trusted decision conversation
Ask KELLY command center
The smartest friend in the room should also show its work.
Ask KELLY like a conversation. KELLY listens, separates what is known from what is uncertain, checks reality when facts are needed, and keeps the proof trail close enough to inspect without turning the page into a control room.
Answer mode
ACT / TEST / WAIT
Proof posture
receipt or watch state
Write boundary
KELLY authority only
answer contract
What happens after you ask?
KELLY forms a source plan, compares your question with live and historical evidence, returns a bounded answer, and creates a private watch lane when the answer needs more reality to close.
regular human audit path
Ask, inspect, and understand the answer.
The page is organized around the same contract every serious KELLY answer should satisfy: a clear question, a bounded response, a visible proof path, and a strict no-public-write boundary.
Ask with context
State the decision, horizon, stakes, place or sector, and what outcome would count as settled.
Get a bounded answer
KELLY separates facts, forecast, uncertainty, next action, and what would change the answer.
Follow the proof path
Serious answers point to source posture, proof receipt, replay, settlement, calibration, or a watch state.
public writes blocked
Your question can guide a response, but public users do not write durable memory or mutate GraphState.
better prompt shape
Decision + horizon + stakes + settlement criteria.
Example: "Should I act now, test first, or wait for the next 72 hours, and what source would change your answer?"
audit expectation
A strong response should show what it knows, what it does not know, what it is watching, and where a user can inspect proof. A weak response should degrade honestly instead of hiding uncertainty.
What users can ask KELLYplain language
KELLY is useful when the question can be bounded, checked, forecast, replayed, or settled later. Users do not need to know the internal machinery; the UI should show what KELLY did with the request.
Make a decision under uncertainty
Use KELLY when you need a next action, a risk check, and a reason to wait or move.
KELLY does: KELLY frames the choice, names the uncertainty, checks available evidence, and returns a bounded answer.
Proof appears: Ask KELLY shows the answer; Glass Lab shows proof, replay, source mode, and protected boundaries.
Boundary: The user request guides attention. It does not directly write memory or GraphState.
Forecast a future event
Use KELLY when a question can later be scored against reality.
KELLY does: KELLY keeps the user forecast and KELLY forecast separate, then prepares a settlement path.
Proof appears: Superforecasting shows probability lanes; calibration shows whether reality improved the model.
Boundary: A user forecast is a probe request. It is not treated as ground truth.
Inspect market predictions
Use Markets Lab for stock and crypto forecasts, source links, cones, and model paper P&L.
KELLY does: KELLY reads approved market data, separates quote from forecast, and tracks settlement/P&L.
Proof appears: Markets Lab should show source, timestamp, forecast, settlement state, and proof link per asset.
Boundary: Read-only analysis. Not financial advice and not trade execution.
Private user lane
Users can have private questions, forecasts, and context beside KELLY. KELLY's shared learning comes from closed reality loops, not from exposing a user's private data.
Conversation flow
Ask
You pose a bounded question in plain language.
Source plan
KELLY chooses what reality to consult before answering.
Evidence
Live and historical artifacts are compared under uncertainty.
Answer
A bounded reply separates known from uncertain.
Watch
Open questions get a private watch lane until reality closes.
Learning Loop Passportcheckingpublic no-write
Current public-safe loop
live_brain_memory_worker
waiting_for_reality
Checking the next public-safe loop snapshot.
memory receipts
0
replay-ready closures
0
rollback ready
0
quarantine ready
0
Next proof checkpoint
checkingSource
KELLY keeps moving through approved source intake, hypothesis creation, reality closure, replay receipts, and memory-cell receipts. Users can watch and ask; public users cannot write ground truth.
Source
checkingObservation
checkingHypothesis
checkingProof
checkingSettlement
checkingReplay
checkingMemory
checkingPassport progress: 0/7 stages visible. This is proof visibility, not a public mutation control.
Open full Glass LabArtifact trace standard
Every KELLY lens should produce an artifact that can be traced backward to sources and proof, and forward to replay, settlement, calibration, memory eligibility, or quarantine.
Answer artifact
What the user sees
Source plan
Where reality is checked
Proof receipt
Why KELLY said it
Replay path
How to rebuild it
Settlement / calibration
How reality scores it
WOW feature
What would change my mind?
Not another chat box: this lens has its own instrument, forecast posture, and artifact trail.
Users want
A friend-level answer without fake certainty: act, wait, or test, with proof one click away.
Proof that matters
Every important output should open backward to source and forward to replay, settlement, calibration, or memory lineage.
Open proof targetConversation flow
Ask
You pose a bounded question in plain language.
Source plan
KELLY chooses what reality to consult before answering.
Evidence
Live and historical artifacts are compared under uncertainty.
Answer
A bounded reply separates known from uncertain.
Watch
Open questions get a private watch lane until reality closes.
High-value forecast
Decision forecast
Ask a real question, then save it as a watched question if the answer needs future evidence.
Use presetAsk KELLY here
In-lab conversation
Name the decision, horizon, stakes, and what would change your mind.
Ask in this labProof artifact
Answer card + watch lane
Answer boundary, source posture, watched-question state, and replay when available.
Open proofDedicated app
Open full lens
Use the dedicated sector app when you want more room for charts, timelines, proof cards, and live updates.
Open Ask KELLY appLens artifact passport
Every serious output must trace to an artifact. The answer stays simple on the surface, the heartbeat visible in the top bar proves the engine clock is present, and these links open the source, health, guide, user controls, or Glass Lab proof when the user wants depth.
Dedicated app
Open Ask KELLY app opens the standalone Ask KELLY surface when a sector needs more room.
OpenLive forecast API
Machine-readable sector output for forecast cards, proof widgets, or status monitors.
Not exposed yetArtifact spine
Eight-step proof chain for this lens: source, observation, hypothesis, proof, replay, settlement, calibration, memory.
Not exposed yetHealth check
Fast check that the standalone lab is reachable and serving the current runtime.
Not exposed yetUser controls
Export, delete, location, privacy, and feedback controls stay visible from every lens.
OpenAsk KELLY inside this lab
The question stays in this sector lens. KELLY answers here with a bounded answer, visible uncertainty, proof path, and the next reality check when more evidence is needed. No page jump, no generic redirect.
KELLY response
This panel shows the answer inside the lab. It does not create public ground truth, memory, or GraphState writes. The proof link is the handoff when you want the moving parts under the magnifying glass.
How KELLY should answer in this lens
Use plain decision language. Name the target, time horizon, stakes, uncertainty, and what evidence would change the answer. KELLY should respond with a bounded artifact, not an engine dump.
Source families
Live signals
Proof windows
User workflow
Ask a plain-language decision question with stakes and horizon.
Add a private forecast or location context without writing public truth.
Save unresolved questions as browser-local watches.
KELLY work behind the lens
Separates facts, forecast, uncertainty, and next action.
Keeps user context out of Prime memory unless public-safe research intent is approved.
Links proof, replay, source posture, or honest degradation.
Artifact trace rail
Every lens output should be traceable backward to source and proof, then forward to replay, settlement, calibration, memory eligibility, or quarantine.
Answer
visible output
Source
reality path
Proof
why it said it
Replay
rebuild path
Score
reality grades it
Lab contract
mode
read-only lens
authority
KELLY Prime
proof target
#prime-research-queue
Public boundary
Users guide attention. They do not directly mutate KELLY memory, GraphState, or canonical reality.
Why this is different
KELLY does not just answer from text. Each lens keeps the answer connected to source posture, proof, replay, settlement, calibration, and memory eligibility so the user can see how the engine knows, what it is watching, and what would change its mind.