Pick a path once — Sarah is the beta default for calm forecasting. Architect opens the full workspace.
Not advice to act — forecasts and proof stay honest. You can switch to Architect anytime via All details.
Loading account…
Use this before acting. Write your decision, smallest next step, and unacceptable outcome — human check, not proof.
The main template answer — expand when ready.
Live pattern from the graph — expand for the full read.
A fast, honest read from what KELLY sees right now — not a final proof score.
Delegation, paper trades, broker stub — practice only.
Human-in-the-loop authorization receipts link delegated paper acts to forecasts — firewall unchanged.
Record paper fills as hash-chained receipts — practice arithmetic only, not audited live PnL.
Delegation, execution, and broker-stub chains — posture signals only, not claim proof.
Simulated broker fills require human-confirmed delegation — not live connectivity.
25-second human check before a real decision.
Take 25 seconds: say the decision, the next-day action, and what would be unacceptable.
Pick a focus to narrow the answer.
Engine heartbeat and proof posture.
Forward-settled scorecard when enough outcomes exist.
Forward-settled scorecard gated by faithfulness posture and ensemble contradiction density.
How the engine thinks — expand when you want the story.
A causal memory engine that writes things down before outcomes arrive, then scores itself honestly.
Local weather, hazards, and nearby teams from this device.
Tap once to let KELLY use this device's approximate location for local weather, hazard, public-health, transport, and nearby-team views. The browser will ask permission first.
Horizon, accuracy, learning loop, and historical reference memory — one slice at a time.
Same stress protocol as the Answer tab — evaluate before acting on calibration reads.
Unified forward proof runway — settlement drive, unlock ladder, and supersession rung in one receipt.
Past public observations build reference classes. They help KELLY compare today with what already happened, but they do not count as live proof unless a forecast was written before the outcome.
21 books in engine loop order — perceive → calibrate → reason → decide → adapt → remember — wrapped by endure and align.
Engine pulse first, then feeds and credentials.
Heartbeat first: loop health, proof samples, and whether the engine is keeping up.
What outside signals are arriving, accepted, or stale.
Credentials and transport setup for each feed.
Read in order: honesty → evidence → audit → math → discovery → future engines.
Full mirror of GET /api/transparency — forward proof, internal ledger, wired flags, and every known gap.
How strong is the live evidence before any claim is written?
Targeted reinforcement for the lowest-confidence nodes without changing the canonical writer.
Live role audit. Primary writes the official claim; Shadow, Critic, Replay, and Continuity write evidence rows only.
These are the claims the engine is allowed or forbidden to make from the evidence it has right now.
Calibration math and causal identification spine.
Structural checks on the current model packet.
Correlation and lag become ranked hypotheses only. Hard causal language waits for intervention or valid identification evidence.
Edges can be sharpened as provisional adapters; hardening is locked behind stronger evidence.
Gated v2 scaffolds for future audit engines. Not a second vote on the live answer; env-off until Phase 3 clears.
Continuity receipts, durable storage, and the upgrade gates that unlock the next build phase.
Live continuity receipts, restore checks, and what the engine is watching across restarts.
Database, backup, cache, and redundant copy state that keeps history from disappearing.
These are measurable upgrade gates. A card changes only when the engine evidence changes.
Manual ticks, decision card, and raw packet — kept separate from the live reading flow.
Use this to show a new observation entering the system. Signals are normalized 0-1.
Developer duplicate of the progressive paper-fill form — same POST /api/execution/paper route.
© 2026 Aaron Moyer. All rights reserved. Causalea™ and KELLY™ are trademarks of Aaron Moyer.
This showcase is proprietary decision-support and research infrastructure. It is not open source, financial advice, betting advice, emergency guidance, medical advice, legal advice, guaranteed profit, guaranteed prediction, or proof of true causality from correlation alone.