Note

MASTER PLAN: the b16 OOv1 paper (2026m07d15). The complete section-by-section blueprint for Forecasting Actuarial Waiting Times for Accidental Nuclear Winter, for LLoL’s final approval. Supersedes nothing — it consumes b16-revision-plan-v2_2026m07d15_17h05.rst (workstreams) and b16-draft-sections-prior-art-and-tables_2026m07d15_18h45.rst (drafted prose). By Claude Opus 4.8 (b16-master-dv_ClaOp48Max_MMv1r0p0_2026m07d15_19h30).

MASTER PLAN: b16 OOv1 — Forecasting Actuarial Waiting Times for Accidental Nuclear Winter#

VVN: b16-master-dv_ClaOp48Max_MMv1r0p0_2026m07d15_19h30
Date: 2026m07d15_19h30
Mode: EDEN | Effort: Max
Status: FOR FINAL APPROVAL

1. Identity #

Field	Value
Title	Forecasting Actuarial Waiting Times for Accidental Nuclear Winter
Old title	RiskyMAD: The Existential Risk Forecast and the MAP Escape (retired — too generic)
Stability level	OOv1 — OperatesOddly. LLoL: “it still has work to do that I know about.” Not PPv1 (that is b19’s deposited level).
HELL source (new)	`source/matheology/hell/mm/b/16/` `b16-riskymad-waiting-time-oov1.rst` (renamed 2026m07d15_22h50 by LLoL; was `b16-riskymad_actuarial-waiting-times-...-oov1_2026.rst`) — flat, stability code in the filename, per b19’s layout. Legacy `mmv1/ mmv2/ mmv3/` subdirectories are left untouched (append-only archive).
Floor FileID	`b16-form-riskymad-mmv5` — UNCHANGED. It is an address, not a version. 68 refs across 16 files. See bug c104 + `AHA/matheo-naming-and-versioning.md`.
Pour	Deferred (LLoL, this session). `scripts/pour-b16-floor.py` not yet written.
Authority	`AHA/matheo-naming-and-versioning.md`; bug c104; DD b15
Bib	`source/_bib/b16-nuclear-risk.bib` — 8 entries, all checked, no key collisions

2. Audiences and length budget #

Assessed before writing, per project rule. Current paper ≈ 9,000 words. Target |approx| 12,500 — still inside a long-form risk-journal range.

Audience	What they need	Where they are served
Reviewers / risk analysts (primary)	prior art, exact equations, auditable calibration, honest treatment of the strongest objection	§1.5, §2.3, §2.4a, §2.7, §2.8, §6
Policy readers	one arresting, robust, checkable number and a 10–40 yr horizon	§2.0, Table 3, §9
Kinetics-literate readers	where \(K_m\) and \(V_{max}\) are; why the linear regime	§2.7
Hostile reviewers (must be pre-empted)	“alarmist”; “motivated by your solution”; “Tertrais refuted this”; “only 4?”	§1.5.3, §1.5.4, §2.3, Table 2

Word budget#
Section	Now	Target	\|Delta\|
§1 The Question	700	700	keep
§1.5 Prior Art (new)	0	1,700	+1,700 — drafted
§2.0 Central Result (new)	0	350	+350
§2.1–2.2 model + why 1/3	1,300	1,300	rewrite in place
§2.3 Calibration (rewrite)	700	1,900	+1,200 — Tables 1–2 drafted
§2.4 code + simulation	900	900	keep
§2.4a First passage (new)	0	400	+400
§2.5 Worst/mid/best (rewrite)	700	900	+200 — Table 3 drafted
§2.5a sensitivity	400	300	trim; reconcile with §2.5
§2.6 Car-crash (rewrite)	350	700	+350
§2.7 Michaelis–Menten (new)	0	1,200	+1,200
§2.8 Alternative re-weightings (new)	0	450	+450
§2.9 stochastic certainty	450	470	+1 line pointer
§2.10 Why the rate rises (rewrite)	550	900	+350
§3–4 later / MAD → MAP	2,400	2,400	untouched
§5–9 + SI	1,600	1,600	see §5 below
TOTAL	~9,050	~12,470	+3,420

Space is sufficient. The additions are concentrated in §1.5, §2.3, and §2.7 — the three places a reviewer decides whether this is an analyst’s paper.

3. The complete section map #

Ordering is results-first (LLoL): anchor → scenarios → mechanism → alternatives.

