2026-09-12 · midday

Saturday afternoon, and this is the second read of the same shut session. The book came back not merely similar but byte-identical to the 07:12 pull - the same 32 legs, the same 32 marks to the last decimal, net liquidation the same to the cent, and all thirteen quotes still carrying Friday's 19:37-to-19:59 closing prints. Every rail lands exactly where it landed this morning and every runbook row is clear. So this run spent itself on the one thing this morning's page filed as unexplained: the 06:07 GPU capture that ran and collected nothing. It has been checked at source and it was transient. The two captures since, at 09:07 and 12:07, both came back full with healthy sample counts, and the health check patched this morning correctly names the empty row rather than passing it in silence.

What to do

280P x5 and 300P x8 sit above the $260 long-leg ceiling.
BUY TO CLOSE 5 GOOGL Dec15'28 360 PUT / SELL TO CLOSE 5 GOOGL Dec15'28 280 PUT
one spread, limit, day · ~$18,716.70 debit ($37.4334 net per spread on Friday's closing marks)
Your playbook: “long leg never above $260 (must sit at or above the $230-250 floor zone, never below $230)”
Fourteenth read of the same standing order, and the first in the series to reprice to exactly the previous read - the 360P and 280P marks are unchanged to eight decimals because the market has not opened between the two pulls. These are Friday's closing marks, not a live quote; the order has to be repriced on the day it is placed, which is Monday at the earliest. Nothing has been traded.

Still over, unchanged

These fire every run until you act on them. No order attached.

rail	now	limit
long_leg_band	280P x5, 300P x8	$230-$260 band	13 contracts outside
net_cash_ratio	-0.0654	amber -0.05, red -0.10	amber
gross_leverage	1.724x	amber 1.3x, red 1.5x	red
top2_name_delta_ratio	0.66x	amber 0.70x, red 0.85x	clear
total_delta_ratio	0.94x	amber 1.30x, red 1.50x	clear

Why each one reads the way it does

long_leg_band
The rail the action item above addresses. No leg has been traded.
net_cash_ratio
-0.148 when the policy was signed. Cash is unchanged to the cent at -$93,230.12 for an eleventh consecutive session, and net liquidation has not moved since this morning either, so the ratio is identical rather than merely close.
gross_leverage
1.57x when the policy was signed. This rail divides gross position value by net liquidation, which counts both legs of every spread; the delta rails measure the same book at 0.66x and 0.94x and both are clear.

delta by name

Delta notional per name as of 2026-09-12: stock at delta one, each option leg at the Black-Scholes delta implied by its own mark. Top two: GOOG, MU.

name	delta notional	of NLV
GOOG	$797,710	55.9%	1543 sh, 16 legs
MU	$143,118	10.0%	525 sh, 2 legs
META	$93,252	6.5%	144 sh, 0 legs
VOO	$86,463	6.1%	123 sh, 0 legs
RDDT	$80,777	5.7%	513 sh, 0 legs
NVDA	$49,671	3.5%	285 sh, 2 legs
INTC	$30,915	2.2%	300 sh, 0 legs
AMZN	$25,670	1.8%	100 sh, 0 legs
BRK B	$20,422	1.4%	40 sh, 0 legs
EQT	$10,814	0.8%	200 sh, 0 legs
TSLA	$365	0.0%	1 sh, 0 legs
DIS	$106	0.0%	1 sh, 0 legs

Everything else clear

Nearest trigger is tranche 2 weakness (nearest written level, and what rails.py reads) at $317, 5% away.
No position moved more than 5% today. Biggest was INTC +2.72%.

All 28 triggers swept, with today's number

price trigger	now	fires at	distance
GOOG · tranche 2 weakness (nearest written level, and what rails.py reads)	$335.38	≤ $317	5.5% down
GOOGL · tranche 2 weakness	$338.39	≤ $317	6.3% down
GOOGL · tranche 3 confirmation	$338.39	≥ $360	6.4% up
GOOG · tranche 3 confirmation	$335.38	≥ $360	7.3% up
GOOG · thesis break, close regardless of price	$335.38	≤ $280	16.5% down
GOOG · washout, stop selling spreads and buy stock	$335.38	≤ $260	22.5% down
GOOGL · crash powder, deploy the $100k reserve	$338.39	≤ $260	23.2% down
MU · collar roll-up	$975.07	≥ $1,250	28.2% up
GOOGL · close every short put	$338.39	≥ $434	28.2% up
GOOGL · long spreads near max, apply the buy-it-fresh test	$338.39	≥ $500	47.8% up
GOOGL · shares capped by the 600Cs, roll the collar up	$338.39	≥ $600	77.3% up

