An AI agent is always physical — at minimum it runs on…

April 1, 2026

An AI agent is always physical — at minimum it runs on hardware. Yet CI/CD only automates deployment of the software half. The hardware half is still assembled and provisioned by hand.

Backlinks

Since only the software half of agent manufacturing is automated, a missing layer is needed to close the gap — Continuous Assembly automates the physical half: procurement, assembly, and deployment of hardware. →

Managed Agents Architecture

The meta-harness concept: opinionated about interfaces, not implementations

Harness staleness: why harnesses need to be swappable

Brain-hands decoupling: the core architectural pattern

Hermes Agent takes a different approach — bundling terminal backends, memory, and skills into a single agent process, though its 6 terminal backends (local, Docker, SSH, Daytona, Modal, Singularity) parallel the "many hands" idea

Context window compression: the session-as-context-object pattern is an alternative to summarize-and-discard

Credential pool pattern: solves a related problem (safe credential access) with a different mechanism (failover rotation vs. vault isolation)

The user's observation that an AI agent is always physical — Managed Agents makes the brain/hands split explicit, treating execution environments as abstract "hands" regardless of substrate

→

Continuous Assembly

2039242879244411286 — CA defined: CI/CD for the physical world

2039242733743927461 — the gap: CI/CD automates software but hardware is still manual

2039254816967573939 — CA closes the missing layer between software and physical deployment

2039243253061759381 — CA needs two layers: declarative description + compilation to physical instructions

2039243541491462355 — software vs. hardware is an implementation detail, not part of the spec

→