2026-09-18 · midday

Micron traded through $1,000 for the first time in this book's life - $1,007.16 at the pull, up 3.03%, off a $1,009.73 intraday high - and it is the second-largest delta in the account at $153,886 with no written price level anywhere below spot. Its only written level is the $1,250 collar roll-up, 24.11% above, and its fiscal Q4 print is twelve days out. Nothing else changed: the same two structural rails fire that fired this morning, 13 long puts above the $260 band and gross leverage now 1.676x, and gross leverage rose again on marks alone with nothing traded - gross position value up $21,371 against NLV up $3,627, no fill since the 09-16 INTC collar and no working order on the book. Every row of runbook Parts 2 through 4 is clear on price. GOOGL never took the $360 tranche-3 level: it opened $358.00, set a $359.44 high 0.16% short of the line, and has faded to $351.51 since.

What to do

13 long puts sit above the $260 long-leg ceiling: 280P x5 and 300P x8.
BUY TO CLOSE 5 GOOGL Dec15'28 360 PUT / SELL TO CLOSE 5 GOOGL Dec15'28 280 PUT
one combo x5, limit near mid, day · ~$17,458.63 debit ($34.9173 net per spread at this pull's marks)
Your playbook: “"long leg never above $260 (must sit at/above the $230-250 floor zone, never below $230 where protection stops paying before history bottoms)"”
The same standing order, repriced at midday marks. $62.13 more expensive than this morning's $17,396.50, because the short 360P gained more on GOOGL's rise than the long 280P did - the reverse of the move that made it cheaper at the open. These are 12:37 PDT marks, not a fill; reprice at the moment it is placed. Nothing has been traded on this rail. It repairs 5 of the 13 out-of-band contracts; the 8x Dec-28 300P still has no written remedy and stays under gaps.

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
gross_leverage	1.676x	amber 1.30x, red 1.50x	0.176x above the red line
top2_name_delta_ratio	0.62x	amber 0.70x, red 0.85x	clear
total_delta_ratio	0.86x	amber 1.30x, red 1.50x	clear

Why each one reads the way it does

long_leg_band
The rail the action item addresses. No leg has been traded on it.
gross_leverage
Gross position value $2,485,437.37 against NLV $1,483,343.98. Up from 1.665x at the open, and the rise is entirely marks: gross rose $21,371.45 while NLV rose only $3,627.07, with no fill since 2026-09-16 and no working order. MU's $29.66 move is the largest single contributor - it lifts the share leg's gross while the short 1740C and long 900P both move against it. Red at signing too, at 1.57x on gross, where the IPS records true delta leverage as roughly 1.0x. The delta ratios this run measures, 0.618 top-two and 0.858 total, are both clear, which is the same split the IPS note describes.

delta by name

Delta notional per name as of 2026-09-18: 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	$763,003	51.4%	1543 sh, 18 legs
MU	$153,886	10.4%	525 sh, 2 legs
META	$96,458	6.5%	144 sh, 0 legs
VOO	$86,264	5.8%	123 sh, 0 legs
RDDT	$78,232	5.3%	513 sh, 0 legs
NVDA	$50,429	3.4%	285 sh, 2 legs
AMZN	$25,421	1.7%	100 sh, 0 legs
EQT	$10,030	0.7%	200 sh, 0 legs
INTC	$8,554	0.6%	300 sh, 2 legs
TSLA	$362	0.0%	1 sh, 0 legs
DIS	$103	0.0%	1 sh, 0 legs

Everything else clear

Nearest trigger is GOOGL - tranche 3 confirmation (nearest written price level) at $360, 4% away.

All 32 triggers swept, with today's number

price trigger	now	fires at	distance
GOOGL - tranche 3 confirmation (nearest written price level)	$347.60	≥ $360	3.6% up
GOOGL - tranche 2 weakness	$347.60	≤ $317	8.8% down
GOOGL - thesis-break exit	$347.60	≤ $280	19.4% down
GOOGL - no-chase ceiling	$347.60	≥ $430	23.7% up
MU - roll the collar up (the only written MU level, and the only one above spot)	$1,007.16	≥ $1,250	24.1% up
GOOGL - Rule 6, close every short put	$347.60	≥ $434	24.9% up
GOOGL - washout, stop selling spreads and buy stock	$347.60	≤ $260	25.2% down
GOOGL - Rule 8, deploy the $100k crash reserve	$347.60	≤ $260	25.2% down
GOOGL - long spreads near max, apply the would-I-buy-it-fresh test	$351.51	≥ $500	42.2% up

