Summary

Style audit of the lecture-python-intro series.

• Audit date: 2026-08-23

• Corpus snapshot: a12d17c0ef

• Lectures audited: 56

• Average overall score: 8.1 / 10

• Average per-category scores: writing 5.2, math 8.6, code 7.3, figures 6.5, references 9.3, links 9.7, admon 10.0

• JAX: out of scope — the qe-jax-* rules target lecture-jax.

• Judgment-review coverage: all lectures reviewed.

Writing, at 5.2, is the weakest category here, and it is very nearly the whole HIGH list. Eighteen of the 19 HIGH lectures are HIGH only because Writing sits at or below the ≤ 4 floor; the nineteenth, markov_chains_I, is floored by Math (3.0) instead. Three of the 18 — geom_series, french_rev and eigen_I — are also floored by Figures, and no lecture in the series is floored by Code, References, Links or Admonitions at all.

Underneath Writing are four rules, and no single one of them accounts for a lecture. qe-writing-008 (39 / 56, 709 occurrences of repeated spaces) is the largest raw count and the only one that sweeps cleanly, but every one of the 18 floored lectures also breaks at least one rule that has to be read: qe-writing-001 (30 / 56) in 16 of them, qe-writing-006 (13 / 56) in 11, qe-writing-004 (18 / 56) in 9. Across the 18 that is 85 hand-edited sites, against 338 whitespace occurrences a script handles. The 2026-05 row put Writing at 7.5 here and left Figures as the weakest category. The reordering since is the judgment layer — that row folds in no review overlay, this one folds one into all 56 lectures — not the lectures: like for like, on the evidence layer alone (history_mechanical.csv), Writing here is 7.5 → 7.7 and Figures 6.5 → 6.5, so on a comparable basis Figures is still the weakest category.

By every other measure this is the strongest series in the corpus: the highest overall (8.1 against 7.7), the smallest share of HIGH lectures (19 of 56, against 197 of 348), nothing below 7.2 — only lecture-python-programming also has no lecture under 7.0 — and neither a build-risk violation nor any of the corpus’s four structural findings. Math especially is not the problem: 8.6, second only to lecture-python-programming, against 7.0 corpus-wide, with qe-math-002 (4 / 56, 13 occurrences) where lecture-python-advanced.myst has it in 20 of its 68. Figures (6.5) sits exactly on the corpus mean and Code (7.3) just under it.

Priority distribution

Priority	Count	%
HIGH	19	33.9%
MEDIUM	0	0.0%
LOW	28	50.0%
NONE	9	16.1%

Top systemic issues across the series

Ranked by how many of the series’ lectures each rule reaches.

1. qe-fig-005 — Descriptive figure names for cross-referencing — 46 / 56 lectures, 174 occurrences.

2. qe-writing-008 — Remove excessive whitespace between words — 39 / 56 lectures, 709 occurrences.

3. qe-fig-008 — Use lw=2 for line charts — 35 / 56 lectures, 217 occurrences.

4. qe-fig-001 — Do not set figure size unless necessary — 30 / 56 lectures, 91 occurrences.

5. qe-writing-001 — Use one sentence per paragraph — 30 / 56 lectures, 54 occurrences.

6. qe-writing-004 — Avoid unnecessary capitalization in narrative text — 18 / 56 lectures, 40 occurrences.

7. qe-fig-004 — Caption formatting conventions — 17 / 56 lectures, 72 occurrences.

8. qe-fig-003 — No matplotlib embedded titles — 15 / 56 lectures, 36 occurrences.

9. qe-ref-001 — Use correct citation style — 15 / 56 lectures, 48 occurrences.

10. qe-writing-006 — Capitalize lecture titles properly — 13 / 56 lectures, 31 occurrences.

Clean across the series

Checked rules with no violation anywhere in the series — the conventions this series already holds to.

