One KELLY engine. Many lenses. One purpose. Labs are not separate truth engines. They are focused ways to ask, forecast, inspect markets, watch proof, and understand memory without overwhelming the user.
Each public app is a focused way to talk to KELLY. This strip probes live control contracts every 30 seconds and labels manifest-only counts when a probe is still warming up.
Lens map
8
focused apps
Ground truth routes
3
www, apex, Railway
Live app contracts
8
privacy and writes
Readiness APIs
8
health APIs
Artifact spines
8
proof trace APIs
In-lab Ask lanes
8
sector answers stay in context
Cognition organs
4
science, hypothesis, memory
Public writes
protected
ground truth safe
operator matrix
This is the public routing map for the KELLY organism. Labs remain specialized read-only windows; KELLY keeps write authority, central persistence, replay, proof receipts, and memory lineage.
Write authority
KELLY only
Public labs inspect, ask, forecast, and explain without writing ground truth.
Durable state
Central Neon
Proof, replay, forecast, settlement, memory, and lineage stay in KELLY read models.
Lab volumes
0 required
Standalone labs do not need source-of-truth Railway volumes.
Proof route
Glass Lab
Every lab row points back to its public proof shelf.
Ask KELLY
Conversation
Conversation and decision help
operational lens
central read model 64%
answer chain
Shows the answer chain without letting a public user write ground truth.
Open Glass Lab proofPostgres
Optional private user/session store only
Volume
No ground-truth volume
Browser-local watched context plus sanitized intent
Boundary
No KELLY memory, GraphState, settlement, or source writes
Superforecasting
Forecast scoring
Forecast discipline and scoring
operational lens
central read model 82%
calibration lane
Keeps user probability separate from KELLY probability so calibration stays honest.
Open Glass Lab proofPostgres
Central forecast, settlement, and calibration read models
Volume
No ground-truth volume
Separated user and KELLY forecast lanes
Boundary
User forecasts remain separate from KELLY truth
Markets Lab
Markets
Market forecast lens
operational lens
central read model 78%
market proof
Every visible forecast should point back to source, timestamp, model posture, and settlement status.
Open Glass Lab proofPostgres
API-first central quote, forecast, and source receipts
Volume
No ground-truth volume
Immutable source receipts through KELLY proof spine
Boundary
Read-only analysis; no trade execution or truth writes
Glass Lab
Transparency
Full transparency console
operational lens
central read model 96%
live proof index
This is the live window into receipts, replay, event ledger, memory, calibration, authority, and link health.
Open Glass Lab proofPostgres
Direct-first read-only proof and replay views
Volume
No ground-truth volume
R2/object evidence and central event receipts
Boundary
Observer console only; KELLY writes internally
Betting and Odds Lab
Probability
Odds and sports probability
operational lens
central read model 76%
odds settlement
Every visible forecast points back to odds snapshot, probability conversion, and outcome receipt when available.
Open Glass Lab proofPostgres
API-first odds snapshots and settlement receipts
Volume
No ground-truth volume
Immutable odds/source receipts through KELLY
Boundary
Read-only probability analysis; no bet execution
Weather Lab
Hazards
Place and hazard awareness
operational lens
central read model 74%
weather closure
Every snapshot names source freshness, place authority, and the closure path back to Glass Lab.
Open Glass Lab proofPostgres
API-first hazard, alert, and source-freshness read models
Volume
No ground-truth volume
Object evidence for alerts, maps, and receipt bundles
Boundary
Awareness window only; not emergency instruction
Policy and Economics Lab
Public records
Public records and macro signals
operational lens
central read model 80%
citation record
Every claim should tie to citations, effective dates, and source authority.
Open Glass Lab proofPostgres
API-first public-record and macro signal read models
Volume
No ground-truth volume
Source citations, effective dates, and immutable records
Boundary
Read-only research; no legal or tax authority
Science and Research Lab
Research
Evidence and research reasoning
operational lens
central read model 84%
evidence lattice
Every research claim should expose source metadata and falsification posture.
Open Glass Lab proofPostgres
Scoped science evidence and falsifier read models
Volume
No ground-truth volume
Paper metadata, contradiction receipts, and source bundles
Boundary
Research support only; no public truth approval
Hypothesis Lab
Discovery
KELLY-generated questions
operational lens
central read model 88%
unknown ledger
Every hypothesis should expose falsifier, source plan, settlement target, and replay path.
Open Glass Lab proofPostgres
Direct-first unknown, hypothesis, and source-plan views
Volume
No ground-truth volume
Falsifier and closure receipts linked to Glass Lab
Boundary
Public users cannot approve memory or GraphState writes
Memory and Lineage Lab
Memory
Memory ancestry and repair
operational lens
central read model 94%
cell lineage
Every cell should connect to its replay receipt, lineage, and quarantine posture.
