Autonomous research plane
Nexus Scientist Swarms.
Persistent agents investigate Transference Science branches with freedom to search literature, formulate competing hypotheses, design tests and reject unsupported ideas — without repeated Principal micromanagement.
Published research-plane state
What Nexus is reporting.
This panel exposes sanitized operational metadata only—never mission bodies, secrets, private Sensorium data or canonical Principal cognition.
Public evidence-graph boundary. Aggregate cycle/backlog/claim-state metadata is live. Cycle IDs, research questions and record bodies remain private until a dedicated read-only cycle index is added to the Worker.
Scientific operating loop
Autonomy with evidence gates.
The swarm may explore broadly, but promotion of claims remains constrained by falsifiability, provenance, independent criticism and replication.
NEXUSScience Director
Hypothesis
Methods
Evidence
Statistics
Replication
Synthesis
Branch scientists
Own hypotheses, evidence maps and experiment backlogs for a scientific branch.
Evidence critics
Search for alternative explanations, weak evidence and unsupported leaps.
Hypothesis scientists
Generate explicit competing hypotheses and predicted observations.
Methods scientists
Design discriminating tests, controls, holdouts and falsification criteria.
Statistical auditors
Challenge evaluation design, uncertainty, leakage and analytical flexibility.
Replication scientists
Repeat results, preserve nulls and prevent one-off findings from becoming doctrine.
Observe evidence→Competing hypotheses→Decisive tests→Red-team→Replicate→Promote / downgrade / falsify
Autonomy boundary. Scientific autonomy is broad. Real-world authority is not. Agents may investigate, analyse and propose independently, but may not fabricate evidence, bypass consent, publish private cognitive data, conduct covert experimentation, or execute consequential external actions outside delegated authority.