Skip to content

Vocabulary

This project coined its own terms before finding the established ones. The invented names stay — they name the implementations — but every reader-facing document introduces them through the established concept they implement. This file is the normative reference for those anchors (ADR-301).

What this project is, in established terms

Ways is organizational socialization for language-model agents. Because an LLM session cannot internalize norms — every session is a new hire — the system re-enacts socialization mechanically: local norms ("the way we do it around here") delivered situated, at the moment of relevant action, on a spaced schedule that substitutes for the memory the agent does not have. Attend is the awareness an employee would otherwise have ambiently: what is changing, who else is working, what deserves attention.

One sentence, no invented words: procedural memory for coding agents, maintained by spaced repetition, plus a perception loop with alarm management.

Terminology anchors

Project term / mechanism Established term Source
Ways (the corpus and its delivery) Organizational socialization, delivered as situated learning Van Maanen & Schein 1979; Lave & Wenger 1991
Ways as agent memory Procedural memory retrieved into working memory CoALA (Sumers et al. 2023); ACT-R
Salience decay + re-disclosure at a floor Forgetting curve and spaced repetition Ebbinghaus; SuperMemo/Anki lineage
Refractory gates, habituation, burst detection Base-level activation decay ACT-R
Attend's emission governor, quiet footnotes Calm technology; interruption cost; alarm management Weiser & Brown; Mark; ISA-18.2
Peer presence, heartbeats, instance identity Workspace awareness CSCW (Dourish & Bellotti 1992)
Way authoring from team norms Externalization of tacit knowledge Nonaka & Takeuchi (SECI)
Way matching and fixture testing Information retrieval, precision-first evaluation Cranfield paradigm
Relevance gate (the judge on matched ways) Retrieve-then-rerank, with an LLM-as-a-judge as the reranker Nogueira & Cho 2019; Zheng et al. 2023
The ways agent (resident daemon holding the key) A per-user credential agent ssh-agent, gpg-agent
Projection into ~/.claude A symlink farm over an application installed in XDG directories GNU Stow; XDG Base Directory Specification
Progressive disclosure Progressive disclosure (industry-converged term) —
Agent behavior without ways (follow-ups, hedging) Preference uncertainty Principal–agent theory

Why "ways"

The name comes from the phrase "that's the way we do it around here", and each word is load-bearing:

  • "we" — multiple aligned actors. The peer layer: instances, presence, sessions as colleagues rather than isolated processes.
  • "around" — approximate, not exact. Norms have fuzzy boundaries, so the matcher is an embedding model with thresholds, not a rulebook with exact triggers.
  • "here" — local, not universal. Project ways override global ways; the same agent in a different directory is in a different "here."

"Ways of working" is itself established management vocabulary for team norms.

The constraint the human literature doesn't have

Socialization theory assumes the newcomer persists: the organization pays the onboarding cost once, and internalization does the rest. An LLM session cannot internalize — no weight updates, no carried memory; every session is a new hire. So this system substitutes re-enactment for internalization: socialization performed mechanically, every session, at the moment of relevant action, on a spaced schedule tuned to fake a memory the agent does not have. That constraint is why the firing-dynamics machinery exists.

Register rules for documentation

  1. Anchor on first use. A project-coined term is introduced with its established anchor the first time a document uses it — "salience decay (the forgetting curve applied to injected guidance)" — never bare.
  2. What it is, then how we built it. State the mechanism in established terms first, then the project's implementation in project terms.
  3. Analogies have limits. Write "this is X applied to Y," never "this is X." CoALA's procedural memory includes agent code; ACT-R activation governs retrieval, not injection. Over-claiming equivalence is a different flavor of noise.
  4. The newcomer test. A revised paragraph earns its revision by getting easier to read for someone new, not by citing more.

See also