📘 Write a textbook chapter, start to finish¶

Status: tutorial. Written 2026-09-15.

Written for a lecturer or course author who wants a chapter of teaching material -- learning objectives, motivation, worked examples, exercises -- and who has not used this pipeline before. Assumed: nothing. This page repeats what other documents also say, deliberately. Not covered here: why each prose rule exists (WRITING-STANDARDS.md) and how a whole book is assembled from chapters (WRITE-A-BOOK.md).

Sister tutorials: a survey, a thesis chapter, a tutorial, a deep-research report.

🧭 Table of contents¶

Is this the genre you want?
What you will have at the end
Before you start
Step 1: the student, the objectives, the slug
Step 2: open the dossier
Step 3: write an outline (optional, recommended)
Step 4: ask for the chapter
Step 5: gate, references, render
Step 6: read the review aids
Step 7: change something
Step 8: chapters into a book
When something goes wrong

⚖ Is this the genre you want?¶

A textbook chapter is for a student who is studying the topic, not yet doing it. It explains, then shows worked applications.

If your reader will be sitting at a keyboard following your steps to a working result, you want a tutorial instead. If your reader is a researcher mapping a field, you want a survey.

This genre is not citation-dense. Most of the content is your own worked examples and exercises; the corpus is there for motivation and background. Citations are still gated -- any that appear must be real -- but a chapter with three citations is normal, not thin.

🎯 What you will have at the end¶

For a chapter you decide to call course/ch3-state-estimation:

Path	What it is
`content/drafts/course/ch3-state-estimation.md`	the chapter -- the canonical copy
`content/dossiers/course/ch3-state-estimation/`	scope, objectives, any evidence kept, every search run
`content/rendered/course/ch3-state-estimation.pdf`	the typeset chapter
`content/rendered/course/ch3-state-estimation.tex`	the same, as LaTeX, for your course template
`content/rendered/course/ch3-state-estimation.md`	a numbered Markdown copy

🔧 Before you start¶

pip install chitragupta-cli
chitragupta init my-course
cd my-course

mkdir -p papers
cp /path/to/your-library.bib papers/bibliography.bib

chitragupta corpus sync
chitragupta corpus ledger

Or clone the repository and cp config.toml.example config.toml instead of the first two lines.

A textbook chapter can be written against a small corpus -- you need enough to motivate the topic and point students onward, not a full literature map. If you have no library at all, the chapter can still be drafted; it simply cites nothing, and the gate has nothing to check.

Every command here also works as python -m chitragupta.<layer> ....

🎓 Step 1: the student, the objectives, the slug¶

The student is your reader, stated concretely: "second-year undergraduates who have had one linear-algebra course and no signals course". Everything -- what is assumed, what is recapped, how fast the worked examples move -- follows from that sentence.

The learning objectives are three to five, each concrete and testable. "By the end of this chapter a student can..."

...state what a state estimator does and why a raw measurement is not enough.
...derive the one-dimensional Kalman update from the two variances.
...hand-compute two steps of the filter on given numbers.
...say when the assumptions fail and what happens then.

Write these before anything else. They are the chapter's contract, and the exercises at the end are how a student checks they hold.

The slug is the path under content/drafts/:

You are writing	Use
one chapter of a course	`course/ch3-state-estimation`
a chapter of a book being assembled	`books/control/ch3-state-estimation`
a standalone handout	`state-estimation-handout`

🗂 Step 2: open the dossier¶

chitragupta draft dossier init \
    content/drafts/course/ch3-state-estimation.md --genre textbook-chapter

That writes eight files. Exactly one is yours to fill in: scope.md. The rest -- evidence.md, rejected.md, sections.md, retrieval.md, steering.md, revisions.md, README.md -- are written for you as the chapter is produced.

What goes in `scope.md`¶

init writes it with every heading present and empty. Five things are yours:

Field	What goes in it	Why it is asked for
`- language:`	a BCP-47 tag: `en-GB`, `en-US`, `en-IN`	ships unset; a chapter whose dialect nobody chose silently gets the model's own
`## Reader`	the student, concretely, including the courses they have had	decides what is assumed, what is recapped, and how fast the worked examples move
`## Covers`	the chapter's content and its learning objectives	this genre has no other home for the objectives, and they are the chapter's contract
`## Does not cover`	what is deferred to a later chapter or a lab	students who see a term with no treatment assume they missed it
`## Glossary`	each recurring term with one definition	a term that drifts between sections is a student's lost afternoon

Set the dialect with the command, so the format is right:

chitragupta draft dossier set-language \
    content/drafts/course/ch3-state-estimation.md en-GB

A filled-in textbook-chapter scope.md -- the whole file is at examples/dossiers/textbook-chapter/scope.md:

# Scope

- genre: textbook-chapter
- language: en-GB
- draft: content/drafts/course/ch3-state-estimation.md
- created: 2026-09-15
- corpus: 214 citekeys, digest `a31f0c4b77de`
- draft digest: not recorded (run `dossier stamp` once the draft is ready)

## Reader

Second-year undergraduates who have had one linear-algebra course and no
signals or probability course beyond a first module. They have seen a
mean and a variance; they have not seen a covariance matrix, and will
not in this chapter.

