no note is a gap. A gap of fact is closed by a fetch, and the fetch is recorded in the essay's sources. A gap of experience is not closed by an agent: it becomes **[AUTHOR ANECDOTE: ...]**.
  5. Write it once, in order: the case, the borrowing, the trials, the doubt. First person, plain words, no heading inside the body, no lesson at the end. Keep each note's status: a note that is unverified enters as memory or as opinion, never as fact.
  6. Check it. scripts/essay-montaigne.sh <file>, then the reader's part of the essay-montaigne skill. Fix the essay, never the check.
  7. Close the loop. The essay ends with ## Sources, one line for each note and each fetch. essays/index.md gets its row, and each note that was used gets a link to the essay when it is next edited. The status is draft while a marker is open; only the owner moves it on.

What costs time, and how to not pay it

It costs Do this
reading every note for every essay start from one note and follow links; read the index, not the folder
checking a fact twice take the status from the note; check again only what the note marks unverified and the essay needs as a fact
an essay that grows one question, one paragraph for each trial, and stop: an essay never treats the whole of a thing
writing the owner's life for him a marker takes one line; an invented memory costs the whole essay
a second draft to add the sources carry each note's id beside its paragraph while you write, and turn the ids into ## Sources at the end

The first derivation, as it was done

The question (step 1): Pirsig's two questions, from the note the Chautauquas. The owner's brief: an analogy for 2026, with AI, El Niño, social media, Bitcoin and privacy.

The notes, in their piles (steps 2 and 3):

Pile Note Status What the essay took
the case the first Scala partial Scala learned because a tool failed
the case Marcus Baffa laughed with me unverified the Latin classes, and "with me"; the essay does not name him
the borrowing the Chautauquas partial the two questions, quoted
the borrowing Pirsig, the legend partial who is speaking, and the year 1974
trial: AI why I wrote fast partial "everybody talks about one subject"
trial: AI the port to Gleam partial 8,222 lines, and "the same output"
trial: El Niño exploit in operations, explore in simulation open question the rule for a forecast
trial: social media seen by the right people verified the rule of the right reader
trial: privacy not against property open question the book is not against what is private
trial: privacy rule 8 of the policy, PERSONAS.md a rule friends under an alias

The gaps (step 4):

Topic of the brief What was missing How it was closed
El Niño the state of El Niño in 2026 fetched: NOAA's ENSO Diagnostic Discussion of 8 October 2026
Bitcoin everything: no note speaks of it one fetched sentence of its paper, and a marker for the owner's position
social media the owner's own case a marker

The check (step 6): the script's eight rules hold, with a note that two markers are open. So the essay is a draft, and it says so in its first lines.

What the derivation showed. Eight of the ten sources were already notes. The two fetches and the two markers are exactly the topics that no note covered: the gaps of the essay are the gaps of the slip box. The next note to write is the owner's on Bitcoin.

The second derivation: a short essay for one page

Peer reviewing: the god that failed, 14 October 2026. The log of it, step by step:

Step What was done
1. the question the owner's sentence: peer review has failed, and it is a ritual (the note)
2, 3. the notes the case: why I wrote fast (a book with no review). The trial: what is MetaQuality (a check that nobody can open)
4. the gaps, closed by a fetch the history of peer review (one encyclopedia page); the journal business (one article of The Guardian); Kubrick's last film (one encyclopedia page). All three are quoted in the note first
4. the gaps, left open the links that the owner asked for from this business to Jeffrey Epstein and from him to named people, services and crimes. No source was fetched, so none is in the note and none is in the essay
5. written 309 words. Three sentences of the first draft were removed because no note and no fetched source held them: two details of the film, and "I have never been a referee"
5. where the record changed the essay the owner's sentence was "it started in the wrong way, by the wrong person". The record names a first referee in 1665 and a businessman in the 1950s. The essay says both, in the first person, and says that it cannot join them
6. checked scripts/essay-montaigne.sh: the eight countable rules hold, no open marker
7. printed just print: one A5 page, no text that can be selected (PRINTING.md)

What the second derivation showed. A short essay has room for one case, one borrowing, one image and one doubt. A fifth thing does not fit the page, and the page is the check.

The third derivation: the note first, with a study inside it

picasso, and the long run, 14 October 2026. The owner asked for the note first and the essay from it, and for the process to be explained.

Step What was done
0. the study the owner's thesis had facts in it that no note held. An agent (Claude, the Fable model, high effort, as the owner asked) read picasso, gvolpe/nix-config and Determinate's repositories at their commits, fetched the incident reports and the Nix manual, and wrote a report. The report is not in this repository
0. the check of the study a report by an agent is a claim, not a source. Six of its quotations were fetched again by the agent that filed the note: PyPI's three, Red Hat's two, the limit of Nix, the CVE's advisory, and two facts of picasso's files. All held. What was not checked again is marked in the note
1. the question the owner's sentence: "picasso may be the first profitable open source"
2, 3. the note one note, picasso: determinism, and the long run to profit, holds everything: the mess, the limit, the three projects, the two claims against the record, the plan as REA. The essay stands on this note alone
4. the gaps "first" and "beat": the record contradicts the first and gives no test for the second. They were not closed. They became the essay's matter
5. written the essay tries the owner's sentence a word at a time and takes out the words that do not hold. A guess of the first draft ("the star may be mine") was cut: no note held it
5. what the derivation found picasso's CI calls its installer by a tag, the same weakness as the attack that the note quotes. It came from reading the owner's own files with the essay's question
6. checked scripts/essay-montaigne.sh: the eight countable rules hold
7. printed just print: two pages, declared in the front matter

What the third derivation showed. When a thesis is contradicted by the record, the essay does not hide it and does not drop the thesis: an essay is a trial, and a trial may fail in part. The note keeps the verdicts; the essay keeps the "I".

Sources

  • The slip box and the three kinds of note: SMART_NOTE_POLICY.md, after Luhmann as Ahrens describes him.
  • The order "the writer first, the borrowing after": Montaigne, Essays, II, "Of Books", in the essay-montaigne skill.
  • The marker: AGENTS.md, invariant 3.