Skip to main content

Curation & Review

Curation keeps capture useful by limiting what stays visible, what gets promoted, and what quietly leaves the active surface.

This page governs the downstream handling of capture items after ingestion.

The purpose of review is not to preserve everything. It is to keep the system usable, selective, and aligned with future reuse.

Core rules​

Curation should follow a few simple conditions rather than a large workflow system.

1. Review should reduce visible clutter, not preserve everything​

The active surface should stay small enough to remain governable.

If review keeps expanding the visible backlog, it is failing.

2. Human judgment is final for promotion and deletion-sensitive actions​

The system may prepare, suggest, and compress.

The human remains the final selector, judge, and operator for actions that materially change preservation or meaning.

3. The system may help with first-pass housekeeping​

The system may:

  • suggest merges
  • archive stale ephemeral items
  • prepare promotion candidates
  • explain why an item is being surfaced

This allows AI to do the heavy lifting without taking over the final judgment layer.

4. When uncertain, keep the lighter state​

When the system is not confident that an item deserves durable preservation, prefer the lighter outcome.

That usually means keeping it local, keeping it working, or letting it age out rather than prematurely promoting it.

Review queues​

The review surface should stay compact.

Use only three queues:

Review now​

Items that need a human glance because the next move is not obvious enough to automate safely.

Merge candidates​

Items that likely duplicate, overlap, or combine into a clearer reusable unit.

Promotion candidates​

Items that likely deserve durable preservation as artifacts.

Everything else should stay out of the main review surface.

This is important. Too many queues recreate clutter as categories.

Auto-housekeeping rules​

Keep auto-housekeeping simple.

Ephemeral aging​

Ephemeral items that remain untouched for a defined period may archive automatically.

This keeps the system from accumulating local debris.

Merge assistance​

The system may collapse likely duplicates only when confidence is high and reversal is easy.

If merge confidence is not high, surface the candidate instead of forcing it.

Promotion restraint​

Promotion should require a clear artifact type and likely reuse value.

If reuse value is unclear, keep the item lighter.

Promotion gate​

Promote only when all of the following are true:

  • the artifact type is clear
  • the title is clear
  • usefulness beyond the current thread is plausible
  • recreating later would cost more than preserving now

This keeps promotion selective and avoids turning every decent note into a durable object.

Merge rule​

Merge when two or more capture items represent the same reusable unit more clearly together than apart.

That is enough for the framework.

The implementation may later distinguish duplicate, fragment, or successor patterns, but the governing rule should stay simple.

Human and system roles​

The system is good at:​

  • first-pass classification
  • surfacing likely merge candidates
  • identifying stale ephemeral items
  • proposing likely artifact destinations
  • preparing compact review explanations

The human is good at:​

  • deciding what is actually worth keeping
  • validating or rejecting promotion
  • resolving ambiguous merges
  • editing meaning where needed
  • protecting the system from preserving noise

This division keeps the system useful without pretending that automatic curation should replace judgment.

Minimal review outcome set​

Every review event should tend toward one of a few outcomes:

  • keep active
  • merge
  • promote
  • archive
  • discard

This is enough to maintain flow without introducing many intermediate states.

Practical guidance​

Prefer fewer visible items​

The review surface should feel like a garden, not a warehouse.

Prefer bounded review​

Each surfaced item should have a visible reason for review.

Prefer reversible automation​

Automatic archiving and high-confidence merge assistance are safer than aggressive autopromotion.

Prefer clear artifacts over many artifacts​

If several fragments belong to one reusable unit, merge first and promote second.

What curation is for​

Curation is not a secondary note-taking layer.

It is the housekeeping logic that keeps capture from flooding the system.

Without curation, cheap capture becomes backlog bloat.

With curation, capture remains cheap while preservation stays selective.

Minimal downstream promise​

A well-governed capture system should make the following true:

  • useful material becomes easier to reuse
  • unclear material remains local until it clarifies
  • low-value material leaves the active surface quietly
  • the human is not forced to browse an ever-growing pile of cards

That is the downstream goal.