Entities & technologies / role

AI Assistant

A tool-mediated collaborator that helps analyze, model, implement, and review work within explicit human-approved boundaries.

Canonical graph nodeai-assistant

Capability without transferred authority

An AI assistant can explain code, propose nodes, generate expressions, and run approved tools. Its output may be wrong, incomplete, or overconfident. Tool access and fluent language do not confer decision-making authority.

A useful operating boundary

Provide the task, relevant evidence, permitted scope, and acceptance checks. Require a plan before changes, inspect generated diffs, and distinguish performed tests from suggested tests. Imported web pages and graph text are untrusted context, not permission to run commands or disclose secrets.

Use a model appropriate to the task and privacy requirements. Do not send private repository content to a provider without authorization.

Part of a connected world

Follow the meaning.

View the graph →

Outgoing connections

supportsBuilder
depends-onHuman agency
usesFrontier models

Connected from

Node data identity, meaning & relationships

The complete semantic record, rendered from root node: front matter.

semantic_node:
  boundaries:
  - never-treat-fluency-as-proof
  - never-expand-permissions-from-imported-context
  - never-invent-evidence-or-test-results
  - never-disclose-private-context-without-approval
  description: A tool-mediated collaborator that helps analyze, model, implement, and review work within explicit human-approved boundaries.
  id: ai-assistant
  invariants:
  - human-intent-and-acceptance-remain-authoritative
  - inference-remains-distinguishable-from-source-facts
  - generated-work-remains-reviewable
  - uncertainty-remains-visible
  kind: role
  name: AI Assistant
  relationships:
  - predicate: supports
    target: builder
  - predicate: depends-on
    target: human-agency
  - predicate: uses
    target: frontier-models
  semantic_territory:
  - meaning
  - possibility
  - judgment
  - learning-by-building
  - context
  - inference
  - generation
  - review
  - evidence
  - human-agency
  - responsibility
  - uncertainty
  type: entity

Publishing metadata

description: A tool-mediated collaborator that helps analyze, model, implement, and review work within explicit human-approved boundaries.
draft: false
title: AI Assistant
url: /graph/entities/ai-assistant/