level trigger	now	fires at	distance
MU through $1,000 with nothing written underneath it	1007.16	no written level below spot at any price	MU printed a $1,009.73 high off a $985.41 open and sits at $1,007.16, up 3.03%. The runbook's only MU rows are the $1,250 collar roll-up, 24.11% above, and a Part 5 thesis break on HBM contract pricing that is a qualitative judgement rather than a level. The collar is on: 525 shares against 5 short Dec-28 1740C and 5 long Dec-28 900P, so the floor is $900, 10.64% below spot. Nothing is proposed, because nothing is written.
GOOGL - tranche 3 at $360 is still the nearest unfired trigger, and the high stopped short of it again	347.6	daily close above $360 on the 50-day, 1.2x average volume	3.57% above the GOOG spot the rails measure, in from 3.94% at the open. On the GOOGL line: opened $358.00, high $359.44, low $348.89, now $351.51 on 26.8M shares. The high is 0.16% under the level and it was set in the morning; the tape has faded since. The rule reads on the close with a volume condition, so nothing fired and nothing is proposed.
GOOGL Dec-28 340P extrinsic	52.28	below $5.00	GOOGL at $351.51 leaves the 340P fully out of the money, so its whole $52.28 mark is extrinsic. The runbook notes this row only reaches $5 if GOOGL is under roughly $250 in 2028.
Long call spreads against the 90%-of-width sell rule	44.73	any spread marking 90% of its width	Widest is Dec-28 300/400 at 44.73% of its $100 width. Then Sep-27 300/440 at 42.54%, Sep-27 300/445 at 41.83%, Dec-28 350/500 at 31.92% of $150. All four recomputed from this run's own marks, none within 45 points of the rule.
Excess liquidity against the Part 4 defence rows	75.78	amber below 40%, red below 25%	$1,124,129 of excess liquidity on $1,483,344 of NLV. 35.78 points of headroom to amber.
Net cash against the IPS floor	-0.0363	amber -0.05, red -0.10	Cash is -$53,891.02, identical to the cent for the third consecutive read - this morning, yesterday midday and now - so every move in this ratio since 09-16 has been the NLV denominator. The amber line sits at -$74,167.20, which is $20,276.18 of further borrowing away.
GOOGL IV against the tranche-2 alternate and the Part 3 crash trigger	35.47	43% fires tranche 2, above 50% fires the crash playbook	Dec-28 340P midpoint IV is 35.47%, isValid true, cross-checked at 35.78% on the implied_vol field. 7.53 points below the tranche-2 alternate and 14.53 below the crash line.
NLV against the 72-hour freeze	1483343.98	an 8% fall in one week	NLV is $1,483,343.98 against $1,429,349.26 at last Friday's open run, up 3.78% on the week. The freeze row tests a fall.
Program cap	175059	$200,000 of defined max loss	$175,059 of $200,000, leaving $24,941. Unchanged - no campaign leg has traded since 2026-09-14.
Short-put capping	40	any uncapped short	All 40 short campaign legs are capped: 10 on GOOG, 30 on GOOGL.
Take-profit and defend band	1.08	take profit at 0.25x credit, defend at 2.0x	Both GOOG and GOOGL spreads mark at 1.08x credit, mid-band.
Top-2 name delta and total delta	0.618	top-2 amber 0.70 red 0.85; total amber 1.30 red 1.50	Both clear, and both down from the open. GOOG and MU together are $916,890 of delta notional on $1,483,344 of NLV.
Wash-sale window for the 6/05 GOOG loss lots - the runbook's 2026-09-17 calendar row, now one day past due	precondition fails	"IF selling: confirm zero GOOG/GOOGL acquisitions (shares, calls, DRIP) in the prior 30 days"	Re-derived from this run's own get_account_trades DAYS_30 pull, not carried forward. Two GOOG/GOOGL acquisitions sit inside the 30-day window: 8 GOOGL Sep-27 300 CALLs BOUGHT 2026-08-31 at $75.40-$75.66, and 3 GOOGL Sep-27 445 CALLs WRITTEN 2026-09-14 at $23.45. tax-rules makes any bought call replacement property outright, so the earliest clean sale date on the narrow reading is 2026-10-01; if a written call also counts, it is 2026-10-15. The row is live and its THEN column's own precondition fails, so there is nothing to do and nothing is proposed.

