2026-08-27 · open

The same four rails as every run since 08-20, unchanged and structural. The standing long-leg close reprices to $37.93 net, another new high in the series. The live-quote endpoint was down for the whole run, so prices came from the broker's own daily bars and the Dec-28 put IV could not be read at all.

What to do

280P x5 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 · ~$18,965 debit ($37.93 net per spread on this run'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 as prior runs, repriced from the book's own marks: $37.93 net against $37.47 at yesterday's midday, $37.02 at yesterday's open, $36.40 on 08-25 midday and $36.76 on 08-21. Sixth consecutive new high in the series and the drift stays one-way, the short 360P richening faster than the long 280P on every run this week. Closing the pair also cuts defined max loss, so it moves the program_cap rail the same direction rather than against it. The 8x Dec-28 300P also trip this rail and remain the separate matter; no order is proposed against them.

Still over, unchanged

These fire every run until you act on them. No order attached.
rail now limit
program_cap $175,059 $100,000 cap 1.8x, $75,059 over
gross_leverage 1.729x 1.50 red / 1.30 amber red
net_cash / NLV -0.071 -0.05 amber / -0.10 red amber

rail	now	limit
program_cap	$175,059	$100,000 cap	1.8x, $75,059 over
gross_leverage	1.729x	1.50 red / 1.30 amber	red
net_cash / NLV	-0.071	-0.05 amber / -0.10 red	amber

Why each one reads the way it does

program_cap
Unchanged to the dollar for a seventh run. Defined max loss is a function of strikes and entry credits, not current marks, so it moves only when a position actually closes. Nothing closed.
gross_leverage
The IPS row this proxies measures total DELTA notional / NLV; this rail measures GROSS position value. rails.json carries its own note, '1.57 at signing on gross; true delta leverage ~1.0x'. A red here is not the same statement as the policy row being red. See plumbing.
net_cash / NLV
Cash is -$99,195 on NLV $1,404,920. Signed at -14.8% and already red that day, so this is the improved end of a known accepted state, not a new break. The IPS written amber action is: no new leverage, no new delta in the top two names.

Everything else clear

Nearest trigger is tranche 3 confirm at $360, 6% away.

All 15 triggers swept, with today's number

price trigger	now	fires at	distance
GOOGL · tranche 3 confirm	$338.91	≥ $360	6.2% up
GOOGL · tranche 2 weakness	$338.91	≤ $317	6.5% down
GOOGL · Rule 8 crash powder	$338.91	≤ $260	23.3% down
GOOGL · Rule 6 close every short put	$338.91	≥ $434	28.1% up
MU · roll the collar	$943.23	≥ $1,250	32.5% up

level trigger	now	fires at	distance
Dec-28 30-delta put IV	not read	> 43%	unevaluated
GOOGL Dec-28 340P extrinsic	$58.00	< $5	+$53.00
Dec-28 300/400 call spread vs width	41.5%	>= 90%	+48.5 pt
Dec-28 350/500 call spread vs width	29.1%	>= 90%	+60.9 pt
Excess liquidity / NLV	74.9%	< 40%	+34.9 pt
NLV change over one week	+0.12%	-8% in one week	+8.1 pt

calendar	date	away
Wash-sale window opens for the 6/05 GOOG loss lots	2026-09-17	21d
Q3 print, the first scheduled adjudicator	2026-10-28	62d
Deploy the $100k reserve at market if no crash by then	2026-10-29	63d
First GOOG vest collar leg (leg 9, +66 sh)	2027-03-13	198d

Spot prices

	last	day
GOOG	$336.00	-0.9%
GOOGL	$338.91	-0.9%
AMZN	$256.39	-1.5%
BRK B	$501.86	-0.6%
DIS	$107.99	-1.5%
EQT	$54.17	-1.1%
INTC	$91.06	+3.2%
META	$568.80	-1.3%
MU	$943.23	+0.5%
NVDA	$225.35	+7.5%
RDDT	$155.77	+0.3%
TSLA	$348.97	+0.9%
VOO	$706.64	+0.3%

Cost of compute

H100 spot $2.06 against the $2.82 monthly composite. No tripwire.
H100 SXM printed $2.06, down 22.8% on the day off a 26-listing book, but its own 24h average is $2.46 and its 7-day is $2.27. The now-print sits 16% under its own 24h average, which is a marketplace-floor spike rather than a level change, and the contract columns did not move with it: getdeploying cloud $4.19 (-1.2%) and Akash $2.71 (flat). B200 is the more real softening, $6.01 against a $7.10 30-day average, and L40S went the other way at +20.8%.
The H100 tripwire still cannot fire. It tests spot against 90 days back and the vast series does not yet reach 90 days, so the 30d and 90d cells read n/a. Nothing here is a confirmed glut signal.

