Skip to content

Beads as an Execution Graph

@doodlestein -- on treating beads as a dependency graph, not a todo list.

Key Insight

Beads are not a flat task list. They're a dependency-tracked execution graph where:

  • Epics group related work
  • Tasks are individual units with clear acceptance criteria
  • Subtasks break tasks into atomic steps
  • Dependencies define what must complete before something else can start (blocks/related/parent-child/discovered-from)

Each bead must be self-contained and self-documenting -- an agent picking it up should have enough context to work without re-reading the entire plan.

What Good Beads Include

  • Relevant background and reasoning
  • Constraints the implementation must respect
  • Acceptance criteria (how to know when it's done)
  • Dependency links to related beads

The QA Loop

After converting the plan to beads, run repeated QA passes:

  • Are dependencies correct?
  • Is any important work missing?
  • Are beads too large or too small?
  • Are tests comprehensive with detailed logging?
  • Are there beads for security, performance, and ops concerns?

Stop when changes are mostly reordering/renaming, not missing work.

Source

View on X