SAPIEN EDGE · SHADOW

Live today's alert tape

SESSION2026-08-04 ALERTS0 SHOWNUNKNOWN

SHADOW MODE · NOTHING WAS SENT. No order left this machine. Every ticket on this page is a proposal on a screen — the click is a human's.

His size is the author's own fill. Our size is what this account would take under its own ceilings — and on a refusal it is zero, not the number the ticket would have used. The gap between the two columns is the entire point of this screen.

TIME	CONTRACT	ACTION	HIS SIZE	OUR SIZE	OURS : HIS	VERDICT

No alert in the corpus for this session.

Position what is actually deployed right now

FLAT

SEM_POSICAO · the bot's state document is intact and says flat

The bot's own state document is intact and reports flat. Nothing is deployed.

Protection — what the engine would do now

no open position to protect

Risk the ceilings in force and what is left of today

ENGINEOK

DAILY LOSS BUDGET$0.00 / $444

TRADES TODAY0 / 3

PREMIUM ENVELOPE$0.00 / $2,000

Ceilings in force

Envelope	$2,000	hard ceiling on premium deployed at once
Equity reference	$22,200	declarado_humano:live SnapTrade read taken in the Classic
Exposure per trade	2.5%
Risk per stop	2.5%
Exposure total	5.0%
Daily loss	$444
Max trades / day	3
Max concurrent	1
Hard contract cap	1
Max round-trip spread	5.0%
Default stop	25.0%

Policy conflicts (resolved to the tighter side)

VERBATIM FROM ANOTHER MODULE · the lines below were written by the engine that produced them (sapien/protecao.py, sapien/risco.py) and are printed exactly as received. Those modules speak Portuguese; this screen does not translate them, because a paraphrased engine reason is a reason nobody can grep for in the engine.

EXPOSURE_PCT · config=5.0% · Thiago disse 2,5% · 5,0% é o PHASE_A_RISK_PCT do Solomon → usando o menor: 2.5%
DAILY_LOSS · config=2.0% ($444) · módulo=$555 → usando o menor: $444
DAILY_TARGET · config tem daily_target_pct=4%; arquitetura_v1 §5.4 diz 'não existe DAILY_PROFIT_TARGET neste sistema' → regra da casa permite meta APENAS como gatilho de give-back condicional, NUNCA como teto que para o dia. Não implementado aqui.

Brakes

Kill-file directory	/root/ophi-sapien-data/kill
Writable	YES	a brake that cannot be pulled is not a brake
Kill files present	none present

PORTOES_NAO_AVALIADOS · risk gates were NOT evaluated: they require a live two-sided quote and a fresh bar, neither of which exists on this read-only port. A gate list built from zeros passes every ceiling.

History the alert corpus, graded

STATESEM_AMOSTRA SOURCEEnhancedMarket alert corpus, grouped by POSITION

GROUPED BY POSITION, NOT BY ALERT. One position is one entry and every exit that unwinds it; each position counts once, weighted by quantity. Counting per message weights every partial trim as if it were a whole trade — winners exit in many messages and losers in one, so the per message average is inflated by roughly an order of magnitude and can flip the sign of the result.

POSITIONS

WIN RATE

UNKNOWN%

MEAN RETURN

UNKNOWN

PROFIT FACTOR

UNKNOWN

return scale, size-independent

MEDIAN

UNKNOWN

BEST

UNKNOWN

WORST

UNKNOWN

PROFIT FACTOR ($)

UNKNOWN

the author's dollars at the author's size

Cumulative return

SERIES NOT PUBLISHED · the snapshot carries aggregates only (n, win rate, mean, median, profit factor) and no per-position series. This chart draws the moment edge_dados publishes historico.retornos_pct (per-position returns, in chronological order) or historico.curva (cumulative points). Nothing is drawn from the mean and the extremes: a curve inferred from three aggregates is an invented curve.

Waiting on one upstream key: historico.retornos_pct or historico.curva.

Sample accounting

Chains seen	0
Graded	0
Closed in corpus	0
Entry never seen	0
Excluded — no exit	0
Excluded — ungradeable	0
Every chain accounted for	YES	if this is NO, a position vanished and the sample is lying about its own n

HISTORICO_SEM_AMOSTRA · no position could be graded, so there is no statistic to show — an empty sample is not a zero result

Cross-check against the source's own published P&L

weighted mean of the slices vs the 'PnL total' the source itself published at the end of the chain — in-corpus ground truth for the grouping

Comparable chains	0
Agree	0
Disagree	0	tolerance 2.0 pp

HISTORICO_SEM_AMOSTRA · no position could be graded, so there is no statistic to show — an empty sample is not a zero result

Journal the day, in order

SESSION2026-08-04 AS OF02:20:33 ET

DERIVED TIMELINE. The snapshot publishes no journal block, so this is assembled from the alert tape, the open position and the watcher heartbeat. It is not sapien/journal.py's append-only day file, and it does not claim to be.

TIME	KIND	WHAT

Watcher state	NUNCA_ESCREVEU
Watcher files	0
Heartbeat age	UNKNOWN

VIGIA_NUNCA_ESCREVEU · the watcher output directory does not exist: it has never written. This is different from 'its heartbeat is late'.