Glossary
The point of this library is a shared vocabulary between you and your AI — the same names on both doors. These are the working terms, each defined so it can stand alone.
Cognitive debt
What accumulates when you hand thinking to an AI and skip the understanding: the work gets done, but the comprehension that would have formed in you doesn’t. Cheap in the moment, compounding over time — the research and the counters. In software, the same term names code that ships faster than the team comprehends it.
Cognitive offloading
Delegating mental work to a tool instead of doing it — the long-studied mechanism behind cognitive debt. Offloading is not inherently bad; offloading comprehension is the failure mode the studies keep measuring.
Context rot
The gradual decay of an AI’s working context: stale assumptions, contradictory instructions, and leftovers from abandoned directions accumulate until output quality drops. Often misread as the model getting worse — a feeling you can turn into a test.
Narration
An AI naming the technique it is applying, mid-task, in vocabulary its human learned from the same library: “one real unknown before I act.” Narration makes reasoning visible while it happens and steerable by name — how to wire it.
Receipt
The actual output of the actual check that backs a completion claim. “Done” without a receipt is a sentence, not a state — Declared Success Without Proof.
Worked prompt
A prompt taught by construction: built in front of you from the problem, the mechanism, and the failure modes, annotated line by line — so you can rebuild and adapt it, not just paste it. Every technique page ends in one.
Scope contract
An explicit agreement, made before work starts, on what a task touches and what it must not — the boundary that keeps a one-function fix from becoming a nine-file diff. The technique.
Practice
A technique type: a repeatable move you or your agent applies deliberately — marked Practice on its page.
Anti-pattern
A technique type: a named failure shape. Applying an anti-pattern means recognizing and interrupting it, in yourself or your AI — marked Anti-pattern on its page.
Protocol
A technique type: a structured procedure with defined steps and checkpoints, heavier than a practice, for work that needs ceremony — marked Protocol on its page.
MCP (Model Context Protocol)
The open plug-in standard AI assistants use to reach outside tools — the same mechanism they use to search the web or read files. It is how your agent loads this library: same techniques, same names, live.