SWARM is a loop-based protocol for nonlinear creativity and signal detection.
It preserves human judgment in environments increasingly shaped by AI systems.
SWARM structures how ideas are spotted, weighed, arranged, refined, and made through repeated loops rather than linear execution.
It is not a productivity system.
Learn more: https://swarmloop.xyz
Modern systems increasingly optimize outcomes for us.
SWARM exists to preserve the human ability to notice, evaluate, and interpret what matters.
Spot → Weigh → Arrange → Refine → Make
A loop is a single exploratory pass.
Work happens inside the loop.
Understanding emerges across loops.
-
Loop— a single exploratory pass. -
Protocol— a pattern that emerges across loops. -
Signal— what begins to persist across loops. -
Loop = movement
-
Protocol = memory
-
Signal = meaning
AI may assist inside the loop by surfacing observations, organizing material, or generating artifacts.
AI does not determine meaning.
Progression through the loop requires practitioner input.
SWARM is designed to be used alongside an AI system.
Open this repository in your AI environment of choice:
- Codex
- Claude
- ChatGPT
- Cursor
- or another AI tool that can work with repository context
Then give the system something real to work on.
Examples:
- “Use SWARM on this problem.”
- “Help me think through this using SWARM.”
- “Use this repository for this task.”
- “Explore this idea using SWARM.”
You do not need to choose a practice in advance.
If a relevant practice exists, the system may suggest it as an entry point into the loop. Otherwise, the system should run a standard SWARM loop.
The loop is:
Spot → Weigh → Arrange → Refine → Make
The AI may assist within each phase, but progression between phases requires practitioner input.
SWARM is not designed to produce immediate final answers.
Clarity emerges through repeated loops, not a single prompt.
Human judgment remains active throughout the process.
canon/— Stable concepts and ontology.protocol/— How SWARM operates.patterns/— Recurring behaviors and diagnostic observations.practice/— Repeatable loop entry points and applications.templates/— Authoring scaffolds for extending the protocol.
- Read
canon/swarm.md,canon/loop.md,canon/protocol.md, andcanon/signal.mdfor the core concepts behind SWARM. - Use
practice/README.mdwhen you want a repeatable entry point such as Loop Logging or a System Check. - Use
patterns/README.mdwhen you want help recognizing what is happening during exploration.
- Do not treat SWARM like a one-shot prompt.
- Do not let AI decide meaning.
- Do not skip practitioner input between loop phases.
SWARM is evolving.
The canon remains stable.
Patterns and practices expand through use.
SWARM was developed by Matt Lambert.
- swarmloop.xyz — SWARM online
- Substack — ongoing writing and exploration
- cardeo.ca — broader creative and systems work