What the system needs from you open items · silence has a defined consequence
open items28
Each item states what the system already tried before asking, and what happens if you stay silent. Silence is safe: every item proceeds on its stated conservative default and the assumption is logged to ASSUMPTIONS.md.
[H-003] whole book SYSTEMATIC_PATTERN · high
what happenedThe adversary returned 'weakened' on 10 of 10 names.
what the system already didDiscrimination was tested mechanically: the critiques are textually distinct and name-specific, so this is not reflexive hedging. The uniformity traces to evidence categories demanded across the whole book.
the questionIs the recurring objection a real doctrinal limit on what we can underwrite, or a gap in the evidence we assemble?
if you say nothingTreat it as an evidence gap and log a build task; do NOT cap conviction without an explicit instruction.
evidenceviews/adversary_reviews.jsonl, DECISIONS.md:Q-001
Or from the CLI: human_queue.py answer H-003 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-008] ALPHA MANDATE_SHORTFALL · high
what happenedALPHA holds 3 of 15 target names, 69.01% in cash.
what the system already didEvery rejection was on the return bar, not on a diversification or correlation gate — nothing was excluded at Stage 4. The pilot universe is 10 names in one supply chain, so 15 slots cannot be filled from it arithmetically.
the questionIs this the gates working correctly, or is coverage too thin to run the mandate? And do you want the cash held, or the universe widened first?
if you say nothingHold the cash and widen coverage. §10: the system reports the shortfall and does not reach to fill the box.
evidenceportfolio_proposal_alpha.json, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-008 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-012] AMD DIMENSION_DEADLOCK · high
what happenedThe analyst ranks AMD ABOVE ARM (conviction 3 vs 2) while the factor model ranks it BELOW (-0.04 vs +0.15) inside the same peer group 'foundry-logic', n=9.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size AMD versus ARM? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap AMD at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:AMD, dossiers/AMD.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-012 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-013] GOOGL DIMENSION_DEADLOCK · high
what happenedThe analyst ranks GOOGL ABOVE META (conviction 4 vs 3) while the factor model ranks it BELOW (+0.31 vs +0.64) inside the same peer group 'hyperscaler', n=7.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size GOOGL versus META? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap GOOGL at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:GOOGL, dossiers/GOOGL.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-013 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-014] INTC DIMENSION_DEADLOCK · high
what happenedThe analyst ranks INTC ABOVE ARM (conviction 3 vs 2) while the factor model ranks it BELOW (-1.04 vs +0.15) inside the same peer group 'foundry-logic', n=9.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size INTC versus ARM? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap INTC at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:INTC, dossiers/INTC.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-014 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-015] QCOM DIMENSION_DEADLOCK · high
what happenedThe analyst ranks QCOM ABOVE ARM (conviction 3 vs 2) while the factor model ranks it BELOW (-0.10 vs +0.15) inside the same peer group 'foundry-logic', n=9.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size QCOM versus ARM? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap QCOM at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:QCOM, dossiers/QCOM.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-015 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-016] MU DIMENSION_DEADLOCK · high
what happenedThe analyst ranks MU ABOVE 000660.KS (conviction 3 vs 2) while the factor model ranks it BELOW (-0.23 vs +0.20) inside the same peer group 'memory', n=4.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size MU versus 000660.KS? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap MU at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:MU, dossiers/MU.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-016 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-017] 8035.T DIMENSION_DEADLOCK · high
what happenedThe analyst ranks 8035.T ABOVE 6146.T (conviction 3 vs 2) while the factor model ranks it BELOW (-0.09 vs +0.28) inside the same peer group 'semicap-equipment', n=11.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size 8035.T versus 6146.T? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap 8035.T at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:8035.T, dossiers/8035.T.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-017 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-018] AMT DIMENSION_DEADLOCK · high
what happenedThe analyst ranks AMT ABOVE SPG (conviction 3 vs 2) while the factor model ranks it BELOW (-0.10 vs +0.29) inside the same peer group 'reit', n=7.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size AMT versus SPG? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap AMT at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:AMT, dossiers/AMT.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-018 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-019] AVGO DIMENSION_DEADLOCK · high
