R
RelightLabs / Attentive.io
Public equities monitoring demonstration
Experimental brief · Not production signal flow
Experimental dry-run artifact

From unstructured evidence to cited workflow decisions.

This static brief demonstrates how an experimental family-office monitoring layer can convert filings, market evidence, and theme observations into structured decision cards with citations, human-review routing, and outcome-tracking hooks.

Demo path A
Synthetic 8-K → ESCALATE

Adverse-event evidence becomes a cited workflow decision.

Demo path B
CBRS → WATCH / INVESTIGATE

A newly relevant ticker enters a thematic watchlist without becoming a trade signal.

Safety posture
Dry run, isolated

No dashboard, alert, Ledger, calibration, ticker-universe, or LLM side effects.

Production isolation

The experimental monitoring layer is enabled locally for this demonstration, but every production-facing switch remains disabled. Outputs are informational decision support, not investment advice and not production Attentive signals.

Dashboard: false
Alerts: false
Ledger: false
Calibration: false
Ticker universe: false

Workflow decision cards

The key product pattern is not “AI predicts returns.” The pattern is: source evidence → structured extraction → cited decision card → human review → outcome tracking.

DemoCo Holdings adverse-event review

Synthetic 8-K-style evidence · Workflow decision
ESCALATE
Class: WORKFLOW_DECISION Direction: NONDIRECTIONAL Severity: 0.72 Confidence: 0.78
What happened
CFO resignation disclosed. Enterprise customer deployments delayed. Prior margin-improvement language withdrawn.
Why it matters
Leadership continuity, deployment timing, and margin language changed in ways that may increase thesis risk or monitoring priority.
Next human action
Review whether the event changes monitoring priority, thesis risk, or the next analyst follow-up. Consider adding to an investment committee brief if tied to a real holding.
Citations
  • synthetic_8k_demo_001 / paragraph_2The company disclosed that its Chief Financial Officer will resign effective immediately.
  • synthetic_8k_demo_001 / paragraph_3Management stated that previously expected enterprise customer deployments have been delayed into the next fiscal period.
  • synthetic_8k_demo_001 / paragraph_4The company withdrew its prior language around sequential margin improvement.

CBRS / AI Infrastructure Loop

Theme onboarding · Watchlist observation
WATCH / INVESTIGATE
Class: WORKFLOW_DECISION Direction: NONDIRECTIONAL Adverse-event severity: N/A Theme confidence: 0.74
What happened
CBRS is newly relevant to the AI Infrastructure Loop as a non-GPU AI accelerator proxy.
Why it matters
The watchlist addition helps monitor first earnings, analyst initiations, customer concentration, lockup dynamics, benchmark claims, and related-name read-throughs without creating a trade signal.
Related names
NVDA, AMD, MU, AVGO, TSM, INTC, VRT, GEV, CEG, VST, TLN, OKLO, FLNC
Next human action
Add to the experimental watchlist and investigate whether AI accelerator alternatives deserve recurring theme monitoring.
Citation
  • foam_theme_observation_cbrs_ai_infrastructure_loop / foam_theme_seedCBRS is newly relevant to AI Infrastructure Loop / AI accelerator alternatives as a newly public non-GPU AI accelerator proxy. It is a experimental watchlist observation, not a production signal.

Outcome tracking

The demo leaves outcomes pending by design. In a real pilot, these fields become the learning loop: analyst usefulness, decision changed, follow-on confirmation, market markout, and calibration cohort.

Decision card Human review Analyst label Follow-on confirmation Outcome status
foam_decision_card_foam_extracted_event_synthetic_8k_demo_001 pending pending pending pending
foam_decision_card_foam_theme_observation_cbrs_ai_infrastructure_loop pending pending pending pending

Dry-run markouts and adapter status

The experimental monitoring layer does not fabricate performance. This demonstration has no market-data fetch and no production Ledger write. Markouts remain pending until a controlled price/Ledger adapter is enabled.

Card Evaluation mode Status Would write to Ledger Calibration cohort
synthetic 8-K decision card dry_run pending_price_data false foam_experimental
CBRS theme observation card dry_run pending_price_data false foam_experimental
LLM adapterdisabled · deterministic demo logic
Ledger adapterdry-run · no production write
Prices adapterpending · no market fetch
Shadow rulesdisabled · no trading-policy eval
Technical feature flags (internal module codename: FOAM)
ENABLE_FOAMtrue
FOAM_DRY_RUNtrue
FOAM_WRITE_LEDGERfalse
FOAM_PUBLISH_DASHBOARDfalse
FOAM_ALLOW_USER_ALERTSfalse
FOAM_ALLOW_PRODUCTION_DISPATCHfalse
FOAM_ALLOW_PRODUCTION_CALIBRATIONfalse
FOAM_USE_LLMfalse

Safety statement

No production dashboard publication, user alert, production Ledger write, current calibration update, production dispatch, production ticker-universe change, or LLM provider call is required for this demo by default.

Working hypothesis: the family-office value is not an autonomous trading bot. It is a workflow layer that turns unstructured evidence into cited judgments, maps those judgments to holdings/themes/theses where useful, and tracks whether the alerts actually changed decisions or proved useful.