Ask Signal Bureau

Answer Engine

Ask the engine

Ask anything about what we track. You get what we know, how we know it, and what new information would update the record at the next cycle — grounded in what's actually happening in the news right now. The first 25 answers are free on a key you mint yourself in one call — no card, for the life of the key (they don’t reset); plans lift the cap. For information only, not advice.

Answers are published at an unlisted public link so you can share them (recent ones may appear on this page; stale links expire to a fresh re-ask after ~36 hours). Don’t include personal or confidential information in your question. Privacy & deletion

Try:
WHEN THE QUESTION SUPPORTS IT, A READ CONTAINS
  • The state of play: what’s actually happening on your question, grounded in tracked coverage — never a hot take, and never a probability of ours.
  • The evidence: the live headlines, article text, and the desk’s own morning synthesis the read is grounded in, source-checked by a second model.
  • The market’s own price: when a prediction market’s terms actually match the question, its live price — quoted and attributed as the crowd’s number.
  • What would change this: the specific developments that would move the answer.

And when a question is outside what we track — or no market matches its exact terms — the engine says so and abstains from the missing piece rather than faking it.

THIS MORNING’S SYNTHESIS · 2026-10-03

Cros is the day's widest cross-domain story: 310 reports across 76 domains — auto, crypto defi, india south asia. Coverage ran 3.5 times its 7-day baseline. It reached business coverage for the first time. AllAfrica Latest Headlines: “Uganda: Passenger Train to Cross the Nile Again As Railways Prepares Kampala-Jinja Service”. Crowdfund Insider: “Cross-Chain Trading Protocol NEAR Intents Halts Services After $3.8 Million Exploit, Pledges Full Compensation”.

Google crossed 75 domains on 389 reports — auto, turkey, tech. It reached latin america coverage for the first time. Cloud Wars: “Hottest AI and Cloud Vendors: #1 Palantir, #2 Google Cloud, #3 Oracle”. The Decoder: “Deepmind researchers propose "Artificial Symbiotic Intelligence" as an alternative to the singularity”.

Thanksgiving crossed 12 domains on 13 reports — cloud computing, artificial intelligence, auto. Coverage ran 9.1 times its 7-day baseline. It reached business coverage for the first time. SiliconANGLE: “Anthropic targets pre-Thanksgiving IPO despite safety concerns and regulatory scrutiny”. Crypto Briefing: “Anthropic targets mega-IPO before Thanksgiving, valued at $965B”.

Trump crossed 70 domains on 529 reports — auto, intelligence, middle east. TIME: “Why the Trump Administration Is Being Accused of Interfering in Brazil's Election”. The Cryptonomist: “Trump announces voluntary AI safety accord with six tech giants”.

Uss George Washington crossed 10 domains on 14 reports — iran regime, middle east, auto. Coverage ran 14.0 times its 7-day baseline. It reached middle east coverage for the first time. Al-Monitor — Middle East: “Aboard the USS George Washington, young sailors adjust to war with Iran”. The Guardian: “Young sailors on the USS George Washington adjust to life at war – in pictures”.

Written each morning by the desk from tracked prediction markets and coverage. Informational only.

How an answer is built

A worked example

Reading as: Ship operatorLedger built 3 Oct 2026

The box above answers live from the record. Below is a worked example of how an answer is built, with every count traced to its reports.

GET https://sigb.ai/api/ask?expand=claims.reports.quotes (POST)

{
  "question": "Is tanker traffic through Hormuz back to normal?",
  "viewpoint": "ship-operator",
  "expand": "claims.reports.quotes"
}

