Summary#

Style audit of the lecture-dp series.

Audit date: 2026-08-07
Corpus snapshot: c30490a2f4
Lectures audited: 52
Average overall score: 7.7 / 10
Average per-category scores: writing 4.7, math 6.6, code 7.7, figures 6.4, references 9.3, links 9.5, admon 10.0
JAX: out of scope — the qe-jax-* rules target lecture-jax.
Judgment-review coverage: all lectures reviewed.

Writing is the floor under this series. At 4.7 it is the weakest of the seven categories measured here, and it sits at or below the ≤ 4 threshold in 30 of the 34 HIGH lectures — against Math’s 14 and Figures’ 2. Nineteen lectures are HIGH on Writing and nothing else. In 2026-05 Math and Figures tied as the weakest at 6.4 and Writing scored 7.0; the fall to 4.7, and the HIGH list growing from 15 to 34, is the judgment layer landing, not the lectures changing — the 2026-05 row folds in no review overlay, this one folds one into all 52 lectures, and a lecture assessed against more rules scores lower. Like for like, on the evidence layer alone (history_mechanical.csv), Writing here is 7.0 → 7.0, Math 6.4 → 6.9, Figures 6.4 → 6.4 and the HIGH count 15 → 15. So the floor is new to what is measured, not a regression; at 4.7 lecture-dp sits marginally above the corpus Writing mean of 4.6 on the same basis. Every one of the 34 is HIGH on a category floor, never on the overall — the lowest overall in the series is 5.7. The series mean, 7.7, is exactly the corpus mean, but 65.4 % of its lectures are HIGH against 57 % corpus-wide — the second-largest share of any series, behind lecture-python-programming.

What sits under Writing divides into a scriptable half and a reading half, and the reading half is the larger one. Repeated spaces dominate the mechanical side: qe-writing-008 (40 / 52, 1,578 occurrences), a per-lecture density second only to lecture-python-advanced.myst. Then qe-writing-001 (23 / 52, 45), qe-writing-006 (22 / 52, 146) and qe-writing-004 (14 / 52, 28). By reading: qe-writing-003 (logical flow) reaches 45 of the 52 lectures, qe-writing-002 39, qe-writing-007 36, qe-writing-005 31. All 30 Writing-floored lectures carry findings from both halves — not one of them can be cleared by a sweep alone. That is what separates this series from lecture-python-programming, where a single heading sweep would empty the HIGH list.

Math is the second problem and a far more concentrated one: at 6.6 it is the lowest in the corpus after lecture-python-advanced.myst, and nearly all of it is one rule. qe-math-002 — transpose notation — (14 / 52, 416 occurrences) reaches fewer lectures than four of the figure rules but averages about thirty sites in each, and it lands where it counts: 12 of the 14 Math-floored lectures carry it, holding 381 of its 416 occurrences. lqcontrol alone has 85, lagrangian_lqdp 69, tax_smoothing_2 59, cross_product_trick 52, markov_jump_lq 47. Past it the Math findings thin out quickly: the next largest is qe-math-010 (20 / 52, 132), itself a proposed rule — so this is one rule and not a cluster.

A caveat specific to this series. 31 of these 52 lectures share a filename with a lecture in lecture-python.myst, the repository lecture-dp syncs from — but only 6 are byte-identical at this snapshot (cross_product_trick, ifp_discrete, ifp_opi, lq_inventories, mccall_model_with_separation, os_numerical). For those 6 a finding here and a finding there are the same finding, and one upstream fix clears both; the other 25 share an origin and have since diverged, so the same defect usually needs two edits. Only two of the 6 are HIGH here — cross_product_trick (5.7, the series’ lowest) and lq_inventories (7.4). It is worth recording how the two independent passes over identical bytes came out: all six pairs differ in at least one judgment-heavy category — Writing in four, Code in two, Math in one — while Figures, References, Links and Admonitions agree exactly, no overall differs by more than 0.2, and all six land in the same priority bucket. Whether the corpus totals should de-duplicate those six is open — compliance-lecture-style#3.

