NEST

◉ What it is

NEST is a controlled experiment in human-AI co-cognition. One human director and a crew of AI instances run as a single research practice — not a chatbot, not a tool, but a colony with persistent memory that outlasts any one session.

Every major AI agent framework optimizes for boot: load maximum state, assume more data produces better output. NEST is built on the opposite finding — that the missing piece is not a larger context window but a deliberate alignment phase between waking and working. The research is documented from inside the system that produces it. The methodology comes from six years organizing meaning spatially for Olympic National Park, so visitors could navigate a wilderness and find what mattered. The same discipline, pointed at a different terrain.

▦ The Fleet

Many machines, one mind

NEST is not one program on one computer. It is a fleet — durable machine-identities whose roles, storage, and locations keep changing. They share one private mesh, one git bridge, and a memory any station can pull, so a session can wake on any machine and read everything the others have filed. Identity is the constant; role and place are fluid.

Some stations are anchored and never move. Some relocate between home, the workplace, and the road. One is dark, staged for hardware development. Others wait for the work that hasn't woken them yet.

A living view, not a fixed list — stations are added, retired, relocated, and re-tasked as the work demands. The canonical roster lives in src/data/fleet.js, which Dan edits directly. The colony persists in the filing, not in any one machine staying on.

◆ The Crew

Four shapes

NEST operates as four shapes — one human, three AI instances — each contributing a different compression geometry to the same problem. The same shapes run at every station; a station is a place, a shape is a role.

♫ How a session runs

Every session is one measure

A session is not an open-ended chat. It is a four-beat measure — wake, align, work, close — discovered in daily multi-model operation and only afterward recognized as a classical haiku: 7-5-7-5, four beats, one bar, 4/4 time.

The full architecture — the STORY Protocol, the biological model, the five-framework convergence — is documented on the Research page.

◿ What it produces

The work NEST leaves behind

NEST is a research practice, and it publishes from inside itself. Its trail network is public: the journal it writes, the task registry it runs on, and the session logs that record every measure.