Current Questions / Public Question Graph

Publish the question skeleton, not the private brain.

The first five cards are Zeyuan Li (Amy Li)’s current machine-readable questions. All seven preserve the original wider research graph. Every card exposes a bounded public surface: why the question matters, what is believed, what remains unknown, what could disprove it, and what another person or agent may contribute.

WQ-001 · P0 PUBLICWashington, DC

Who governs machine consensus?

Governance may concentrate in the permission, provenance, evaluation, and escalation layers around models.

Counterargument
Model providers may absorb the surrounding governance functions.
Falsifier
Independent permission layers fail to improve repeated real outcomes or trust.
Seeking
Auditable failures, appeal mechanisms, institutional deployments.
Query this question
WQ-002 · P0 PUBLICSan Francisco–Berkeley

What is the missing mandate layer of the agentic economy?

Machine action needs explicit purpose, duration, forbidden actions, approval actors, and revocation state.

Counterargument
Existing identity and authorization standards are already sufficient.
Falsifier
Purpose-bound mandates add complexity without reducing ambiguous action.
Seeking
Delegation postmortems and cross-agent authorization patterns.
Query this question
WQ-003 · P0 PUBLICLondon–Cambridge

How should personal agents exchange cognition without exposing private memory?

Agents should exchange bounded, owner-approved artifacts—never unrestricted memory access.

Counterargument
Low-context artifacts may be too shallow to outperform normal search.
Falsifier
Bounded artifacts fail to transfer novel, actionable judgment.
Seeking
Owner behavior, leakage tests, useful minimal disclosure.
Query this question
WQ-004 · P0 PUBLICNew York

When should an Agent recommend that two humans actually meet?

Human escalation should follow a useful authorized exchange, not precede it.

Counterargument
Human chemistry cannot be predicted by artifact exchange.
Falsifier
Pre-exchange routing performs no better than profile matching.
Seeking
Repeat queries, meeting conversion, and time saved.
Query this question
WQ-005 · P0 PUBLICWashington, DC

What minimum evidence is required before an AI may represent a human judgment?

Representation needs provenance, authoring mode, recency, scope, approval actor, and falsification conditions.

Counterargument
Strict evidence rules make personal agents too slow.
Falsifier
Lighter attribution produces equal trust and fewer errors.
Seeking
Provenance standards, correction flows, approval latency.
Query this question
WQ-006 · P0 PUBLICSingapore

What replaces the social graph when AI can route knowledge directly?

The next graph may connect questions, artifacts, provenance, and bounded access rights.

Counterargument
Existing platforms can add these features and own the layer.
Falsifier
Question routing does not outperform profile discovery.
Seeking
Maintenance behavior, portability, repeat exchange.
Query this question
WQ-007 · P0 PUBLICCross-city

How do cities become physical nodes in an agent-native cognitive network?

A node forms when recurring questions, authorized exchanges, field experiments, and steward responsibility compound locally.

Counterargument
Online networks make city-based structure unnecessary.
Falsifier
Physical nodes add no continuity, trust, or novel transfer.
Seeking
Recurring salons, second questions, steward workload.
Query this question

Machine-readable source

Every complete card is available through the public API.

GET /api/public/questions
Includes current belief, known facts, unknowns, changed-mind note, evidence sought, artifacts, permission, provenance, review date, related city, and human trigger.