what happenedThe analyst ranks AVGO ABOVE MRVL (conviction 3 vs 2) while the factor model ranks it BELOW (-0.40 vs -0.20) inside the same peer group 'foundry-logic', n=9.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size AVGO versus MRVL? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap AVGO at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:AVGO, dossiers/AVGO.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-019 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-020] DUK DIMENSION_DEADLOCK · high
what happenedThe analyst ranks DUK ABOVE 0002.HK (conviction 3 vs 2) while the factor model ranks it BELOW (-0.10 vs +0.36) inside the same peer group 'utility', n=10.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size DUK versus 0002.HK? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap DUK at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:DUK, dossiers/DUK.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-020 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-021] EPD DIMENSION_DEADLOCK · high
what happenedThe analyst ranks EPD ABOVE 0857.HK (conviction 3 vs 2) while the factor model ranks it BELOW (-0.03 vs +0.13) inside the same peer group 'energy-dividend', n=10.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size EPD versus 0857.HK? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap EPD at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:EPD, dossiers/EPD.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-021 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-022] IBM DIMENSION_DEADLOCK · high
what happenedThe analyst ranks IBM ABOVE CSCO (conviction 3 vs 2) while the factor model ranks it BELOW (+0.00 vs +0.04) inside the same peer group 'tech-dividend', n=3.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size IBM versus CSCO? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap IBM at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:IBM, dossiers/IBM.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-022 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-023] NEE DIMENSION_DEADLOCK · high
what happenedThe analyst ranks NEE ABOVE 0002.HK (conviction 3 vs 2) while the factor model ranks it BELOW (-0.10 vs +0.36) inside the same peer group 'utility', n=10.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size NEE versus 0002.HK? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap NEE at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:NEE, dossiers/NEE.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-023 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-024] O DIMENSION_DEADLOCK · high
what happenedThe analyst ranks O ABOVE SPG (conviction 3 vs 2) while the factor model ranks it BELOW (+0.00 vs +0.29) inside the same peer group 'reit', n=7.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size O versus SPG? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap O at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:O, dossiers/O.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-024 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-025] WEC DIMENSION_DEADLOCK · high
what happenedThe analyst ranks WEC ABOVE 0002.HK (conviction 3 vs 2) while the factor model ranks it BELOW (-0.05 vs +0.36) inside the same peer group 'utility', n=10.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size WEC versus 0002.HK? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap WEC at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:WEC, dossiers/WEC.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-025 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-026] WPC DIMENSION_DEADLOCK · high
what happenedThe analyst ranks WPC ABOVE SPG (conviction 3 vs 2) while the factor model ranks it BELOW (-0.20 vs +0.29) inside the same peer group 'reit', n=7.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size WPC versus SPG? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap WPC at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:WPC, dossiers/WPC.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-026 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-027] XOM DIMENSION_DEADLOCK · high
what happenedThe analyst ranks XOM ABOVE 0857.HK (conviction 3 vs 2) while the factor model ranks it BELOW (-0.32 vs +0.13) inside the same peer group 'energy-dividend', n=10.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size XOM versus 0857.HK? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap XOM at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:XOM, dossiers/XOM.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-027 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-028] VST DIMENSION_DEADLOCK · high
what happenedThe analyst ranks VST ABOVE GEV (conviction 3 vs 2) while the factor model ranks it BELOW (-1.09 vs -0.29) inside the same peer group 'power-thermal', n=12.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size VST versus GEV? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap VST at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:VST, dossiers/VST.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-028 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-029] VST (ALPHA) VETO_UNRESOLVED · high
what happenedpm-review recommended vetoing VST in ALPHA, the name is still held at 6.40% of capital, and no decision was ever recorded in policies/vetoes.yaml.
