HScreen

rendered 2026-08-22 10:45 UTC

System freshness every step, when it last ran and how fresh its data is

scanned steps37
failing1
scanned at2026-08-22T10:44:56Z UTC
A step is fresh if it ran within its cadence (nightly 30h, weekly 8d) and its artifact is not older than that. reported steps (dossiers, adversary, approved books) are event- or decision-driven and are meant to age between runs — their age is shown, never failed. A MISSING artifact fails the nightly the same way a crash would.
stepstatusdata ageexpectedlast run (UTC)
state filesfresh0.1hnightly2026-08-22T10:04:04Z
transcript backfillfresh0.6hnightly2026-08-22T10:06:46Z
depot embedfresh13.6hnightly2026-08-22T10:06:47Z
desk-technicalfresh0.5hnightly2026-08-22T10:12:35Z
desk-quantfresh0.4hnightly2026-08-22T10:20:10Z
desk-optionsfresh0.4hnightly2026-08-22T10:20:12Z
desk-positioningfresh0.4hnightly2026-08-22T10:20:45Z
earnings calendarfresh0.3hnightly2026-08-22T10:24:03Z
dividend metricsfresh0.3hnightly2026-08-22T10:29:17Z
desk-macrofresh0.1hnightly2026-08-22T10:36:02Z
credibilityfresh0.2hnightly2026-08-22T10:31:54Z
visibilityfresh0.2hnightly2026-08-22T10:32:12Z
tail dependencefresh0.2hnightly2026-08-22T10:32:58Z
dependence matrix (canonical)fresh0.2hnightly2026-08-22T10:35:13Z
ER marked to live pricefresh0.1hnightly2026-08-22T10:35:59Z
chain centralityfresh0.1hnightly2026-08-22T10:35:59Z
kelly derivedfresh0.1hnightly2026-08-22T10:35:59Z
freeze boardfresh0.1hnightly2026-08-22T10:35:59Z
payout durabilityfresh0.0hnightly2026-08-22T10:43:43Z
dividend qualityfresh0.0hnightly2026-08-22T10:43:43Z
stage 1+2 eligibility — alphafresh0.0hnightly2026-08-22T10:43:46Z
stage 1+2 eligibility — incomefresh0.0hnightly2026-08-22T10:43:48Z
stage 3+4 shadow optimise — alphafresh0.0hnightly2026-08-22T10:44:18Z
stage 3+4 shadow optimise — incomeFAILED16.0hnightly2026-08-22T10:44:56Z
book driftfresh0.0hnightly2026-08-22T10:44:56Z
tier two watchlist (D-027)fresh0.0hnightly2026-08-22T10:44:56Z
calibration statistical unit (A4)fresh0.0hnightly2026-08-22T10:44:56Z
veto closure (A5)fresh0.0hnightly2026-08-22T10:44:56Z
render terminalfresh2.7hnightly2026-08-21T18:46:43Z
archive snapshot (append-only research record)fresh1.0hnightly2026-08-21T18:46:43Z
rebuild search index over the archivefresh1.0hnightly2026-08-21T18:46:44Z
public mirror → DigitalOceanfresh2.7hnightly2026-08-22T08:00:14Z
calibration resolvefresh73.4hweekly2026-08-19T09:23:24Z
dossiers (newest)reported73.4hevent2026-08-19T09:23:23Z
adversary reviewsreported73.4hevent2026-08-19T09:23:23Z
approved proposal alphareported2.7hmanual2026-08-22T08:05:12Z
approved proposal incomereported42.3hmanual2026-08-20T16:28:31Z

Health

Theme registrygreen148 tickers; no "unclassified" bucket exists
Renormalisation rateREDup to 1 component(s) absent on eligible names
ER basisgreenevery dossier carries scenario probabilities
Ledger labellingRED29 call row(s) missing situation_type
Tail-dependence inputsgreenevery held pair has >=500 observations
DAG assertiongreenchain graph acyclic
Regime daemonREDvol_regime.json absent — SHORT-HEDGE addendum, parallel session
Side-session mergegreeninterface acknowledgment published for the short/hedge session

The freeze board v1.4 — the design is frozen; changes need evidence, not enthusiasm

From v1.4 the design is frozen. Changes are justified by ledger evidence or measured defects, never by new design enthusiasm. New scoring components, gates, weights or agents require (a) a calibration finding naming the deficiency, or (b) a reproduced defect with an artifact. STANDING EXCEPTION: anything fixing a fail-open path, a provenance gap, or a point-in-time violation is always in scope — integrity outranks the freeze.