## Covers

Why a raw measurement is not enough; the one-dimensional estimator built
from two variances; two steps worked by hand; a faded example the
student finishes; and an honest section on which assumptions break.

Learning objectives -- by the end a student can:

1. state what a state estimator does and why a raw reading is not enough;
2. derive the one-dimensional update from the two variances;
3. hand-compute two steps on given numbers;
4. say when the assumptions fail and what happens then.

## Does not cover

The matrix form. That is chapter 4, and reaching for it here would cost
the derivation its arithmetic-only property.

Implementation. There is no code in this chapter -- students who want to
build one are pointed at the lab, which is a tutorial.

## Glossary

- **State** -- the quantity we want to know and cannot measure directly.
  Never "the system" or "the value".
- **Estimate** -- our current best guess of the state, always paired
  with its variance. A number without its variance is never called an
  estimate in this chapter.
- **Gain** -- the weight given to a new measurement against the current
  estimate, between 0 and 1.

Putting the objectives in ## Covers is the convention for this genre: they are scope, they are what the exercises test, and a later revision that quietly drops one is then visible.

🗺 Step 3: write an outline (optional, recommended)¶

A textbook chapter has a shape students expect, and declaring it is cheap:

chitragupta draft dossier init \
    content/drafts/course/ch3-state-estimation.md \
    --genre textbook-chapter --outline

Edit content/dossiers/course/ch3-state-estimation/outline.md. Three fields per section:

Field	What goes in it	What the skill does with it
`brief:`	steering in your own words -- what to emphasise, what to leave out, how long	consumed once, never appears in the chapter
`claim:`	your own prose, where you want a specific statement made	rewritten and grounded; anything the corpus cannot support is reported rather than shipped
`queries:`	a `-` list of search terms	run verbatim instead of the skill inventing sub-themes

A section needs at least a brief: or a claim:. In this genre most sections have only a brief: and no queries: at all -- a worked example and a set of exercises are yours to design, not the corpus'. An outline here that is almost all briefs is correct, not thin.

A worked example -- the whole file is at examples/dossiers/textbook-chapter/outline.md:

## Learning objectives

brief: The four objectives, verbatim as agreed. No prose around them.

## Why a raw measurement is not enough

brief: Motivate before mechanism. One concrete scenario -- a robot whose
wheel encoder drifts -- carried through the whole chapter.

queries:

- state estimation sensor noise motivation

## The idea in one dimension

brief: Build the update from two variances. Arithmetic only, no matrix
algebra -- that is chapter 4.

## Worked example: two steps by hand

brief: Fully worked, every number shown. Then a faded version where the
student supplies the second step.

## When the assumptions fail

brief: Non-Gaussian noise and an unmodelled bias. Honest about what
breaks; this is where students who go on to research start.

queries:

- kalman filter assumption violation bias

## Exercises

brief: Six, mixed difficulty, hints for the last two. At least one that
cannot be done by pattern-matching the worked example.

## Summary and where to go next

Validate:

chitragupta draft dossier outline \
    content/drafts/course/ch3-state-estimation.md --check

🗣 Step 4: ask for the chapter¶

Draft a textbook chapter on one-dimensional state estimation into content/drafts/course/ch3-state-estimation.md, for second-year undergraduates with one linear-algebra course. The objectives and outline are in the dossier.

"Textbook chapter", "lecture notes", "course reader" or "worked-examples handout" selects textbook-chapter-writer.

What it does: establishes the objectives, states scope and prerequisites, motivates the need before the mechanism, writes worked then faded examples, writes exercises with hints or solutions, closes the loop with a summary, and reads the whole thing back as the student before handing it over. Where it cites, it cites real citekeys from your corpus only.

✅ Step 5: gate, references, render¶

chitragupta draft gate content/drafts/course/ch3-state-estimation.md

If the chapter cites nothing, the gate passes trivially -- that is expected for this genre, not a warning sign.

If it does cite, build the references section from exactly the gated keys:

chitragupta draft references content/drafts/course/ch3-state-estimation.md

Render:

chitragupta draft render \
    content/drafts/course/ch3-state-estimation.md --format pdf
chitragupta draft render \
    content/drafts/course/ch3-state-estimation.md --format tex
chitragupta draft render \
    content/drafts/course/ch3-state-estimation.md --format md

Then stamp:

chitragupta draft dossier stamp \
    content/drafts/course/ch3-state-estimation.md

🔍 Step 6: read the review aids¶

All advisory, all exit 0.

chitragupta draft style content/drafts/course/ch3-state-estimation.md
chitragupta review verbatim scan \
    content/drafts/course/ch3-state-estimation.md
chitragupta review figure content/drafts/course/ch3-state-estimation.md

What each is for here:

Aid	Reads for	Why it matters in teaching material
`review verbatim`	wording shared with any parsed source, cited or not	the one that matters most. A chapter that reuses a source's wording is a copyright problem in a way a private research note is not
`draft style`	defect markers, an acronym never expanded at first use, dialect	a student meets every acronym for the first time; an unexpanded one costs them the paragraph
`review figure`	overlapping nodes, overlong labels, and the edge list	confirm the edge list against your own prose -- a diagram that wires the concepts up wrongly teaches the wrong thing convincingly

review uncited is deliberately gentler on this genre: prose with no citation is the normal state of a textbook, not a finding.