what the system already didReconciled every RECOMMENDED VETOES paragraph in the pm-review memos against policies/vetoes.yaml and the live books. AMD, PFE and KMB were recorded; this one was not. The system cannot decide it — a veto removes capital, so P10 gates it on you.
the questionHonour the veto on VST (it leaves at Stage 1 with reason ben_veto), or decline it on the record?
if you say nothingThe position stays. Nothing is removed automatically — but the recommendation remains OPEN and the nightly keeps reporting it, because an unactioned veto is not the same as a declined one.
evidencepm_review_alpha.md, policies/vetoes.yaml, portfolio_proposal_alpha.json, veto_closure.json
Or from the CLI: human_queue.py answer H-029 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-030] 0857.HK (INCOME) VETO_UNRESOLVED · high
what happenedpm-review recommended vetoing 0857.HK in INCOME, the name is still held at 4.40% of capital, and no decision was ever recorded in policies/vetoes.yaml.
what the system already didReconciled every RECOMMENDED VETOES paragraph in the pm-review memos against policies/vetoes.yaml and the live books. AMD, PFE and KMB were recorded; this one was not. The system cannot decide it — a veto removes capital, so P10 gates it on you.
the questionHonour the veto on 0857.HK (it leaves at Stage 1 with reason ben_veto), or decline it on the record?
if you say nothingThe position stays. Nothing is removed automatically — but the recommendation remains OPEN and the nightly keeps reporting it, because an unactioned veto is not the same as a declined one.
evidencepm_review_income.md, policies/vetoes.yaml, portfolio_proposal_income.json, veto_closure.json
Or from the CLI: human_queue.py answer H-030 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-031] GIS (INCOME) VETO_UNRESOLVED · high
what happenedpm-review recommended vetoing GIS in INCOME, the name is still held at 4.40% of capital, and no decision was ever recorded in policies/vetoes.yaml.
what the system already didReconciled every RECOMMENDED VETOES paragraph in the pm-review memos against policies/vetoes.yaml and the live books. AMD, PFE and KMB were recorded; this one was not. The system cannot decide it — a veto removes capital, so P10 gates it on you.
the questionHonour the veto on GIS (it leaves at Stage 1 with reason ben_veto), or decline it on the record?
if you say nothingThe position stays. Nothing is removed automatically — but the recommendation remains OPEN and the nightly keeps reporting it, because an unactioned veto is not the same as a declined one.
evidencepm_review_income.md, policies/vetoes.yaml, portfolio_proposal_income.json, veto_closure.json
Or from the CLI: human_queue.py answer H-031 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-032] CVX (INCOME) VETO_UNRESOLVED · high
what happenedpm-review recommended vetoing CVX in INCOME, the name is still held at 4.40% of capital, and no decision was ever recorded in policies/vetoes.yaml.
what the system already didReconciled every RECOMMENDED VETOES paragraph in the pm-review memos against policies/vetoes.yaml and the live books. AMD, PFE and KMB were recorded; this one was not. The system cannot decide it — a veto removes capital, so P10 gates it on you.
the questionHonour the veto on CVX (it leaves at Stage 1 with reason ben_veto), or decline it on the record?
if you say nothingThe position stays. Nothing is removed automatically — but the recommendation remains OPEN and the nightly keeps reporting it, because an unactioned veto is not the same as a declined one.
evidencepm_review_income.md, policies/vetoes.yaml, portfolio_proposal_income.json, veto_closure.json
Or from the CLI: human_queue.py answer H-032 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-033] 0941.HK (INCOME) VETO_UNRESOLVED · high
what happenedpm-review recommended vetoing 0941.HK in INCOME, the name is still held at 4.40% of capital, and no decision was ever recorded in policies/vetoes.yaml.
what the system already didReconciled every RECOMMENDED VETOES paragraph in the pm-review memos against policies/vetoes.yaml and the live books. AMD, PFE and KMB were recorded; this one was not. The system cannot decide it — a veto removes capital, so P10 gates it on you.