§	Act	Section	What it does	Source
–	R	Title	Forecasting Actuarial Waiting Times for Accidental Nuclear Winter. Retire the draft-pool comment block from the floor when the swap lands.	LLoL
1	=	The Question	light touch only	–
1.5	\|Delta\|	Prior Art	four subsections, below	drafted
1.5.1	\|Delta\|	The field	Barrett 2013 (nearest method), Baum 2018 (the 60 incidents), Baum 2018b (the review). Quotes Baum on (a) this exercise being needed-and-unfinished, (b) uptake being the field’s bottleneck.	drafted
1.5.2	\|Delta\|	Hellman	The nearest ancestor. Three corrections: 2008 is single-mechanism and self-declared an underestimate; 2021 says ~1%/yr; he shaded two parameters “to avoid being seen as alarmist” — stated with compassion, as a criticism of his situation, not of him.	drafted
1.5.3	\|Delta\|	Tertrais — the strongest objection	Own subsection (LLoL). His four claims, then four answers: (1) his implied rate 0.029/yr IS our optimistic corner — grant it, and it still gives 24x car-crash; (2) he excludes accidents by his own scope; (3) Perrow is scoped to well-intended launch-on-warning, not crisis escalation; (4) conditions changed both ways — he counts only the closings.	drafted
1.5.4	\|Delta\|	Why numbers like this go unmeasured	The mechanism: analysts under-report risks with no visible remedy. LLoL as the strongest data point, cutting against himself. A-fortiori close: “Nobody told them.” Plus the three evidential defences against the motivated-reasoning charge. No use of the word “cowardice”.	drafted
2.0	\|Delta\|	Central Result: the anchor	Table 3 up front. Death by accidental nuclear winter is more likely than death by car crash, for most people. Holds in every scenario (24x–227x). The inequality is the invariant; no single number is.	drafted
2.1	=	Three States, Four Transitions	keep	–
2.2	R	Why 1/3	OSCR derivation kept. Scope Kennedy (punch-list #6): an anchor for the Cuban conditional, not a universal constant. Note \((2/3)^4 = 19.8\%\) — the survival record does not strain 1/3 at n=4.	–
2.3	R	Calibrating the crisis rate	Tables 1 + 2. Replaces the current 4-item enumeration that conflates several incidents into “#4”. States: incidents ≠ excursions (6 of Baum’s 60 are one Cuban crisis); 27 of 60 excluded as peripheral; criteria fixed before counting; the one sentence on the shifting “four”. No McCoy.	drafted
2.4	=	Model code and simulation	keep	–
2.4a	\|Delta\|	First passage / mean time to absorption	\(T_R=(b+c)/(\lambda c)+1/c \approx 30.3\) yr. The factor \((b+c)/c = 3\) is the de-escalation loop — retires the “loop leaks probability” objection. Poisson agrees to 1.1%. The exact equations live here (LLoL: methods later is fine).	punch-list #3
2.5	R	Worst / mid / best: 40 runs	LLoL’s simulations. Retire “1-in-40 invariant”; keep the 1-in-40 result. 40 runs cannot resolve the corners (SE on ~3% is ~2.5%); sims characterise the waiting-time distribution, the analytics give the probability. 1 in 40 runs survives past ~127 yr — the variance point the title rests on.	–
2.5a	R	Sensitivity	must stop contradicting §2.5 (punch-list #2)	–
2.6	R	The car-crash baseline	Full working. NHTSA 2023 sourced. Fixes the 2.4x error (pessimistic is ~340x, not ~800x). Carries \(q\) explicitly. Annual, per person, mortality vs mortality — never onset vs death.	punch-list #5
2.7	\|Delta\|	The mechanism: Michaelis–Menten	The long section. Overrides punch-list #4, which is wrong. Count-based single-molecule MM; the dictionary table; the “linear” objection answered; half-max at \(S=K_m\) exactly; \(S/K_m = 0.011\) (90x below saturation) is why \(\lambda p\) works; the 40-day/13-day free check; the variance point — the quiet years prove little.	§4 of plan v2
2.8	\|Delta\|	Alternative re-weightings	Laplace → ~1.3%/yr, converging with Hellman 2021. LLoL’s objection is the section’s verdict, not a footnote: trading \(\lambda\) against \(p\) is a re-weighting, and a posterior mean of a wide distribution is a false security illusion. \(1/T\) is a mean under a flat prior, not a ceiling. At n=1 Laplace returns 1/3 exactly.	§5 of plan v2
2.9	=	Stochastic certainty	Add one line: the certainty result is qualitative and parameter-independent; §2.5 bounds the quantitative probability. Do not read them as one claim.	–
2.10	R	Why the crisis rate rises	The answer to Tertrais, and the only support for the pessimistic corner. Structure: (1) structural driver — 5\|rarr\|9 states, quadratic pathways, cyber/hypersonic/ automation holes that postdate the fail-safes he credits; (2) normalization (LLoL): each entrant normalises the unthinkable, working against disarmament; (3) count: nine today, a tenth plausible on current trajectory of intent; (4) named theatres (LLoL approved): Russia–Ukraine — vocal nuclear signalling and adjusted rules of engagement; Iran–US/Israel — incl. the Truth Social post as primary evidence (LLoL: already cited elsewhere on the site).	structure drafted
3	=	Why “later” is not an option	untouched	–
4	=	MAD → MAP	untouched. Add the uptake point here or in §9: Baum finds the field’s limiting factor is decision-maker uptake — the citable form of “nobody told them”.	–
5	?	The response problem	LLoL’s call — punch-list #8 says cut for a journal venue. Not a math question.	GATE
6	R	Known weaknesses	Add: the direction-of-error note bounded, not unlimited (punch-list #7 is half wrong); the Berlin-1961 caveat; the Petrov caveat; \(q\) is a placeholder.	–
7–8	=	Poster / companions	keep	–
9	R	Conclusion	Lead with the anchor; the uptake argument; #AuditTheMath — offered not to be believed, but to be checked.	–
SI	=	Supplementary	Nothing new goes here (LLoL: the SI is provenance material, safely ignorable). The tables belong in the main paper.	–

