Glossary

The Loop - the team-scale way of working with AI that compounds: Intake → Thoughts → Workstreams → Creation, run as a cycle.

Intake - the stage where signals enter (transcripts, chats, files, GitHub issues). Skill: intake-review.

Thoughts - the stage where signal becomes structured understanding through research and synthesis. Skill: thought.

Workstreams - the stage where understanding becomes an executable spec (tracks, phases, deliverables). Skills: create, iterate.

Creation - the stage where you build from the spec and ship. Skills: execute, finalize.

Workstream lifecycle - the four workstreams_* skills together (create, iterate, execute, finalize). The loop splits them: create/iterate shape the spec (Workstreams stage); execute/finalize do the creation (Creation stage).

fai - the command you run to get the loop's skills in your AI tool.

fai serve - a standalone stateless MCP server that serves the skills without a session.

MCP prompt - how a skill surfaces in your AI tool; named fai: plus the skill path with / written as _ (e.g. fai:workstreams_create).

Track / Phase / Deliverable - the structure of a workstream: a track groups phases, a phase groups deliverables, a deliverable is one concrete unit of work.

Compounding - each pass building on the last, so context grows instead of resetting.

Honest compounding - compounding where you own and steward the loop and nothing is fabricated, only the work you actually did.