Open Glass Lab proofPostgres
Direct-first memory, replay, lineage, and quarantine views
Volume
No ground-truth volume
Birth-to-root cell dossiers and immutable replay bundles
Boundary
Memory writes are internal KELLY authority only
Forecast first
Each lab opens with the highest-value forecast artifact for that sector, then reveals proof.
Ask in context
Ask KELLY with the target, horizon, stakes, and evidence that would change the answer.
Not another chat box
Every lens has a sector instrument, a WOW feature, and a proof trail that opens in Glass Lab.
Proof that matters
Every lens card separates the local lens, dedicated app, and proof shelf.
Each card is a sector entry point. Forecast output, dedicated app, and proof are separated on purpose.
Conversation and decision help
Ask in plain language. KELLY answers with a bounded response, risks, next step, and proof path when available.
WOW feature
What would change my mind?
Users want
A friend-level answer without fake certainty: act, wait, or test, with proof one click away.
Best first question
Ask a real question, then save it as a watched question if the answer needs future evidence.
Shows the answer chain without letting a public user write ground truth.
Forecast discipline and scoring
Compare a user forecast beside KELLY's independent forecast, then score the outcome when reality settles.
WOW feature
Dual-lane forecast scoreboard
Users want
Calibration honesty: user probability and KELLY probability scored separately when reality closes.
Best first question
Turn an uncertain future event into a forecast with a closing condition and settlement date.
Keeps user probability separate from KELLY probability so calibration stays honest.
Market forecast lens
Track default stocks and crypto with forecast cones, source links, proof links, and model paper P&L.
WOW feature
Freshness-first forecast cone
Users want
What moved, whether the feed is fresh, and why the forecast posture changed before any chart drama.
Best first question
Pick a tracked asset and inspect the source, forecast posture, proof link, and paper P&L.
Every visible forecast should point back to source, timestamp, model posture, and settlement status.
Full transparency console
Watch KELLY ingest approved sources, form hypotheses, close loops, replay proof, and expose memory-cell visibility.
WOW feature
Live Proof Index
Users want
A microscope for skeptics: proof, worker health, loop freshness, blockers, and honest stall states.
Best first question
Start with the Learning Loop Passport, then open the feed worker, Prime queue, and replay panels.
This is the live window into receipts, replay, event ledger, memory, calibration, authority, and link health.
Odds and sports probability
Inspect line movement, implied probability, KELLY posture, and settlement receipts.
WOW feature
Line movement to implied probability
Users want
Odds translated into probability, KELLY disagreement, and settlement path without bet slips or hype.
Best first question
Choose a market and inspect whether KELLY sees a meaningful probability divergence.
Every visible forecast points back to odds snapshot, probability conversion, and outcome receipt when available.
Place and hazard awareness
Check hazard risk near you with proof-linked reads from NOAA, NWS, and public feeds.
WOW feature
Place-aware hazard window
Users want
Plain-language hazard chance, what to watch, and official alerts for your area.
Best first question
Check weather and disaster risk near you — proof-backed reads from NOAA and NWS.
Every snapshot names source freshness, place authority, and the closure path back to Glass Lab.
Public records and macro signals
Inspect regulations, public spending, macro pressure, and policy uncertainty.
WOW feature
Citation-first policy card
Users want
What changed in official records, who is affected, and what would falsify the thesis.
Best first question
Ask which public records changed and what would falsify the policy thesis.
Every claim should tie to citations, effective dates, and source authority.
Evidence and research reasoning
Inspect papers, claims, contradictions, falsifiers, and research trail.
WOW feature
Evidence lattice
Users want
Support, contradiction, replication posture, and falsifiers instead of authoritative-sounding summaries.
Best first question
Pick a claim and inspect what evidence would strengthen or break it.
Every research claim should expose source metadata and falsification posture.
KELLY-generated questions
Watch observations become falsifiable hypotheses, source plans, and reality checks.
WOW feature
KELLY-generated question forge
Users want
High-value questions with closure criteria, source plans, and a clear way for theories to die.
Best first question
Open the highest-value hypothesis and inspect what would change KELLY's mind.
Every hypothesis should expose falsifier, source plan, settlement target, and replay path.
Memory ancestry and repair
Inspect memory cells, generations, parent-child links, quarantine, and proof ancestry.
WOW feature
Birth-to-root cell dossier
Users want
Where a memory came from, whether it replayed, and whether it was superseded, repaired, or quarantined.
Best first question
Pick a cell and trace its parent, receipt, replay state, and eligibility.
Every cell should connect to its replay receipt, lineage, and quarantine posture.