The agenda: all of them as one worklist¶

AGENDA.md explains every section of an agenda file in full; what follows is the short version for this genre.

chitragupta review agenda content/drafts/course/ch3-state-estimation.md

It reads the aids' filed JSON and never runs an aid, so run them first with --write. Items are [unattended] (safe to repair automatically: prose, short verbatim runs, missing-citekey) or [surfaced] (yours to judge).

A chapter that cites lightly produces a short agenda, and that is the expected shape here:

# Agenda: content/drafts/course/ch3-state-estimation.md

> **Review aid, not a gate.** This report is evidence for a human
> judgement, never a verdict. No draft is blocked by what it says.

- Draft: `content/drafts/course/ch3-state-estimation.md`
- Command: `chitragupta review agenda content/drafts/course/ch3-state-estimation.md`

## Sources

- Citation provenance: read
- Verbatim scan: read
- Citation coverage: not run
- Multi-source synthesis: read, no item class defined
- TikZ layout check: read
- Uncited prose: read
- Quotation integrity: not run
- Claim support: read
- Prose (style_check): read
- Dossier drift: read

## Summary

- 3 prose
- 1 verbatim-run

## Findings

### prose

- `5a8bf91ec244` [unattended] (2. Why a raw measurement is not enough):
  chitragupta.AcronymNotExpanded: 'RMS' used before first expansion
- `c14d0b7e9a33` [unattended] (5. When the assumptions fail):
  chitragupta.DefectMarker: 'obviously' (1x)
- `7e2a55c1b09d` [unattended] (6. Exercises): chitragupta.DialectDrift:
  'analyze' -- scope.md records en-GB

### verbatim-run

- `1561cffaa101` [unattended] (2. Why a raw measurement is not enough):
  10-word verbatim run citing `bar_shalom_estimation_2001`

How to read that as the author. All four are [unattended], which is typical for a chapter whose content is mostly your own: an acronym a student meets before its expansion, a hedge word, a US spelling in a chapter that declared en-GB, and one phrase that matches a textbook you cited. Hand them off -- ask to "work the review agenda" and agenda-reviser repairs them one at a time, re-running the gate after each and logging every attempt in revisions.md.

The DialectDrift finding is only possible because scope.md records language: en-GB. Leave that line unset and this whole class of finding silently does not exist.

To check a round of edits helped:

chitragupta review agenda content/drafts/course/ch3-state-estimation.md \
    --baseline content/review/course/ch3-state-estimation.agenda.json

That re-runs the aids and reports each finding as resolved, persisting, new or accepted -- read the new list, not just the count.

📝 Step 7: change something¶

Never re-run the genre skill. Ask for a revision:

Exercise 4 is harder than the worked example prepares them for. Replace it with one that uses the same numbers as the worked example, and move the current one to the end as a challenge.

That selects draft-reviser, which edits only what is affected and logs it in revisions.md.

After a hand edit:

chitragupta draft gate content/drafts/course/ch3-state-estimation.md
chitragupta draft dossier stamp \
    content/drafts/course/ch3-state-estimation.md

Back it up -- content/drafts/ is gitignored:

chitragupta draft dossier export course/ch3-state-estimation

📗 Step 8: chapters into a book¶

If this chapter is one of many, the book track keeps them consistent: a signed outline (chitragupta draft spec), per-chapter acceptance records (chitragupta draft unit), a cross-reference check (chitragupta draft registry check) and an assembly step that produces one LaTeX book. Start at WRITE-A-BOOK.md before drafting chapter two -- retrofitting the outline afterwards costs more than declaring it now.

🚑 When something goes wrong¶

What you see	What it means	What to do
The chapter cites nothing	normal for this genre	nothing; the gate passes trivially
The chapter is citation-heavy and reads like a survey	the wrong genre ran	ask for a textbook chapter explicitly, or revise toward worked examples
Worked examples feel abstract	the motivating scenario was never fixed	give one concrete scenario and ask `draft-reviser` to carry it through
`gate` says `FAIL`	a citekey is not in the corpus	correct it or drop the claim
`[missing-binary]` from `render`	no `pandoc`/`pdflatex`	install them; the `.md` chapter is unaffected
Students need to do rather than study	you want the other genre	see WRITE-A-TUTORIAL.md