dira decision memory for AI coding agents
$dira whycst-0002
int-0003Replace the tool pile — kazi + dira + a subset of skills is the whole stackactive 2026-07-30
cst-0002The entry set is closed at five kindsactive 2026-07-29
int-0003Replace the tool pile — kazi + dira + a subset of skills is the whole stack
cst-0002The entry set is closed at five kinds
cst-0002 ◆ active 2026-07-29

Arising from int-0003 — Replace the tool pile — kazi + dira + a subset of skills is the whole stack

The entry set is closed at five kinds

intent, decision, question, constraint, note. That is the whole vocabulary, at every tier.

The set is closed because each plausible addition is a specific tool dira has promised not to become:

- a **task** kind → dira becomes a tracker, and kazi already owns execution - an **event** kind → dira becomes a calendar - a **journal** kind → dira becomes Obsidian, the tool being deleted - a **person** kind → dira becomes a CRM - a **metric** kind → dira becomes a dashboard

The strongest evidence the five are sufficient: the personal and venture tiers were added to the design without needing a sixth (dec-0006). A tier-spanning model that still fits in five kinds is a model worth defending.

`note` is the pressure valve — the place a thought goes when it is not yet any of the other four. Notes surface once in the next brief and then decay, so the valve cannot silently become a sixth store. A note that matters gets promoted into an intent, a decision, or a question.

Adding a kind requires superseding this constraint.

Alternatives on record — 0

no alternatives recorded — a constraint states a rule rather than choosing between options

Every decision in this repo is recorded like this one.

Keep your own →