deferred — triggers read from live state

Mutual information between signal dimensions (v1.4 4c)
>=50 resolved ledger rows		0/50
HMM regime model, market-level (v1.4 section 6)
>=2 quarters of vol-harvest outcomes, then must beat the hand-built thresholds out of sample		0.0/2
Convex solver replacing greedy Stage 4 (v1.4 section 6)
eligible universe > 300 names		51/300
KL-vs-implied as a Stage-2 component at 0.10 (v1.4 4a)
options-implied coverage >=60% of eligible names		0.0435/0.6
Experience-rated credibility Z per theme/situation (v1.3 2.3)
>=30 resolved rows in a grouping cell		0/30

standing rejections — read the reason before relitigating

Markov / binomial machinery on stock outcomes
non-stationary p, correlated trials, reflexive state — the assumptions fail exactly where it matters. The probabilistic layer lives in scenario probabilities, options-implied distributions, and the calibration loop		REJECTED, standing
Stochastic-calculus option pricing
the system READS market prices; modelling them adds error, not edge		REJECTED
Fitted ML for alpha on fundamentals
effective n ~ 120 autocorrelated cross-sections. Fitting remains licensed on the revision tape ONLY		REJECTED, standing
Full copula fitting for tail dependence
data-hungry; the empirical lambda_L coefficient answers the question asked (v1.4 Part 2, built)		REJECTED
Uncapped position sizing (Kelly at full conviction)
Kelly requires knowing p. Measured: our central estimates moved ~50% in one day on evidence improvement alone (NVDA 28->14, 2327.TW 22->6). Half-Kelly absorbs that estimation error; uncapped does not. DECISIONS Q8		REFUSED by the build agent, Ben may overrule explicitly

post-freeze changes

none — and an empty log is the freeze working. A change cannot be entered here without a linked finding: a calibration result or a reproduced defect with an artifact.

Self-test

=== 1. scripts parse ===
  PASS  all scripts compile  — 95 files

=== 2. dossiers validate ===
  PASS  36 dossiers validate

=== 3. cited chunk_ids resolve ===
  PASS  every cited chunk_id appears in that name's evidence pack

=== 4. depot alignment ===
  PASS  depot vectors match their text  — ALIGNMENT OK — every sampled row's vector matches its own text (self_sim >= 0.85)

=== 5. adversary coverage ===
  PASS  all 36 dossiers have a CURRENT L1 review

=== 6. append-only stores ===
  PASS  views/adversary_reviews.jsonl is 66 valid JSONL rows
  PASS  views/monitor_state.jsonl is 109 valid JSONL rows
  PASS  calibration.jsonl is 145 valid JSONL rows
  PASS  views/heartbeat.jsonl is 3 valid JSONL rows

=== 6b. ledger labelling (P0.3 fields on live rows) ===
  FAIL  no NEW unlabelled call rows (P0.3 fields on live rows)  — 26 unlabelled of 77 call rows; baseline is 16 pre-fix rows that can never be labelled. Any excess means a writer is not calling ledger_fields.enrich(). Tickers: ['000660.KS', '005930.KS', '2327.TW', '6146.T', 'ABBV', 'AMD', 'ASML', 'CVX']
  FAIL  every call row carries theme_tag  — 24 of 77 rows missing theme_tag (baseline 16 pre-fix)

=== 7. state files carry priced facts ===
  PASS  143 state files, price+source+date present

=== 8. desk outputs ===
  PASS  desk_technical.json has stages  — 10 entries
  PASS  desk_quant.json has profiles  — 76 entries
  PASS  desk_options.json has chains  — 7 entries
  PASS  desk_positioning.json has positioning  — 5 entries

=== 9. portfolio pipeline reproduces ===
  PASS  Stage 1/2/4 rerun produces the same holdings  — 0 holdings, cash 100.0%

=== 10. provenance on every figure ===
  PASS  every figure block carries a source_id

============================================================
16/18 checks passed

FAILURES:
  no NEW unlabelled call rows (P0.3 fields on live rows): 26 unlabelled of 77 call rows; baseline is 16 pre-fix rows that can never be labelled. Any excess means a writer is not calling ledger_fields.enrich(). Tickers: ['000660.KS', '005930.KS', '2327.TW', '6146.T', 'ABBV', 'AMD', 'ASML', 'CVX']
  every call row carries theme_tag: 24 of 77 rows missing theme_tag (baseline 16 pre-fix)
Rendered from canonical artifacts only (T1) — no terminal database exists. Every figure is read from the system's own files at render time; where an artifact is missing the panel names the dependency rather than showing an empty box. No execution path exists anywhere in this interface (T5), not even a link.