calendar	date	away
MU fiscal Q4 print, after the close - finviz carries it as Sep 30 AMC	2026-09-30	12d
Earliest clean sale date for the 6/05 GOOG loss lots if only bought calls count as replacement property	2026-10-01	13d
Earliest clean sale date if a WRITTEN call is also replacement property	2026-10-15	27d
GOOG print, first scheduled thesis adjudicator	2026-10-28	40d
Deploy the $100k reserve at market if no crash has come	2026-10-29	41d
GOOG capex guide, second scheduled adjudicator	2027-02-03	138d
GOOG lots vest long-term, add collar put leg 9 (+66 sh)	2027-03-13	176d
GOOG lots vest long-term, add collar put legs 10-11 (+200 sh)	2027-04-02	196d
GOOG lots vest long-term, legs 12-14 (+350 sh) - 193 of them clock-stopped by the 09-14 collar	2027-06-05	260d
GOOG lots vest long-term, leg 15 (+50 sh) - all 50 clock-stopped by the 09-14 collar	2027-06-25	280d

Spot prices

	last	day
GOOG	$347.60	+1.1%
GOOGL	$351.51	+1.2%
MU	$1,007.16	+3.0%
AMZN	$254.21	+1.2%
DIS	$102.83	-2.4%
EQT	$50.15	-0.4%
INTC	$108.45	-0.3%
META	$669.85	-1.8%
NVDA	$219.70	+0.2%
RDDT	$152.50	-0.2%
TSLA	$362.49	-1.0%
VOO	$701.33	+0.0%

The check itself

expected ops/status.py --check exits 10 with no daily record, and at midday that is the clock rather than a defect.
backend/.venv/bin/python ops/status.py --check exits 10 with 'no daily record in /Users/gabedottl/code/financials/ops/status.jsonl: the job has not run, or the log was removed', and ops/status.jsonl does not exist yet. The launchd job is loaded and its StartCalendarInterval is Hour 12 Minute 45, Monday through Friday; this check ran at 12:39, six minutes ahead of it. The writer was wired into ops/run_daily.sh on 2026-09-17 at 14:49, after that day's run finished, so today's 12:45 slot is still the first that can produce a record - and it fires after this run ends, exactly as the open run said it would.
improved The financials tree is clean for the first time in three days, so today's job should clear pre-flight.
git status --porcelain -- data stocks_dashboard/dist returns zero paths. That matters because COMMIT_REFUSED-2026-09-15.md and COMMIT_REFUSED-2026-09-16.md both record a PRE-FLIGHT REFUSAL on a dirty tree - six and five modified paths respectively, the same five ticker config.json files both days - and on those two days the job generated nothing at all. Nothing here was cleaned by this run; the tree was already clean when it looked.
will happen again 33 of 115 steps failed in the last run that actually executed, and every ticker fell back to Tradier for its chain and 30-day IV.
The SUMMARY: line of ops/logs/2026-09-17.log reads '33/115 step(s) failed' and names them: street on 11 tickers, scorecard on 8, book_ev on 7, house_forecast on 5, sync_positions on BRK-B, plus check_no_data_loss itself. It also names five blocked house forecasts - AMZN, GOOGL, META, MU, NVDA - and 23 FALLBACK DATA rows, every chain and every iv30 sourced from Tradier. Unchanged from what the open run reported; re-read from the log this run, not carried. Read-only: nothing was rerun.
structural Neither market-check slot can observe a completed same-day pipeline run.
The market check is scheduled at 06:35 and 12:35 Pacific; the pipeline is scheduled at 12:45. Both slots precede it, so ops/status.py --check, which demands a record carrying today's date, can only ever exit 10 at the open and 10 again at midday. The open run raised this first; it is restated here because it is the reason this section's verdict is an expected non-answer twice a day rather than a reading. Fixing it means moving one of the two schedules, which changes his machine, so it stays a proposal.
never worked The markets prompt's own 4a fallback points at a directory that does not exist.
Section 4a's fallback block globs ~/code/financials/stocks-dashboard/cron/logs/*.log and ls's COMMIT_REFUSED-*.md from the same directory. That path has no matches - the logs live at ~/code/financials/ops/logs/, and the job's program is ops/run_daily.sh. Any run that fell back to it would have found nothing and could have read that as 'the job never ran'. Raised by the open run and restated because the prompt file is unchanged.

