Capability Progression
Internal reference — what each tutorial unlocks and what it requires.
| Tut | Problem | Config added | Prompts | Infrastructure used |
|---|
| 01 | Context loss kills progress | [workspace], [[agent]] | one-shot | beads, session, reconciler |
| 02 | Named agents for different jobs | Multiple [[agent]] | mayor, worker | agent claim (assign to named agent) |
| 03 | Hand-feeding tasks one at a time | — | loop | claim (atomic self-claim via ready queue) |
| 04 | One agent too slow | More [[agent]] entries | — | — (just config + existing hooks) |
| 06 | Restart from scratch on multi-step work | [formulas] | — | formula parser, molecules (bead DAG) |
| 07 | Reusable formulas with specific context | gc mol create --on | — | attached molecules, Store.Update |
| 08 | Need more hands when work piles up | [[pools]], scale_check | polecat | pool manager, scale_check shell eval |
| 09 | Agents stepping on each other’s files | dir on [[agent]], GC_DIR env | scoped-worker | resolveAgentDir, MkdirAll |
| 05b | Agents die silently | [daemon], [agent.health] | — | health patrol, restart |
| 05c | Manual maintenance chores | [orders] | — | order gates, event bus |
| 05d | Multi-repo orchestration | [projects.*], scope | — | project scoping, agent replication |
| 10 | Multiple projects need isolated task tracking | [[rigs]], prefix, routes | — | InitRigBeads, deriveBeadsPrefix, routes.jsonl |
Last modified on March 19, 2026