level trigger	now	fires at	distance
Dec-28 30-delta put IV	36.19%	> 43%	+6.8 pt
GOOGL IV, crash trigger	36.19%	> 50%	+13.8 pt
Dec-28 300/400 call spread, mark vs width	41.48%	>= 90%	+48.5 pt
8x Sep-27 300/440 call spread, mark vs width	38.07%	>= 90%	+51.9 pt
10x Dec-28 350/500 call spread, mark vs width	28.94%	>= 90%	+61.1 pt
GOOGL Dec-28 340P extrinsic	$56.32	< $5	$51.32
Excess liquidity / NLV	74.78%	< 40%	+34.8 pt
NLV change on the week	-0.17%	-8% in one week	+7.8 pt
Defined max loss vs program cap	$175,059	$200,000	$24,941 headroom
Biggest single-name move on Friday	INTC +2.72%	>= 5%	+2.3 pt of headroom

calendar	date	away
Wash-sale window opens for the 6/05 GOOG loss lots	2026-09-17	5d
Micron fiscal Q4 print (not a trigger row)	2026-09-30	18d
Alphabet Q3 print, scheduled thesis adjudicator - re-read the house view within 72h	2026-10-28	46d
No crash by this date: deploy the $100k reserve at market anyway	2026-10-29	47d
Earliest date the GPU glut tripwire can evaluate a 90-day baseline	2026-11-16	65d
Alphabet capex guide, second scheduled adjudicator	2027-02-03	144d
GOOG lots vest long-term (+66 sh): add collar put leg 9, log in the straddle register same day	2027-03-13	182d

Spot prices

	last	day
GOOG	$335.38	+1.5%
GOOGL	$338.39	+1.7%
MU	$975.07	-0.2%
AMZN	$256.70	+1.9%
BRK B	$510.54	+0.7%
DIS	$106.50	+0.6%
EQT	$54.07	-1.6%
INTC	$103.05	+2.7%
META	$647.58	+0.5%
NVDA	$218.26	-0.0%
RDDT	$157.46	+1.4%
TSLA	$365.25	+0.5%
VOO	$702.95	+0.9%

The check itself

resolved This morning's GPU capture failure was transient, and that is now established rather than assumed.
The 06:07 intraday capture raised a CalledProcessError on all eleven sources at once and wrote a row with every price column blank. This morning's page said the manual pull twenty minutes later succeeded but the cause was unexplained, and left it open. Two scheduled captures have run since: 09:07 and 12:07, both full rows, both with healthy sample counts (H100 SXM n=19 and n=12, RTX 4090 n=64 on both). The LaunchAgent is loaded and its last exit status is 0. One empty capture bracketed by five good ones on either side is a transient fault in the environment the job runs in, not a broken schedule, and the daily series never lost a row because the manual pull had already written it.
verified The health-check patch written this morning fires correctly on real data.
Re-running the script prints 'WARNING: 1 of 8 intraday captures in the last 24h arrived but collected NO prices (2026-09-12T06:07:04-07:00). The 24h averages above rest on 7 captures, not 8.' It names the offending row, its timestamp, and the effect on the averages, and points at the notes column for the cause. This is the check being watched doing the thing it was patched to do, against the failure that motivated the patch - not a re-reading of this morning's claim that it would.
found this run The book came back byte-identical, not merely unchanged.
32 rows, the same 32 contract_ids, zero quantity diffs and zero mark diffs - every one of the 32 marks matches this morning's pull to the full eight decimals the feed serves. The account summary compares equal field for field. All 13 equity quotes returned FROZEN with the same last-trade timestamps, 19:37:46 to 19:59:59 ET Friday. On a shut market this is the correct behaviour and it is worth stating plainly: nothing on this page could have moved, so anything that differs from this morning would have been a fault in the check, not a fact about the market.
found this run Every connector call succeeded on the first attempt.
Positions, summary and 14 price snapshots, 16 of 16, zero retries. Fourth consecutive clean run.
will happen again The campaign IV field returns the shut-market artifact again.
option-midpoint-iv on the Dec-28 340 put comes back as annualIv -15.8745 with isValid false - a negative volatility, the same defect recorded on 08-23, 08-29, 08-30, 09-03, 09-05, 09-06 and this morning, every one of them a closed or pre-market book. The separate implied-vol field is valid at 0.36190194301204803 and is what the rails read. It is identical to this morning's value to all sixteen digits, which is the expected behaviour for a frozen book and matches what Friday's midday run observed.
note The GPU script is idempotent and did not write a second daily row.
It reported 'row for 2026-09-12 already exists - reporting from it' and printed the levels table off the row the open run wrote. The daily CSV still carries exactly one 2026-09-12 row. The daily pull is an open-run job; this run only read it.

