← Back to all fieldnotes
Paperclip · Medium confidence / demo-stage

Paperclip: Orchestration Without Agent Theatre

What the demo shows

Paperclip is a demo-stage meta harness. Existing agents keep their own runtimes while working in a company-like hierarchy of objectives, managers, agents, tasks, reviews, routines, artifacts, and decisions. The observed pattern is explicit work instead of unbounded free-form agent chatter.

Controls worth studying

Task hand-offs, review points, per-agent secret grants, approval gates, and an artifact trail make a multi-agent system easier to inspect. A human provides direction and approval; agents produce work that can be reviewed rather than relying on an invisible conversation.

What remains open

This is research, not a claim of production readiness. Availability, authorization detail, failure handling, scaling beyond the observed demo, and CEO-agent hallucination risk are open questions. Local-first operation also needs verification.

How the pieces connect

Explore the original interactive diagram. Its controls include guided views, search, trace, and theme switching. On a narrow screen, use the diagram controls or open it full size.

Open diagram full size ↗

More notes will be added as this work develops.