Fintheon Arbitrum — 5-Seat AI Deliberation Chamber

Fintheon Arbitrum is a deliberation engine — five AI analysts seated at a virtual round table, each arguing from a distinct role, then converging on a weighted verdict. When a high-impact RiskFlow item crosses the IV threshold, or when the 5 PM ET session cron fires, the chamber convenes automatically. You can also trigger it manually at any time. The output is a structured read you can act on: a consensus probability, a chamber confidence score, a plain-language digest, and a dissent summary if the seats disagreed. Arbitrum never places trades and never takes actions — the call is always yours.

Fintheon Arbitrum is unrelated to the Arbitrum Ethereum L2 blockchain network. When disambiguation is needed, this feature is referred to as “Fintheon Arbitrum.”

The five seats

Each seat runs from a distinct persona and analytical role. The weighted average of all five seat outputs produces the final consensus probability.

Seat	Persona	Role	Weight
1	Harper	Lead Analyst / CAO	30%
2	Oracle	Forecaster	30%
3	Feucht	Risk Manager	20%
4	Consul	Quantitative	10%
5	Herald	Bear Case	10%

Harper and Oracle carry the most weight because synthesis and probabilistic forecasting are the highest-leverage inputs for a verdict. Feucht’s risk management perspective acts as a floor check. Consul and Herald provide quantitative grounding and the contrarian bear case respectively.

When the chamber fires automatically

The chamber fires without your intervention in two situations: IV threshold trigger — When a RiskFlow item’s IV score reaches 8.5 or above and the source is a top-tier commentator or a tracked party of interest, the chamber opens immediately. This is the event-driven path: a single high-conviction headline from a credible speaker can bring all five seats to the table. Session cron — Every weekday at 5:00 PM ET, the chamber runs a session-end deliberation. This verdict feeds directly into the Post-Market Daily Brief (PMDB) as the “Chamber Read” section, which publishes at 5:15 PM ET.

If you see a Chamber Read in the PMDB but didn’t trigger Arbitrum yourself, it came from the 5:00 PM session cron. You can fetch the underlying verdict from the Aquarium panel or via the API.

Triggering the chamber manually

You have two paths to trigger a manual deliberation: From the UI — Navigate to Sanctum in Consilium and open the Aquarium panel. The deliberate button runs a chamber session on the current market context. Via API — Send a POST request with your question, an optional category, and optional context text. The chamber returns a verdict synchronously.

POST /api/arbitrum/deliberate
Content-Type: application/json

{
  "question": "Is the current /ES IV environment driven by macro risk or a sentiment overreaction?",
  "category": "volatility",
  "context": "VIX spiked 14% this session following FOMC minutes. /ES is at session lows."
}

{
  "verdict_id": "arb_20260508_174312_a3f",
  "persisted": true,
  "consensus_probability": 0.73,
  "confidence": 0.81,
  "dissent": {
    "seat": "Herald",
    "magnitude_pp": 18,
    "rationale": "Headline risk from after-hours commentary not yet priced in — sentiment tail remains open."
  },
  "digest_text": "The chamber reads the IV spike as predominantly macro-driven. Harper and Oracle converge on 73% probability that this is a genuine risk repricing, not sentiment noise. Feucht confirms the technical level at 5,420 as a key invalidation. Consul notes that earnings-season positioning exacerbates the move. Herald dissents: after-hours commentary risk remains open and the tail is not fully closed."
}

Request fields

Field	Type	Required	Description
`question`	string	Yes	The market question for the chamber to deliberate
`category`	string	No	Topic category, e.g. `volatility`, `macro`, `custom` (defaults to `custom`)
`context`	string	No	Additional context injected into each seat’s prompt
`rounds`	number	No	Override the number of deliberation rounds (uses chamber default if omitted)