• qe-admon-003 — Use tick count management for nested directives

• qe-code-005 — Use quantecon timeit for benchmarking

• qe-fig-010 — Plotly figures require latex directive

• qe-math-006 — Use aligned environment correctly for PDF compatibility

• qe-math-007 — Use automatic equation numbering, not manual tags

• qe-math-008 — Explain special notation (vectors/matrices)

• qe-math-013 (proposed) — Reference equations via {eq}`label`

Series-level recommendations

1. qe-writing-008 — collapse repeated spaces (39 / 56, 709 occurrences). Free and scriptable, so do it first; 338 of the occurrences are inside the 18 Writing-floored lectures, 89 of them in french_rev alone. On its own it clears no HIGH lecture — each of the 18 also breaks a rule that needs reading.

2. qe-writing-006 — sentence-case the headings (13 / 56, 31 headings). The best ratio of floored lectures reached to work done in the series: it reaches 11 of the 18, and 29 of the 31 headings are in those 11. Scriptable, but the check is heuristic and leans on a curated proper-noun list, so read the diff before committing it.

3. qe-writing-004 — drop the mid-sentence capitals (18 / 56, 40 sites). Reaches 9 of the 18, concentrated in time_series_with_matrices (7 sites) and eigen_I (5). Same heuristic caveat, and small enough to check by hand.

4. qe-writing-001 — one sentence per paragraph (30 / 56, 54 blocks). The largest single blocker — 16 of the 18 — but a reading pass rather than a sweep, because splitting a paragraph changes its rhythm. 31 of the 54 blocks sit in the floored lectures.

5. Take the floored lectures cheapest first. On top of the sweeps, observed_distributions and laffer_adaptive need one hand-edited site each, msy_fishery two (and no whitespace work at all), then cagan_ree, networks, solow, tax_smooth, unpleasant and greek_square at three each. The tail is time_series_with_matrices (9 sites), complex_and_trig (8) and french_rev (8).

6. qe-fig-005 — name the figures (46 / 56, 174 figures). The widest-reaching rule in the series, a pure sweep, and it unlocks {numref} cross-referencing — but it lifts the series average, not the HIGH list. The three Figures floors are geom_series, french_rev and eigen_I; qe-fig-005 is only 14 of their 135 figure findings, and all three are floored by Writing regardless. qe-fig-008 (35 / 56, 217 calls) and qe-fig-001 (30 / 56, 91 overrides) are the same kind of item: broad, cheap, no effect on priority.

7. markov_chains_I — the one Math floor (3.0). 22 of its 23 math findings are qe-math-004 (4 / 56, 15 occurrences) and qe-math-010 (proposed) (8 / 56, 100 occurrences), and the latter is still a proposed rule rather than a registry one. So this lecture’s priority turns partly on action-style-guide #18; qe-math-004 is worth doing either way.

8. Do the work upstream, in QuantEcon/lecture-python-intro. This repository only measures. One file needs care about where the edit lands: short_path is byte-identical with the lecture-dp copy at this snapshot, so fixing it once clears both series. It is the only substantive shared file here — status and zreferences also match another series byte for byte but are boilerplate, while lake_model, lln_clt and mle share a filename with lecture-python.myst and have diverged, so each of those copies needs its own edit.

Lectures ranked by priority (lowest score first)

Scores are 0–10 per category; Overall is the mean of the in-scope categories, and Priority follows spec §4. A dash means the category is not applicable to that lecture. Click a lecture for its full report.