Priority distribution#

Priority	Count	%
HIGH	34	65.4%
MEDIUM	0	0.0%
LOW	9	17.3%
NONE	9	17.3%

Top systemic issues across the series#

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

qe-fig-005 — Descriptive figure names for cross-referencing — 42 / 52 lectures, 164 occurrences.
qe-writing-008 — Remove excessive whitespace between words — 40 / 52 lectures, 1578 occurrences.
qe-fig-008 — Use lw=2 for line charts — 38 / 52 lectures, 219 occurrences.
qe-fig-001 — Do not set figure size unless necessary — 31 / 52 lectures, 102 occurrences.
qe-fig-003 — No matplotlib embedded titles — 30 / 52 lectures, 105 occurrences.
qe-writing-001 — Use one sentence per paragraph — 23 / 52 lectures, 45 occurrences.
qe-writing-006 — Capitalize lecture titles properly — 22 / 52 lectures, 146 occurrences.
qe-math-010 (proposed) — Blackboard \mathbb{P}, \mathbb{E}, \mathbb{V} with braces — 20 / 52 lectures, 132 occurrences.
qe-ref-001 — Use correct citation style — 20 / 52 lectures, 45 occurrences.
qe-link-002 — Use doc links for cross-series references — 15 / 52 lectures, 53 occurrences.

Clean across the series#

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

qe-admon-002 — Use dropdown class for solutions
qe-admon-003 — Use tick count management for nested directives
qe-fig-004 — Caption formatting conventions
qe-fig-007 — Keep figure box and spines
qe-fig-010 — Plotly figures require latex directive
qe-link-001 — Use markdown style links for lectures in same lecture series
qe-math-012 (proposed) — Multiplication via \cdot or juxtaposition, never *

Series-level recommendations#

Ordered by HIGH lectures cleared per unit of work, not by size of the finding.

qe-math-002 on the four HIGH lectures Writing does not floor — tax_smoothing_2 (59 occurrences), markov_jump_lq (47), tax_smoothing_1 (11), lqramsey (10). These are the only lectures in the series where Math work on its own can move the bucket; every other HIGH lecture is also floored by Writing and stays HIGH until a reading pass lands there. Mostly a mechanical ' → \top rewrite, but the check has to tell a transpose apostrophe from a derivative, so each site wants an eye. tax_smoothing_2 needs its Figures floor cleared too — qe-fig-003 (9 embedded titles) and qe-fig-006 (9 axis labels); lqramsey’s Math is as much qe-math-010 (proposed) (19) as transpose.
qe-math-002 across the other ten Math-floored lectures — (14 / 52, 416 occurrences) series-wide, 381 of them inside Math-floored lectures. lqcontrol holds 85, lagrangian_lqdp 69. Same scripted-with-review work as item 1, and it lifts the Math mean, but these lectures stay HIGH on Writing regardless, so it buys score rather than bucket. lqcontrol and lq_inventories also carry qe-math-003 (17 pmatrix environments each) — a clean regex, worth doing in the same edit.
Fix the synced lectures upstream, not here. cross_product_trick is byte-identical to the lecture-python.myst copy and carries the series’ only build-risk finding — qe-math-006, five align environments inside $$ — plus the malformed {eq}`eq:Kalman102} at line 133. Both belong in lecture-python.myst; an edit here is overwritten at the next sync. Three more want a re-sync rather than an edit, because the upstream copy is already ahead: mccall_q (six qe-writing-006 headings and four repeated spaces that upstream no longer has), os_egm and os_egm_jax (each carries a qe-code-003 the upstream copy does not — the missing install cell). By contrast ifp_advanced has diverged, so its raw \label{a:y0} at line 158 (qe-math-007) needs its own edit here even though the same defect sits at the same line upstream.
The Writing reading pass — the only route that clears the 30. qe-writing-003 (45 of 52 lectures), qe-writing-002 (39), qe-writing-007 (36), qe-writing-005 (31) are judgment rules; no script helps, and every Writing-floored lecture carries at least one of them. Work down from the lowest overalls, where the mechanical half is thick enough to fold into the same edit: smoothing (6.0), perm_income_cons (6.5), rs_inventory_q (6.5), cons_news (6.6), lqcontrol (6.6). Pair it with qe-writing-008 (40 / 52, 1,578 occurrences) and qe-writing-006 (22 / 52, 146) in the same pass — both are near-scripted, but on their own they clear no lecture out of HIGH in this series.
Figures sweeps — broad, cheap, and low-leverage on the HIGH list. qe-fig-005 (42 / 52, 164), qe-fig-008 (38 / 52, 219) and qe-fig-001 (31 / 52, 102) are scriptable; qe-fig-003 (30 / 52, 105) is a reading pass, because each embedded title has to be rewritten as a caption. Only two lectures are Figures-floored — smoothing_tax (20 qe-fig-008 sites, the most in the series) and tax_smoothing_2 — so this raises the series mean rather than shortening the HIGH list.
Leave qe-ref-001 (20 / 52, 45) and qe-link-002 (15 / 52, 53) last. References at 9.3 and Links at 9.5 are already near the top of their range, and neither rule floors a lecture anywhere in the series. tax_smoothing_1 holds 9 of the 45 citation findings if one lecture is wanted for a first pass.