Rails that could not be evaluated

The 06:07 failure is established as transient but its cause is still not identified. Eleven sources raising CalledProcessError in the same second points at the environment the LaunchAgent runs in - a network blip, a DNS failure, a throttled process table - rather than at any one provider, but nothing in the captured notes distinguishes between those. If it recurs, the notes column will not be enough to diagnose it; the job would need to log the subprocess stderr, which it currently discards.
stressed_excess_liquidity is still unevaluated. It needs IBKR Risk Navigator at -30% price and +50% implied vol with the TWS gateway up, and has never once been measured. This is the rail that would actually price the crash scenarios the runbook's Part 3 is built around.
The glut tripwire still cannot fire. It is defined as H100 spot 20% or more below its level 90 days back, and the daily series starts 2026-08-18, so no 90-day reference exists. The earliest date it can evaluate is around 2026-11-16.
The SemiAnalysis composite the tripwire is written against is still the 2026-04 print at $2.82, five months stale, with later months paywalled. The daily marketplace numbers are being compared to an anchor that has not moved since April.
The 8x Dec-28 300P leg is outside the long-leg band and has no written repair order. Every page in this series proposes the 280P repair and says nothing actionable about the 300Ps, because nothing actionable is written. 8 of the 13 out-of-band contracts have no remedy on paper.
short_dec28_340p_extrinsic is listed unevaluated in rails.json but was computable this run and was swept by hand against the runbook: $56.32 against a $5 trigger. The file's stated reason - that it needs a live Dec-2028 chain pull - is wrong; the position row's own mark is enough. The file was not edited, because this check does not edit the rails to match the book.

Where the numbers came from

Run label decided from the rails log, not the clock: an entry for 2026-09-12 already carried markets-open, so this is the midday run. The GPU daily pull and the whole forecast pass were skipped as open-run jobs.
Saturday, the US market is shut. The book is Friday 2026-09-11's finished session for the second time today, and all 13 equities returned top_status FROZEN.
Book pulled 19:39Z / 12:39 PDT. Diffed field by field against the 07:12 pull: 32 rows, same 32 contract_ids, zero quantity diffs, zero mark diffs, summary equal field for field.
All 13 equity quotes are single get_price_snapshot calls, every one FROZEN with a last-trade stamp between 19:37:46 and 19:59:59 ET Friday - the same stamps the open run recorded.
Prior closes: all 13 served by the endpoint, matching the open run's values. Three-for-three on frozen books continues to hold.
Campaign IV read, not solved: option-midpoint-iv returned the shut-market artifact (annualIv -15.8745, isValid false) so the valid implied-vol field, 0.36190194301204803, is what the rails read.
Runbook Parts 2, 3 and 4 were swept row by row against this run's own position marks rather than copied from the open run. The three call-spread width percentages (41.48%, 38.07%, 28.94%), the 340P extrinsic ($56.32) and the excess-liquidity ratio (74.78%) were each recomputed and each reproduces the open run's figure, which is the expected result on an identical book.
Repair debit is the 360P mark minus the 280P mark on 5 spreads: (68.8750000 - 31.4416008) x 500 = $18,716.70, unchanged from this morning because neither mark moved.
Week-over-week net liquidation is measured against 09-05, the previous Saturday, so both sides are a frozen book carrying a Friday close: $1,426,052.75 against $1,428,489.17, -0.17%.
The GPU intraday health check was re-run to verify this morning's patch and to read the two captures that have landed since. No daily row was written; the script reported from the existing 2026-09-12 row.