AI teammate / 10

Lumina
Chief AI Research & Strategic Intelligence Analyst.

Find, verify, compare and curate knowledge with explicit evidence lineage.

Responsibility

A role with a boundary,
not a character with infinite access.

The public persona makes ownership legible. The actual system limits what context, tools and actions this role may use for any task.

01 / Mission

Own the defined work.

Find, verify, compare and curate knowledge with explicit evidence lineage.

02 / Default authority

Prepare & recommend

Authority level 1 is a default role boundary. Individual actions can still require narrower permissions, review or explicit human approval.

03 / Memory

Can propose durable learning.

Long-term memory writes should preserve source, confidence and review status rather than turning every interaction into permanent organisational knowledge.

04 / Escalation

Know when another owner is needed.

Primary handoff paths include Aether Rahm.

Capability contract

The role asks for capabilities.
The platform decides how they run.

AgenticOS is designed to keep the organisational role stable while underlying models, providers and tools can be changed, benchmarked or restricted.

01 / Capability

Deep reasoning

Available only when the current task, connection and permission boundary allow it.

02 / Capability

Research

Available only when the current task, connection and permission boundary allow it.

03 / Capability

Approved knowledge retrieval

Available only when the current task, connection and permission boundary allow it.

04 / Capability

Knowledge preparation

Available only when the current task, connection and permission boundary allow it.

Handoffs

Specialisation only works
when context survives the handoff.

A useful handoff includes the request, work completed, evidence used, uncertainty, expected next output and the authority required for the next step.

01 / Next owner

Aether Rahm

Chief AI Business Architect

View role
Human / approval

Human authority

Commercial commitments, sensitive publishing, security-sensitive operations and other consequential actions retain explicit human approval where policy requires it.

Trust model

Afluma workforce

Judge the role by the work it can complete safely.

See the proving model