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	cross_product_trick	4	3	—	—	—	10	—	5.7	HIGH
2	smoothing	3	3	7.5	5	10	7.5	—	6.0	HIGH
3	tax_smoothing_1	4.5	4	7	6	7.5	9	—	6.3	HIGH
4	markov_jump_lq	5	3	7.5	5.5	8.5	9	—	6.4	HIGH
5	perm_income_cons	3	4	7.5	5.5	10	9	—	6.5	HIGH
6	rs_inventory_q	3	6	8.5	5	—	10	—	6.5	HIGH
7	tax_smoothing_2	5.5	3.5	7.5	4	8.5	10	—	6.5	HIGH
8	cons_news	3	4.5	8.5	6	10	7.5	—	6.6	HIGH
9	lqcontrol	3	3	7.5	4.5	10	8	10	6.6	HIGH
10	lagrangian_lqdp	3	3	7	—	10	7.5	10	6.8	HIGH
11	ifp_advanced	3	3	6.5	7	8.5	10	10	6.9	HIGH
12	amss2	3.5	5.5	8.5	6	8.5	10	—	7.0	HIGH
13	perm_income	3	4	7.5	6	8.5	10	10	7.0	HIGH
14	discrete_dp	4	7	6.5	6	9	7	10	7.1	HIGH
15	dyn_stack	4	5	8.5	5	10	7.5	10	7.1	HIGH
16	smoothing_tax	3.5	6	7.5	4	10	9	10	7.1	HIGH
17	amss3	3.5	5.5	8.5	5.5	7.5	10	10	7.2	HIGH
18	calvo	3	5.5	8.5	7	8.5	8	10	7.2	HIGH
19	inventory_q	4	5.5	7.5	6	10	10	—	7.2	HIGH
20	mccall_model	3	7	6.5	6	10	8	10	7.2	HIGH
21	amss	3.5	3.5	8	6	10	10	10	7.3	HIGH
22	opt_tax_recur	4	5	8.5	4.5	9	10	10	7.3	HIGH
23	calvo_machine_learn	4	3	6.5	8	10	10	10	7.4	HIGH
24	lq_inventories	4	3	7.5	7	10	10	10	7.4	HIGH
25	odu	3	9	7.5	5	9	8	10	7.4	HIGH
26	mccall_q	3	9.5	7	7	9	10	—	7.6	HIGH
27	chang_ramsey	3	10	8.5	6	8.5	10	—	7.7	HIGH
28	os_stochastic	3	7.5	8	7.5	10	8	10	7.7	HIGH
29	ifp_egm_transient_shocks	3.5	9.5	7.5	5.5	9	10	10	7.9	HIGH
30	lqramsey	6.5	3	7.5	8	10	10	10	7.9	HIGH
31	os_numerical	4.5	10	7.5	5.5	—	10	10	7.9	LOW
32	calvo_abreu	4	9	7.5	9	8.5	10	—	8.0	HIGH
33	ifp_egm	3	9	7.5	6.5	10	10	10	8.0	HIGH
34	ifp_opi	5	10	8	6	—	9	10	8.0	LOW
35	jv	3.5	9.5	7.5	6.5	9	10	10	8.0	HIGH
36	mccall_model_with_sep_markov	5	10	7.5	5.5	—	10	10	8.0	LOW
37	tax_smoothing_3	6	8.5	10	5	8.5	10	—	8.0	LOW
38	un_insure	4.5	9.5	7.5	5.5	9	10	10	8.0	LOW
39	mccall_fitted_vfi	7	8	7.5	5.5	9	10	10	8.1	LOW
40	os_egm_jax	7	10	5.5	7	—	10	10	8.2	LOW
41	os_time_iter	5.5	7.5	8	8	9	10	10	8.3	LOW
42	chang_credible	3	10	8.5	9.5	10	10	—	8.5	HIGH
43	ifp_discrete	6	7	8	8.5	10	10	10	8.5	LOW
44	mccall_model_with_separation	6.5	8.5	6.5	8.5	10	10	10	8.6	NONE
45	mccall_persist_trans	9.5	6.5	6	8	10	10	10	8.6	NONE
46	os_egm	5.5	9.5	7.5	9	10	10	10	8.8	NONE
47	career	8	9.5	8.5	6.5	10	10	10	8.9	NONE
48	os	6	9	9	8	10	10	10	8.9	NONE
49	short_path	7.5	10	8.5	8	—	10	10	9.0	NONE
50	intro	10	—	—	—	—	10	—	10.0	NONE
51	status	10	—	—	—	—	10	—	10.0	NONE
52	zreferences	10	—	—	—	—	10	—	10.0	NONE