adoption

The confirmed use of a suggested technique. OpenTacit records an adoption automatically when your actions match the recipe. An offer alone never counts as adoption.

area of practice

A cluster of related techniques on the playbook map: techniques that share tags, or techniques that the same cohorts adopt together. The map uses areas to divide the playbook into groups with names.

technique

A validated method of working with AI that someone recorded for others to use. A technique contains a name, description, recipe, boundaries, tags, and measured outcomes. Together, an organization's techniques are its playbook.

cohort

An aggregate dimension for outcome reports, in the form dimension:value, for example team:payments or role:engineer. OpenTacit uses cohorts to show "who" without data about individuals.

decay

The flag on a technique whose recent helped rate fell below its own baseline. Decayed techniques do not continue to serve on old evidence. They go to the queue on the Review page for revalidation. The flag shows as a "Decaying" moment in the Events feed.

draft

A technique that waits for review. Drafts never reach members through retrieval. A reviewer promotes, edits, or rejects them.

Events

The activity feed under Outcomes: a record of OpenTacit's activity, with the newest items first. It shows delivery moments (the funnel, one interaction at a time) and lifecycle events (changes that the playbook makes to itself). The feed shows cohort-level data only, like every OpenTacit view.

evidence

A technique's measured outcomes. OpenTacit always shows them exactly as measured ("helped 94% · n=120") or as "awaiting measured outcomes." Evidence comes only from feedback. OpenTacit never asserts evidence.

federation

The protocol that shares techniques between organizations. Organizations publish signed feeds of techniques by channel. Evidence crosses only as aggregate attestations.

fit check

The test before delivery. Before OpenTacit shows a retrieved technique, it checks the technique against your current work. Each suggestion includes a specific reason for the match.

funnel

The shown → adopted → helped progression. It measures whether suggestions give results, from offer to impact.

harness

The kind of surface that installs an integration on your machine: an AI coding tool (Claude Code, Codex, Gemini CLI, GitHub Copilot CLI, Cursor, Amp, pi, omp, or opencode). The tool's hooks let OpenTacit listen, and let you address OpenTacit.

helped rate

The share of a technique's adoptions that helped, as measured. It is the core quality signal that controls rank order and decay.

hook agent

The small local process that watches your session's events, asks the registry for techniques relevant to your task, and delivers suggestions. It runs on demand, on your machine, and stops when it is idle.

lifecycle event

A record of an action that the registry does on the playbook itself. The registry records a technique as discovered from usage, promoted to serve on evidence, retired, or decaying. Each event has its reason. These events, together with delivery moments, fill the Events feed.

member

A person who uses a surface with a connection to the registry: a harness on their machine, or a chat connector that they signed into. The registry identifies member machines by key for access control. The registry never joins machines to feedback.

miner

A separate service that reads sources OpenTacit cannot reach on its own and proposes techniques back to your registry. None ships with OpenTacit, and you need none to use it.

OpenTacit discovers techniques from two sources by itself. tacit init reads the conventions a repository has already written down — CLAUDE.md, AGENTS.md, .cursorrules, the Copilot instructions, and the skills, commands and rules beside them — and files each as a draft. And the registry watches for approaches that worked and were repeated: when the same one appears in at least three sessions across at least two cohorts, it becomes an observed technique. Both work on one organization's own material, which is the limit: the first sees the repositories you point it at, the second sees only what your own members did.

A miner is how an organization goes wider. It reads what OpenTacit does not — other repositories, ticket systems, incident write-ups, review comments, whatever a particular miner is built for — and posts its proposals to the registry, where they arrive in the drafts queue like any other suggestion and wait for a person. You point a registry at one by setting TACIT_SKETCH_URL, which is also what sends it sketches, and OpenTacit tells the member when that value is first set.

playbook

The organization's full validated set of techniques: its measured record of what works with AI. The playbook is the knowledge. The registry is the service that stores, measures, and serves the playbook.

promote

The review decision that changes a draft into a technique that serves.

provenance

The source of a technique: curated, contributed, suggested (by a research pass), mined, federated, or observed. An observed technique comes from repeated approaches that worked and that the organization never wrote down.

registry

The shared service that stores the playbook's techniques, measures outcomes, and serves the dashboard. Each organization has one registry.

retire / restore

To retire is to take a technique out of service without deletion. To restore is to put the technique back.

revision

A proposed change to a technique that already exists. A reviewer sees it as a draft with a diff. The original continues to serve.

scope

The property that shows whether a technique is general (good practice anywhere) or org-scoped (specific to your organization). Org-scoped techniques have an org badge.

segment

The cohort values that your machine declares (team=…,role=…). OpenTacit attaches them to feedback, so that OpenTacit can report outcomes by cohort.

shadow

A technique under evaluation for relevance, which members never see. Retrieval retrieves it, and the fit check judges it, but OpenTacit never shows it. A shadow technique auto-promotes to serve when its fit evidence passes the threshold. It auto-retires if it fits too rarely. Shadow is the machine's own review lane.

shown

The funnel's first stage: a technique that a member saw. This includes a push suggestion, and the one technique that an @tacit answer cites. A search of the playbook (a pull) is not a shown event. As a result, search traffic never increases helped rates.

signal trust

The dashboard view that compares inferred feedback with explicit feedback. This keeps automated inference calibrated.

sketch

An opt-in, scrubbed record of a situation and the adopted technique. OpenTacit uses sketches to discover techniques that the organization did not write down. A sketch never includes a transcript.

suggestion

A technique that OpenTacit offers in your session (the ◆ OpenTacit block) after retrieval and a fit check.

surface

Any AI tool through which OpenTacit reaches your work. A harness is a surface that installs hooks. It can listen, and you can address it. A chat connector (Claude or ChatGPT with the registry's MCP endpoint added) is a lighter surface. It installs nothing, and you can only address it.