the questionHonour the veto on 0941.HK (it leaves at Stage 1 with reason ben_veto), or decline it on the record?
if you say nothingThe position stays. Nothing is removed automatically — but the recommendation remains OPEN and the nightly keeps reporting it, because an unactioned veto is not the same as a declined one.
evidencepm_review_income.md, policies/vetoes.yaml, portfolio_proposal_income.json, veto_closure.json
Or from the CLI: human_queue.py answer H-033 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-062] CEG DIMENSION_DEADLOCK · high
what happenedThe analyst ranks CEG ABOVE GEV (conviction 3 vs 2) while the factor model ranks it BELOW (-0.45 vs -0.34) inside the same peer group 'power-thermal', n=12.
what the system already didBoth readings are internally sound: the analyst's rests on the revision tape and company disclosure, the quant's on cross-sectional ratios against real comparables. Neither is a data error — they disagree about which evidence should dominate.
the questionWhich dimension should size CEG versus GEV? Or should the pair be treated as one position and sized on the weaker of the two?
if you say nothingKeep the analyst's ranking but cap CEG at the weight the quant ranking would justify — i.e. size on the WEAKER read.
evidencedesk_quant.json:CEG, dossiers/CEG.yaml, pm_review_alpha.md
Or from the CLI: human_queue.py answer H-062 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-005] correlation inputs UNSTABLE_INPUT · normal
what happened4 correlation pair(s) driving position haircuts are unstable across sub-windows: 005930.KS|NVDA halves=[0.604, -0.019]; NVDA|005930.KS halves=[0.604, -0.019]; 005930.KS|000660.KS halves=[0.485, 0.835].
what the system already didTested three ways (probe_correlation_validity.py): FX is not the driver, and the optimiser already uses weekly returns. The instability is real, so the code now applies the MORE CONSERVATIVE of the two half-window estimates.
the questionIs a conservative (higher) correlation the right default for sizing, or would you rather size on the full-window number and accept the estimation risk?
if you say nothingConservative: assume the higher correlation. Underestimating correlation between two holdings is the error that costs money.
evidenceportfolio_proposal_alpha.json:correlation_stability
Or from the CLI: human_queue.py answer H-005 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-007] FMP consensus EBITDA PROVENANCE_CONFLICT · normal
what happenedFMP consensus ebitdaAvg sits BELOW reported operating income on 11 of 56 universe names (20%) — arithmetically impossible for a forward EBITDA.
what the system already didTested whether this was staleness rather than a field defect: consensus EPS is current (0-3% off reported trend) while EBITDA is 17-35% low on the SAME rows, so it is field-specific. Also found FMP consensus carries no as-of timestamp at all, so it cannot be used point-in-time (A-008).
the questionDo we drop FMP consensus EBITDA entirely and derive it from reported statements, or keep it with a warning for the 80% where it reconciles?
if you say nothingDerive EBITDA from reported statements; never read the consensus field. Use estimates.sqlite (dated per note) for anything point-in-time.
evidenceBACKLOG.md, ASSUMPTIONS.md:A-008
Or from the CLI: human_queue.py answer H-007 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.
[H-009] options / positioning coverage (re-raised) EVIDENCE_BLIND · normal
what happenedRE-RAISED: identical to H-004, which was closed by a build-agent TEST write on 2026-08-17, not by Ben. No decision was made. 61% of ALPHA gross exposure has NO usable option data, so there is no market cross-check on those theses.
what the system already didFutu has no single-stock chain for Korean listings; the liquidity gate correctly refuses to publish a number rather than computing one from a thin chain. Positioning has no short-interest or institutional-flow source at all.
the questionAccept the blind spot and hold these names anyway, or require a market cross-check before a name can exceed some weight?
if you say nothingHold them, but flag in the pm-review that a majority of gross carries no market cross-check.
evidencedesk_options.json, desk_positioning.json
Or from the CLI: human_queue.py answer H-009 "…"
The box needs serve.py running. On the static file:// build it has nowhere to post — use the CLI there.