Killfeed · Lab
Killfeed Lab
Public experiments in signal extraction
Lab is where Killfeed tests the Signal Extraction Engine against noisy, ambiguous, and high-stakes domains.
These are not opinion pages and not finished products. They are structured experiments in turning fragmented public inputs into signals you can scan, inspect, and challenge.
What belongs in Lab
High-noise domains
Conflicting inputs, inconsistent reporting, fast-changing conditions.
Public-signal methods
Official alerts, public advisories, structured updates, and attributable sources.
Transparent outputs
Signals, confidence, source visibility, and clear limits on what the system can claim.
Principles
Attribution first
Signals link back to their source whenever possible.
No narrative layer
Lab tracks signals and change, not blame or opinion.
Uncertainty is visible
Confidence and conflict are part of the output.
Same engine, new domains
Lab experiments reuse the same Killfeed signal pipeline used elsewhere on the site.
Live experiments
Information Inspection
In this experiment, we headline per-article diagnostic signals and create a Reliability score based on weighted inputs, genre calibration, and a band label. Explore the articles in the feed; we sometimes run daily crawling and scraping to add new sources.
Investment signal inspection
Browse KB articles where Has Investment Signal is checked: primary signal type, impact, scope, horizon, and model confidence—still grounded in the same Reliability banding as Information Inspection.
CrowdPulse · TSA checkpoints
Operational crowd and checkpoint stress signals—congestion, delays, and traveler-reported pressure—when official wait-time views are late or sparse. The page uses a fixed demonstration snapshot so the demo stays stable.
More experiments
Iran Conflict
A live experiment in measuring escalation using public signals such as alerts, advisories, transport disruption, and official statements.
Status: Up next
Election Signals
Exploring whether public procedural and official signals can be structured without editorial framing.
Status: Concept
Cyber Threat Signals
Experiment in converting advisories, disclosures, and operational notices into severity signals.
Status: Concept
Market Stress Signals
Exploring whether disruption indicators can be separated from narrative noise.
Status: Concept
Request a proposal
Lab experiments are one slice of what the Signal Extraction Engine can do. If you have a domain, data sources, or workflow you want to explore with a scoped engagement, use the form on the main site to share requirements and timeline.
Open proposal form