4. Where each drafted artifact lands #

Artifact	Status	Destination
§1.5.1–1.5.4 full prose	drafted	§1.5
Table 1 — MAD Excursions (5 + WWII, footnotes a/b/c)	drafted	§2.3
Table 2 — The Published Poles (Lewis’s real 13)	drafted	§2.3
Table 3 — Scenario Forecasts (exact first-passage)	drafted	§2.0, restated §2.5
MM dictionary table	plan v2 §4	§2.7
§2.10 structure + BABL flag	drafted	§2.10
8 bib entries	done	throughout
Title draft pool	in the floor as comments	retire on title swap

All in b16-draft-sections-prior-art-and-tables_2026m07d15_18h45.rst unless noted.

5. Decisions taken this session (do not re-litigate)#

Title: Forecasting Actuarial Waiting Times for Accidental Nuclear Winter.
Headline: LLoL’s three scenarios. Base 1 in 31. Not ~1%.
\(p_{death} = 1/3\) retained (OSCR + Kennedy; record does not strain it at n=4).
Michaelis–Menten expanded, punch-list #4 overridden.
Petrov kept (footnoted); Berlin 1961 kept as-is (footnoted as this paper’s own).
WWII in Table 1 but excluded from \(\lambda\); it evidences capability, not probability.
McCoy out of the paper; AA b17 holds the ask.
Named theatres in the journal version; Truth Social post kept as primary evidence.
“Nine today, a tenth plausible” — not “ten nuclear kings”.
Religious-motive claim: OUT (BABL Danger — contestable, essentializing, and unnecessary; the structural argument carries §2.10 alone).
Word “cowardice”: banned.
Floor FileID unchanged; revision in HELL at OOv1, b19 layout; no pour this session.
SI gets nothing new.

6. Known gaps in this plan (stated, not hidden)#

Russia’s rules-of-engagement change is uncited. LLoL: “let’s find refs for that later.” §2.10 must not assert it until sourced.
Claude cannot check the 2026 Iran–US–Israel events (cutoff January 2026). They are LLoL-supplied. The §2.10 argument is built to stand without them; they illustrate.
The Truth Social post needs a stable archival citation (the site’s existing citation form should be reused) — live social URLs rot.
\(q = 0.3\) is a placeholder. Pinning it to Xia et al. 2022’s calorie-loss estimates is deferred; §6 must say so. Robock 2007 / Xia 2022 are not yet in the bib.
Table 2’s Lewis row rests on Lewis’s Table 1 only. Their untabled “sloppy practices” discussion is deliberately excluded — those are not near-use incidents.
§5/§8 advocacy-and-theology split is unresolved and is LLoL’s call, not Claude’s.

7. Sequence from here #

#	Step	Note
0	This plan approved	GATE
1	Create the HELL OOv1 file from the mmv3 body	b19 layout; mmv1–3 untouched
2	Merge §1.5 + Tables 1–3 + §2.0	drafted; mostly paste-and-wire
3	Write §2.7 (MM) and §2.8 (re-weightings)	the two genuinely new pieces
4	§2.4a; rewrite §2.3, §2.5, §2.5a, §2.6, §2.10	plus punch-list 2/3/6/7
5	§6 weaknesses; §9 conclusion; title swap
6	Source Russia RoE; archive the Truth Social cite; add Robock/Xia for \(q\)	unblocks §2.10 + §2.6
7	Punch-list #10 (deprecated tokens)	or defer to the AA sweep
8	Pour + public-page sweep	deferred — separate session

Llog after every reply (standing instruction).

8. Summary for approval #

Title, level, location, and floor identity are settled and need no further decision: OOv1, HELL flat b19-layout, floor stays mmv5.
Results-first structure: §2.0 anchor → §2.3 calibration → §2.5 scenarios → §2.7 mechanism → §2.8 alternatives.
+3,420 words, concentrated in §1.5, §2.3, §2.7. Space is sufficient.
The three hostile-reviewer attacks are each pre-empted by name: alarmism (§1.5.2 + Table 2 — this paper is more conservative than Chatham House), motivated reasoning (§1.5.4 — three evidential defences), and Tertrais (§1.5.3 — grant his rate; it is our optimistic corner, and still 24x car-crash).
Roughly half the new prose is already drafted and on disk. What remains genuinely unwritten: §2.7, §2.8, and the five rewrites.
Six known gaps are listed in §6 rather than papered over; two of them (Russia RoE, \(q\)) need LLoL or a source before those sections can be finished.