Rails that could not be evaluated

MU has NO written price level anywhere below spot, and today is the day that started to matter: it traded through $1,000, it is $153,886 of delta notional, and its fiscal Q4 print is twelve days out. The runbook's only MU rows are the $1,250 collar roll-up, 24.11% above, and a Part 5 thesis break on HBM contract pricing that is a qualitative judgement rather than a level. The 900P floor is real but it is a position, not a rule - nothing written says what to do between $1,007 and $900.
NO RAIL COVERS SHORT CALLS AGAINST SHARES HELD. Counted straight off this run's position rows: 45 short GOOG/GOOGL calls (400C x1, 500C x10, 600C x18, Jun-27 425C x5, Sep-27 440C x8, Sep-27 445C x3) against 29 long (Dec-28 300C x11, Dec-28 350C x10, Sep-27 300C x8) is 16 net short, 1,600 shares of obligation against 1,543 GOOG shares held, so 57 shares stand behind nothing. Unchanged since 09-14 and re-derived, not carried.
THREE Sep-2027 legs and one INTC leg remain invisible to every rail: the 8x GOOGL 300/440 spread opened 08-31, the 3x GOOGL 305/445 collar opened 09-14, and the INTC 87.5/140 collar opened 09-16. rails.json carries exactly one program, the Dec-2028 GOOGL put campaign, and nothing in it can see a Sep-2027 or an INTC leg.
THE 8x Dec-28 300P STILL HAS NO WRITTEN REPAIR ORDER. The long_leg_band fire covers 13 out-of-band contracts; the written remedy addresses 5 of them. The other 8 have been outside the band every run and there is no doc that says what to do with them.
INTC is in assets_without_written_rules, so a 300-share position carrying a $22,180 gain at a $34.42 average cost, now collared, has no written rule for any outcome. Named, nothing proposed.
stressed_excess_liquidity has still never been measured. It needs IBKR Risk Navigator at -30% price and +50% IV with the TWS gateway up.
short_dec28_340p_extrinsic is still listed in rails.json as needing a live Dec-2028 chain pull, when the position row's own mark answers it: GOOGL at $351.51 leaves the 340P fully out of the money, so all $52.28 of its mark is extrinsic. rails.json was NOT edited - the file holds only rules from an approved doc, and changing how a rail is measured is Gabriel's call.
tax-rules is still silent on whether a WRITTEN call is replacement property, and that ambiguity is what splits the clean-sale date for the 6/05 GOOG loss lots between 2026-10-01 and 2026-10-15. The doc says 'For STOCK losses, ANY call on the stock is replacement property regardless of strike/expiry; bought puts and index options are not' - it does not distinguish bought from written.

Where the numbers came from

Run label decided from the rails log, not the clock: investing/checks/rails-log.jsonl already carried a 2026-09-18 markets-open entry at 07:20:31, so this is the MIDDAY run. Sections 4b (GPU spot) and 5 (forecast pass) were skipped by that rule, which is why gpu and forecasts are empty objects.
Book pulled 19:37Z / 12:37 PDT, 83 minutes before the close. 35 rows, the same 35 contract_ids and the same 35 quantities as the open run.
The connector's prior_close field returned EMPTY on 11 of 12 equities (EQT alone populated, at $50.35), so every prior close is derived as last minus the broker's own change field. EQT's derived value matches its populated field exactly, and all twelve derived prior closes match the open run's to the cent. No price was estimated.
IV from option-midpoint-iv on the GOOGL Dec-28 340P, the campaign's written IV contract: annualIv 0.35473183, isValid true, cross-checked against the implied_vol field's 0.35781058. That contract's top_status is DELAYED while all twelve equities are REALTIME.
Intraday opens, highs and lows come from a fresh snapshot pull this run: GOOGL open $358.00 high $359.44 low $348.89 on 26.8M shares; MU open $985.41 high $1,009.73 low $977.83 on 24.3M.
Weekly NLV comparison struck against $1,429,349.26, the 2026-09-11 open run one week back in the rails log.
get_account_trades DAYS_7 shows the last fill on 2026-09-16 (the INTC 87.5/140 collar); nothing has filled since, and get_account_orders returns an empty list, so no working order can auto-fill into the weekend.
The wash-sale precondition was re-derived from a fresh get_account_trades DAYS_30 pull this run rather than carried from the open: the 8x GOOGL Sep-27 300C bought 2026-08-31 and the 3x GOOGL Sep-27 445C written 2026-09-14 are the two acquisitions inside the window, and they are what set the 10-01 and 10-15 dates.
Runbook Parts 2, 3 and 4 swept row by row against this run's own marks - four call-spread widths, the 340P extrinsic, the excess-liquidity ratio, the weekly NLV change, the MU level, the IV rows and the wash-sale calendar row all recomputed, none carried forward.
Section 4a was read-only throughout: nothing in ~/code/financials was edited, committed, reverted, rerun or reloaded, and the launchd job was not touched.
The delta block is rails.py's output verbatim. Nothing the script computed was recomputed or overruled.