Every rate, per chip and per source

Dollars per GPU-hour. Vast is a marketplace floor and runs well below contract rates, which is what the cloud and hyperscaler columns are there to show. Bold rows are current generation.

	vast	cloud	hyper	akash	n
H100 SXM	$2.06	$4.19	$7.89	$2.71	26
H100 NVL	$2.69	-	-	-	9 thin
H200	$4.08	$4.50	$10.44	$4.45	21
B200	$6.01	$7.88	$15.18	-	21
A100 SXM4	$0.83	$2.00	$3.67	$1.87	40
A100 PCIE	$0.93	-	-	-	29
L40S	$0.64	-	-	-	22
RTX 4090	$0.34	-	-	-	64
RTX 5090	$0.40	-	-	-	64

H100 by source. SFCompute clearing price $1.88. Composite is $2.82 as of 2026-04 (PRE-SIGNAL tightening; latest print as of 2026-08-18, so the anchor is four months stale).

source	now	24h avg	day
vast median	$2.06	$2.46	-22.8%
sfcompute	-	$1.88	-
getdeploying	$4.19	-	-1.2%
akash	$2.71	$2.71	+0.0%

The check itself

source broke get_price_snapshot is down for the whole run.
Ten calls between 06:37 and 06:46 PT all returned 'the connector's server returned an error', for equities and for the GOOGL Dec-28 340P alike. search_contracts failed the same way. get_account_positions, get_account_summary and get_price_history all worked, so this is not a dead connector, it is one endpoint. Consequence: every price in this report is a daily-bar close rather than a live quote, and the Dec-28 30-delta put IV could not be read at all, which puts the t2_iv rail into unevaluated for the first time in the series.
source broke DuckDuckGo HTML no longer returns search results.
https://html.duckduckgo.com/html/ answers HTTP 202 with a challenge page and zero result rows. markets.md names it as one of two working WSB discovery routes. The other route, rendered Reddit HTML, still works through connected Chrome and was used instead.
never worked the gross_leverage rail has never measured the policy row it cites.
rails.json maps gross_leverage to gross_position_value / net_liquidation and takes its 1.30 amber / 1.50 red thresholds from the IPS. But the IPS row carrying those thresholds is 'Total delta notional / NLV', shown at ~1.2x at signing, and the IPS separately states effective leverage is '~1.0x on a beta basis (not the 1.57x IBKR reports)'. So the rail applies delta-notional thresholds to a gross-notional quantity. rails.json's own note admits the mismatch. It has fired red on every run of this series and that red has never been a statement about the policy metric. Either the rail needs its own thresholds or it needs per-leg deltas, which are the same input top2_name_delta_ratio is already blocked on.
will happen again the runbook's 11x Dec-28 300/400 call tranche is not eleven spreads.
The book holds 11x long GOOGL Dec-28 300C against exactly 1x short Dec-28 400C, so ten of the eleven short legs are absent. Part 2's '90% of width' row is reported as a single 300/400 percentage, 41.5% today, computed off one short leg and applied to all eleven contracts. The unpaired 300Cs are covered by the 10x short 500C, 18x short 600C and 1,543 GOOG shares, but they are not the defined-width structure the trigger names.
will happen again rails.py reads the GOOGL spot rails off the GOOG quote.
The script evaluated every GOOGL price rail against GOOG $336.00 rather than GOOGL $338.91. The gap is about 0.9% and changes no verdict today, but the watch_price distances in this report are computed against GOOGL's own quote and so differ from the script's.
will happen again the SemiAnalysis composite anchor is four months stale.
gpu-index.semianalysis.com has not published past the 2026-04 print of $2.82. Every daily GPU number is quoted against an anchor from April. 'Flat' at the composite is unpublished, not observed.

Rails that could not be evaluated

t2_iv - the Dec-28 30-delta put IV was not read because the only route is the snapshot endpoint, which was down all run. First time this rail has gone unevaluated.
stressed_excess_liquidity - needs IBKR Risk Navigator (-30% price, +50% IV) with the TWS gateway up. Never yet measured.
top2_name_delta_ratio - needs per-leg option deltas, which summary.json does not carry. ~0.90 at signing and already red.
short_dec28_340p_extrinsic - rails.json still classes this unmeasured pending a live Dec-2028 chain pull, but the position mark answers it: $58.00 of extrinsic against the <$5 watch, with the put $1.09 in the money at GOOGL $338.91.
The vast GPU series still does not reach 90 days, so the H100 tripwire's own test cannot run.
SemiAnalysis has published no composite since 2026-04.

Where the numbers came from

get_account_positions and get_account_summary: succeeded (one retry on positions).
get_price_snapshot: FAILED on every one of ten attempts between 06:37 and 06:46 PT, for equities and for the option contract alike. search_contracts failed too.
Prices therefore come from get_price_history ONE_DAY bars: last = today's bar close-so-far, prior_close = the 2026-08-26 bar close. Both are broker numbers. Nothing was estimated.
Price set was captured 06:40-06:42 PT and is internally consistent as one snapshot. A later 06:46 pull showed GOOGL had moved to 341.26, so the levels here run a few minutes behind the tape.
GOOGL contract id 208813719 confirmed by its 2026-08-25 close of 346.96 matching the prior close recorded on 08-26.
Dec-28 30-delta put IV: not read this run, the only route is the snapshot endpoint.