Understand
the signal.
Start with the reported change, why it could matter and what your agent should check next. The detailed measurements are available when you need to inspect the decision.
What the connection adds
Rarebird gives a compatible agent source-linked reporting, a conditional interpretation, market checks and reasons to review, wait or reassess. The agent can bring those into your conversation. A review signal asks for investigation; it does not place a trade or deliver a notification.
Connect your agentRead the plain-language view
| View | Meaning |
|---|---|
| Review now | A fresh information change, measured response and applicable checks qualify for a closer review. |
| Keep watching | The development may matter, but confirmation is missing, the move is advanced, or the information is no longer fresh. |
| Reassess | The evidence contradicts the premise or measured behavior goes against the hypothesis. |
| Waiting for data or analysis | A reliable current view is not available. Older evidence may remain visible, with alerts held. |
| Outside this focus | The story does not match the selected business, funding or risk context. |
Explore the optional numbers
On the intelligence page, open See the numbers and checks for the exact values, thresholds, timestamps and model baseline. Open Read the reasoning and sources for the full hypothesis, original reports, uncertainty and revision history.
| Measurement | What it describes |
|---|---|
| Asset return | How the identified venue market changed over the displayed window. |
| Market comparison | A named comparison market. Stock tokens use QQQx; crypto uses peers that exclude the asset itself. |
| Residual move | The part of the move left after adjusting for the fitted comparison-market exposure and baseline. |
| Relative volume | Recent traded volume compared with the median of earlier windows. |
| Spread and book impact | Observed trading costs estimated from the displayed order book. These exclude fees and execution latency. |
| Beta and fit | The estimated relationship to the comparison market, its sample size and descriptive fit. These are not causal estimates. |
The default graph shows observed asset movement against its comparison market. It is not a forecast. Stock graphs use related xStock venue prices, not a consolidated underlying-share tape.
Pilot coverage and initial rules
The pilot covers NVIDIA, Apple, Microsoft, Alphabet, Bitcoin and Ethereum. Stocks are measured through identified Kraken xStock USD pairs; Bitcoin and Ethereum use same-venue spot markets and self-excluding peer comparisons.
The initial policy requires fresh reporting and completed analysis, a meaningful information change, a supported reported condition, a pre-information baseline and current traded prices. Confirmation also checks participation, spread, book costs and whether the move is already extended or fading.
For the exact heuristic thresholds, fitting windows, pair identities and collection schedule, read the quantitative methodology. Thresholds remain research rules awaiting calibration.
Find the move before choosing a story
The independent market scan can start an investigation without a headline thesis. It looks for unusual 15-minute, 30-minute and one-hour moves relative to the market comparison, then checks them against earlier scored windows. Each comparison uses a model fitted before that window starts.
Volatility context helps distinguish an isolated dislocation from a changing environment. Recent checked interpretations can support, oppose or leave the move unexplained; that link does not establish its cause. The same existing venue data powers the scan.
A detected episode freezes its first reference and fitted model, then measures later observations. Repeated scans do not restart the same episode. Historical rarity is not the probability of a profitable trade, and discovery does not bypass the review-alert checks. Open Independent discovery and frozen outcomes inside the optional numbers for the research record.
Agents receive this through quantitative.research, including its scan time, quality state, windows and episode records, even when no news construction is available.
What the forward record measures
Decisions and reference quotes are frozen before 15-minute, one-hour and four-hour outcomes are observed. The detailed view separates all constructions from qualified alerts and counts pending or missed observations. Returns are before fees and are not executable profit and loss or evidence of outperformance.
Read it through your agent
Call newsbeat_intelligence with a resolved beat_id and an optional context: fundamentals, liquidity or risk. REST accepts the same selection:
GET /api/v2/intelligence?beat_id=b_bb964843350e&context=fundamentalsResponses include signals for reporting and interpretation, constructions for context-qualified decisions, quantitative for measurements and forward records, and alerts for the measured review journal. Inspect source freshness, analysis status, measurement times and hold reasons. Public reads do not start generation.