Data handling, as documented while the surface was liveWhale Watcher combines a scheduled activity baseline with wallet reads from live providers, briefly cached. Its signals describe activity; they are not findings of wrongdoing. See what you can use today for every surface side by side.
What it does
Whale Watcher is a supporting signal surface for Stablecoin Vault readiness. It focuses on large stablecoin and treasury-like wallet monitoring, assembling multiple analytical layers for a single wallet address into one view — activity history, behavioral signals, AI-assisted narratives, token holdings, and protocol usage patterns — so the movement of significant wallets can be carried into a vault-readiness review. It is a read-only intelligence surface: no custody, no wallet keys, no account required.What you can learn
- How active a wallet has been — per-day on-chain activity intensity over a rolling 12 weeks.
- Which behavioral patterns its activity resembles — five deterministic signals, each with a confidence level and the evidence behind it.
- What that adds up to, in plain language — narrative cards built from those signals, with their sources and caveats.
- What it holds and which protocols it uses — stablecoin holdings and the share of on-chain volume by DeFi protocol.
- What its long-term exposure looks like — the Quantum Exposure Card and a recommended migration path.
How to read it
The workspace is organized into vertical sections, top to bottom:1
Wallet header
ENS, address, and observability badge.
2
12-week activity heatmap
Per-day on-chain activity intensity over a rolling window.
3
Adversarial signals panel
Deterministic behavioral exposure heuristics.
4
Narrative cards
Template-based, optionally AI-rewritten, summaries of the signals.
5
Quantum Exposure section
Quantum Exposure Card and Quantum Vault Readiness Card.
6
Wallet holdings strip
Token portfolio, as balances or trading exposure.
7
Protocol affinity bar
Share of on-chain volume by DeFi protocol.
12-week activity heatmap
The heatmap visualizes per-day on-chain activity intensity over a 12-week rolling window. Each cell represents one day; color intensity corresponds to that day’s activity score. While scheduled data loads, the activity section shows a skeleton heatmap grid with the label “Indexing scheduled wallet activity.” and asyncing badge. The full grid renders once the scheduled response is available.
Adversarial signals
Adversarial signals are behavioral exposure heuristics derived from wallet transaction history and the 12-week scheduled activity data. These are deterministic, explainable heuristics.
Each signal carries a score, a confidence level that reflects data completeness rather than accusatory certainty, a reason written in observation language (“resembles”, “may indicate”, “increases exposure”), and the on-chain evidence that produced the score. When no signal meets the rendering threshold, the UI shows a low-noise empty state.
Narrative cards
The narrative engine is deterministic — it templates from signal evidence and never invents numbers, protocols, or chains. An optional AI layer may rewrite the prose but must leave evidence values unchanged.AI narrative generation is optional. If no AI provider is configured, the workspace displays template-based narratives only.
- Headline (≤120 characters, derived from primary signal evidence)
- Body and key points
- Data source badges with freshness information
- Confidence level and caveats
Wallet holdings strip
The strip shows a wallet’s stablecoin holdings from a WalletWall-owned BigQuery cohort: the newest daily snapshot of the top Ethereum wallets by par stablecoin balance, with holdings of at least 1,000 at par, at most 15 holdings shown and the cut disclosed.Protocol affinity bar
The Protocol Affinity Bar shows the fraction of the wallet’s on-chain volume that flowed through each DeFi protocol (Uniswap, Aave, Curve, etc.). It is derived from the baseline and live event data assembled by the signal engine.Quantum section
The Quantum Exposure section renders two cards stacked vertically:- Quantum Exposure Card — composite score ring, per-component breakdown, behavioral signals, source caveats, and the policy simulator.
- Quantum Vault Readiness Card — recovery readiness signals, dormancy feed findings, and the Migration Path panel. The migration path is derived at render time from the wallet’s exposure score, value at risk, dormancy, and wallet structure. It recommends one of:
monitor,fresh-wallet,split-wallet,multisig,treasury-custody, orvault-prototype. Whenvault-prototypeis recommended, the panel surfaces a vault candidate row and the WalletWall Vault research disclosure automatically. The panel also includes a Vault Readiness block showing a Stablecoin Vault recommendation for this wallet.
Data freshness and provenance
Signals degrade gracefully when the baseline is unavailable, and every narrative card carries data source badges with freshness information. See Data Serving Architecture for which of these inputs are live provider reads, which are scheduled snapshots, and when a fallback answers.
Important boundaries
- Signals are not findings. They are not findings of wrongdoing and do not imply intent, identity, or legal status.
- Absence from the holdings cohort is not a zero balance. A wallet outside the cohort shows no holdings strip.
- AI never changes the evidence. The optional AI layer may rewrite prose, but it must leave evidence values unchanged.
- The homepage signal feed is illustrative. The homepage animates sample on-chain alert rows and links into the Whale Watcher route when a row is selected. It is a product-navigation preview, not a live provider stream and not a scheduled query result. The source-backed activity, deterministic signals, and source/caveat badges render after a wallet is opened.
Technical reference
Data flow
Signals degrade gracefully when the baseline isnull. AI narrative rewriting is optional and only active when a provider key is configured. See Signal Engine for the full taxonomy.
Implementation notes
Signal fields and keys
Signal fields and keys
Signals are computed by the behavioral-signals engine. Each carries
score (0.0–1.0), confidence, reason, and evidence. Signals with score < 0.3 are computed but not displayed.Expected Dune row fields (16 columns)
Expected Dune row fields (16 columns)
address, label, category, tx_count_48h, unique_receivers_48h, usd_volume_48h, last_seen, activity_tier, activity_day, week_number, day_of_week, tx_count_day, unique_receivers_day, usd_volume_day, intensity_score, day_activity_tierHeatmap color intensity corresponds to the intensity_score field.Signal engine dependency
Signal engine dependency
Whale Watcher depends on the signal engine to produce
WalletSignal[] facts before narratives and the Protocol Affinity Bar can be rendered. The narrative engine converts those facts into structured NarrativeCard objects. The signal engine requires:- A
HistoricalWalletBaselinefrom the Dune scheduled query (may benull— signals degrade gracefully) - Live wallet events from Alchemy or Etherscan
Holdings route
Holdings route
The Wallet Holdings Strip reads
/api/wallet-portfolio. In production the route serves the owned cohort, labelled provider: "bigquery" and dataMode: "balance". A wallet outside the cohort returns inDataset: false.The legacy scheduled Dune feed remains the route’s fallback. It supports two query shapes, which the API detects from the columns present: balance-denominated positions, or historical DEX trading exposure.Related pages
Holder Wall
Ownership map and holder analytics — treemap, leaderboard, and wallet drawer.
Quantum Intelligence
The full Quantum Exposure Score specification, migration path table, and approved framing behind the Quantum Exposure section.
Signal Engine
The deterministic fact layer that produces
WalletSignal[] from the scheduled baseline and live wallet events.Stablecoin Vault & Vault Simulator
The vault-readiness review you can take a wallet to after reading its signals here.
Wallet Profile
The compact single-wallet overview that hands off to this workspace.
Data Serving Architecture
Which of Whale Watcher’s inputs are live provider reads and which are scheduled snapshots.