Understanding the verdict output

Every verdict returns the same set of fields whether it came from an automatic trigger or a manual call. consensus_probability — A value between 0 and 1 representing the weighted average probability across all five seats. The Aquarium displays this as a percentage (e.g., 73%). A reading above 0.7 indicates strong chamber agreement on the directional thesis. confidence — The chamber’s confidence in its own consensus, also 0–1 (displayed as x/10 in the UI). Low confidence alongside a high consensus probability suggests the seats agreed directionally but with soft conviction — worth noting before acting. dissent — Present when one or more seats diverged meaningfully from the consensus. Shows which seat dissented, by how many percentage points (magnitude_pp), and the seat’s stated rationale. A large dissent magnitude is a signal to investigate the contrarian case, not ignore it. digest_text — The plain-language synthesis of the chamber’s deliberation. This is the text that appears in the Aquarium panel, in the IV widget hover peek, and in the PMDB Chamber Read section. trigger — How the chamber was invoked: scheduled (session cron), iv_threshold (RiskFlow event trigger), or manual.

Where to read verdicts

Sanctum → Aquarium panel — The primary surface for Arbitrum output. Shows the live verdict card with consensus percentage, confidence bar, digest text, and dissent badge. Individual seat cards display each analyst’s probability, confidence, and rationale. IV score widget hover — Hover over the IV score widget in Strategium to see a peek of the latest verdict’s digest text alongside the IV aggregate reading. PMDB Chamber Read — The Post-Market Daily Brief includes a dedicated “Chamber Read” section when a session verdict exists for the day. This section leads with the consensus and flags any dissent before the brief’s bullet points.

Fintheon Arbitrum produces analysis only. It does not submit orders, set stops, or take any position in the market. All trading decisions remain with you.

API reference

POST /api/arbitrum/deliberate
Content-Type: application/json

{
  "question": "Your market question here",
  "category": "macro",
  "context": "Optional context string"
}

GET /api/arbitrum/latest

GET /api/arbitrum/latest?trigger=session

GET /api/arbitrum/:id

GET /api/arbitrum/latest response

{
  "verdict": {
    "id": "arb_20260508_170012_c9e",
    "created_at": "2026-05-08T21:00:12.000Z",
    "consensus_probability": 0.68,
    "confidence": 0.76,
    "digest_text": "Session read: moderate bearish tilt on /NQ into overnight. Oracle and Harper align on 68% probability of continuation lower. Feucht flags 18,900 as the line — close below confirms. Herald notes social sentiment is crowded short, adding contrarian risk to the thesis.",
    "dissent": null,
    "trigger": "scheduled",
    "rounds_total": 3,
    "rounds_complete": 3,
    "phase": "complete",
    "seats": [
      { "role": "Lead", "probability": 0.71, "confidence": 0.82, "rationale": "..." },
      { "role": "Forecaster", "probability": 0.69, "confidence": 0.79, "rationale": "..." },
      { "role": "Future PM", "probability": 0.65, "confidence": 0.74, "rationale": "..." },
      { "role": "Quant", "probability": 0.63, "confidence": 0.68, "rationale": "..." },
      { "role": "Skeptic", "probability": 0.55, "confidence": 0.61, "rationale": "..." }
    ]
  }
}

Trigger type values for ?trigger= query parameter

Value	Description
`event`	IV threshold trigger (RiskFlow item ≥ 8.5)
`session`	Weekday 5:00 PM ET session cron
`manual`	Manually triggered via the UI or API

Get Started

Workspace

AI Features

Trading Tools

Configuration

Fintheon Arbitrum — 5-Seat AI Deliberation Chamber

The five seats

When the chamber fires automatically

Triggering the chamber manually

Understanding the verdict output

Where to read verdicts

API reference

​The five seats

​When the chamber fires automatically

​Triggering the chamber manually

​Understanding the verdict output

​Where to read verdicts

​API reference

The five seats

When the chamber fires automatically

Triggering the chamber manually

Understanding the verdict output

Where to read verdicts

API reference