#	Lecture	Writing	Math	Code	Figures	References	Links	Admon	Overall	Priority
1	geom_series	3	8.5	7.5	4	—	10	10	7.2	HIGH
2	linear_equations	5.5	6.5	7.5	6	—	7.5	10	7.2	LOW
3	networks	4	6	7	5	8.5	10	10	7.2	HIGH
4	french_rev	3	10	7.5	3	7.5	10	10	7.3	HIGH
5	eigen_I	3.5	10	7.5	3.5	—	10	10	7.4	HIGH
6	heavy_tails	6	5.5	6	5	10	9	10	7.4	LOW
7	inflation_history	3	10	6	4.5	8.5	10	10	7.4	HIGH
8	markov_chains_I	6	3	7	7.5	9	9	10	7.4	HIGH
9	time_series_with_matrices	3	6	7.5	7	10	8	10	7.4	HIGH
10	business_cycle	6	—	7	7	—	10	—	7.5	LOW
11	bivariate_dist	6.5	5.5	6	5	10	10	10	7.6	LOW
12	mobility	4.5	6	8.5	7.5	7	10	10	7.6	LOW
13	greek_square	4	7.5	7	6.5	9	10	10	7.7	HIGH
14	monte_carlo	5.5	5	7.5	8	—	10	10	7.7	LOW
15	inequality	4	9	6.5	5	10	10	10	7.8	HIGH
16	intro_supply_demand	4.5	10	7.5	6.5	—	8	10	7.8	LOW
17	lln_clt	8	4.5	7.5	8	—	9	10	7.8	LOW
18	long_run_growth	5	—	7.5	6	8.5	10	10	7.8	LOW
19	simple_linear_regression	4.5	8.5	7.5	6.5	—	10	9.5	7.8	LOW
20	solow	4	8	7	8	—	10	10	7.8	HIGH
21	tax_smooth	3	9.5	6	6	10	10	10	7.8	HIGH
22	ar1_processes	6	7.5	6	7	9	10	10	7.9	LOW
23	complex_and_trig	3	9.5	7	5.5	10	10	10	7.9	HIGH
24	equalizing_difference	5	8	7.5	5.5	9	10	10	7.9	LOW
25	laffer_adaptive	4	10	7	6	8.5	10	10	7.9	HIGH
26	lake_model	4.5	10	7.5	5.5	—	10	10	7.9	LOW
27	money_inflation_nonlinear	3	9.5	6.5	6.5	10	10	10	7.9	HIGH
28	olg	4.5	9.5	7	6.5	10	8	10	7.9	LOW
29	prob_dist	5	7	7	8.5	—	10	10	7.9	LOW
30	unpleasant	3.5	8	7.5	6	10	10	10	7.9	HIGH
31	lp_intro	3.5	6.5	7.5	8.5	10	10	10	8.0	HIGH
32	msy_fishery	3.5	10	6	6.5	10	10	10	8.0	HIGH
33	supply_demand_multiple_goods	4.5	9.5	6.5	7.5	—	10	10	8.0	LOW
34	cons_smooth	4.5	10	7.5	6	9	10	10	8.1	LOW
35	cagan_adaptive	5	10	7	6.5	9	10	10	8.2	LOW
36	mle	5	10	7	7.5	—	10	10	8.2	LOW
37	money_inflation	4.5	9.5	7.5	6	10	10	10	8.2	LOW
38	scalar_dynam	4.5	10	6	8.5	—	10	10	8.2	LOW
39	supply_demand_heterogeneity	4.5	9.5	7	—	—	10	10	8.2	LOW
40	pv	4.5	10	8.5	7	—	10	10	8.3	LOW
41	eigen_II	4.5	9.5	7.5	—	9	10	10	8.4	LOW
42	markov_chains_II	6.5	10	7.5	6.5	8.5	10	10	8.4	LOW
43	observed_distributions	4	10	7.5	7	10	10	10	8.4	HIGH
44	about	8	—	—	—	—	9	—	8.5	LOW
45	cagan_ree	4	10	8.5	7	10	10	10	8.5	HIGH
46	input_output	7.5	10	7.5	6	8.5	10	10	8.5	LOW
47	short_path	5.5	10	7.5	8	—	10	10	8.5	LOW
48	schelling	5	10	7	9	10	9	10	8.6	NONE
49	bayes_intro	7.5	9.5	10	5.5	—	10	10	8.8	NONE
50	cobweb	7.5	10	7.5	6.5	10	10	10	8.8	NONE
51	troubleshooting	8.5	—	—	9	—	9	—	8.8	NONE
52	commod_price	7.5	10	9	8	10	10	10	9.2	NONE
53	fitting_distributions	6.5	10	10	8.5	—	10	10	9.2	NONE
54	intro	10	—	—	—	—	10	—	10.0	NONE
55	status	10	—	—	—	—	10	—	10.0	NONE
56	zreferences	10	—	—	—	—	10	—	10.0	NONE