GET 200 application/json

{
  "schema": "sb.answer.v1",
  "question": "Is tanker traffic through Hormuz back to normal?",
  "state": "example",
  "answer": "No. Ships cross at a small fraction of the pre-war rate of about 125 a day, while oil-flow estimates differ on how near normal supply is.",
  "series": {
    "dailyCrossings": [
      {
        "date": "2026-08-31",
        "value": 5
      },
      {
        "date": "2026-09-08",
        "value": 9
      },
      {
        "date": "2026-09-09",
        "value": 6
      },
      {
        "date": "2026-09-10",
        "value": 7
      },
      {
        "date": "2026-09-15",
        "value": 12
      },
      {
        "date": "2026-09-16",
        "value": 3
      },
      {
        "date": "2026-09-17",
        "value": 4
      },
      {
        "date": "2026-09-22",
        "value": 3
      }
    ],
    "tenDayAverage": [
      {
        "date": "2026-09-09",
        "value": 12
      },
      {
        "date": "2026-09-14",
        "value": 14
      },
      {
        "date": "2026-09-23",
        "value": 15
      }
    ],
    "disputed": {
      "date": "2026-09-09",
      "values": [
        6,
        11
      ]
    },
    "newest": "2026-09-22",
    "prewar": 125
  },
  "notFound": [
    "No daily count after 22 Sep.",
    "No tanker-only count.",
    "No first-hand ship tracking."
  ],
  "ledger": {
    "claims": 1598,
    "computedAt": "2026-10-03T00:20:25+00:00"
  }
}

Example response shape: state says the answer wording is an example. Every series value is read from a claim in the ledger.

GET answer/where-from?expand=+1

{
  "level": "answer",
  "label": "Where this answer comes from",
  "children": [
    {
      "level": "where from",
      "label": "8 daily counts and 3 averages",
      "children": {
        "count": 12,
        "unfold": "expand=+1"
      }
    },
    {
      "level": "where from",
      "label": "1 report",
      "children": {
        "count": 1,
        "unfold": "expand=+1"
      }
    },
    {
      "level": "where from",
      "label": "JPMorgan, as reported",
      "children": {
        "count": 1,
        "unfold": "expand=+1"
      }
    },
    {
      "level": "assertions",
      "label": "Outlets disagree",
      "children": {
        "count": 2,
        "unfold": "expand=+1"
      }
    }
  ]
}

Each level folds below one step; ask for one more, down to the article link.

Example the answer wording, tiers, chair lines and follow-up questions below were written by hand to show the shape of an answer; every count, date, quote and link inside it is real

Is tanker traffic through Hormuz back to normal?

Reading as: Ship operator · from 2,647 documents and 1,598 claims, 1 Sep to 1 Oct
Ledger built 3 Oct 2026

No. Ships cross at a small fraction of the pre-war rate of about 125 a day, while oil-flow estimates differ on how near normal supply is.

How sure: each crossing count is a single press report of one data firm’s figures. On oil flows, estimates differ.

Daily crossings, 31 Aug to 1 OctFilled dots: daily crossings as reported, between 3 and 12 vessels a day. Hollow squares: the reported 10-day average, between 12 and 15. On 9 Sep two reports give 6 and 11. No daily count after 22 Sep. Before the war, about 125 a day.051015no dailycount6 or 11?31 Aug1 Oct↑ before the war: about 125 a day, far above this scale
  • daily count
  • 10-day average
  • two counts for one day

Real every point: a number as written in a claims_hormuz.json claim, on the day the sentence names

Conditions

Real values and sourcesExample tiers

Could not find
  • No daily count after 22 Sep. The newest daily figure in the ledger is for 22 Sep; averages run to 23 Sep.
  • No tanker-only count. Reports count commodity vessels, commercial vessels or all large ships; none counts tankers alone.
  • No first-hand ship tracking. Every figure here is a press report of a data firm’s count.

Same question, other chairs

  • Ship operator Crossings are a small fraction of normal, and many move under US escort.
  • Insurer War-risk premiums were reported on 24 Sep: “War-risk insurance premiums for Hormuz voyages are running between 6% and 9%.”
  • Energy trader Oil-flow estimates run close to normal even though crossings are low. The reports do not say why.

Follow-up questions from the same data

Same question, same answer. The first time anyone asks a new question, the answer is filed with its date. The next person who asks gets that same answer until the next update.

Real every claim, count, date and link: claims_hormuz.jsonReal gaps: absences in the same file