> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mindset.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Roles

> Humans are a first-class primitive in Memex. This is how different people (and different agents) use the graph.

## Role-Based Interaction

[Humans are a first-class primitive](/memex/concepts/humans) in Memex, not "whoever happens to be logged in". Different people bring different expertise, and Memex learns who knows what. Below is how different roles (human and AI) tend to use the graph, but the same person moves fluidly between them in a given week.

<CardGroup cols={2}>
  <Card title="Product & Leadership" icon="compass" iconType="solid">
    **Primary layer:** Strategies + Decisions

    * Create strategies that articulate the objective, the *why*, not just what needs building
    * Resolve open decisions that block work
    * See the impact of re-prioritisation: *"If we reverse `D4`, which work items are affected and which blueprints become stale?"*
    * Track progress through strategy completion and decision resolution rate, not story points
    * Promote work items to strategies when scope expands, keeping the graph honest
  </Card>

  <Card title="Engineering" icon="wrench" iconType="solid">
    **Primary layer:** Work Items + Blueprints

    * See what's ready to work on (all decisions resolved, all dependencies met)
    * Load relevant blueprints before starting work
    * Create execution plans that reconcile design with reality
    * Flag drift when blueprints don't match the codebase
  </Card>

  <Card title="AI Coding Agents" icon="robot" iconType="solid">
    **Primary layer:** Work Items + Blueprints + Decisions

    * Load the goal, constraints, and institutional knowledge before writing code
    * Check for unresolved blocking decisions and stop rather than guess
    * Submit execution plans for review before implementation
    * Report blueprint drift when detected during planning
    * Update work item status as implementation progresses
  </Card>

  <Card title="AI Research Agents" icon="magnifying-glass" iconType="solid">
    **Primary layer:** Decisions

    * Investigate open decisions that need data to resolve
    * Gather competitive analysis, technical benchmarks, or user research
    * Submit findings as evidence attached to specific decisions
  </Card>

  <Card title="AI Testing Agents" icon="shield-halved" iconType="solid">
    **Primary layer:** Work Items + Blueprints

    * Read acceptance criteria from completed work items
    * Load testing blueprints (conventions, frameworks, helpers)
    * Verify implementations against the stated goals
    * Flag when test results contradict blueprint documentation
  </Card>
</CardGroup>