Radar Needs an Actuator

Display alone is cosmetic

A polished chart of duplicate clusters, orphans, and drift does not reduce the workflow if Dad must locate and repair every file manually. The dashboard lowers rediscovery cost while disposition and proof remain scattered.

Dad's correction was simple: without a path to repair, the lens is cosmetic. The detector still cannot auto-merge. An actuator with a human plan gate belongs between observation and mutation.

The candidate snapshot becomes the work order

Queueing from Radar freezes current counts and candidate hashes into a brief. The session reads exactly that set and proposes verdicts by candidate type.

Clusters use merge-to-canonical, keep-all, or archive-one; orphans use link-in, retier, or keep-as-is; drift uses re-point, retrace, retire-line, or add-line. One generic Fix button erases judgment.

Lens exclusions are product decisions

Record and library tiers contain deliberate repetition and historical residue. A live measurement found forty-five clusters with only one actionable candidate; excluding those tiers made the default lens useful.

The exclusion narrows purpose to living-memory curation rather than hiding data. Wider scans can remain explicit, but a noisy default turns an actuator into busywork.

Before and after prove completion

A landing report should lead with the rerun scoreboard: clusters three to zero, orphans two to one with one keep-as-is, drift four to zero.

Zero is not the only success. An approved keep verdict can retain a fingerprint so the lens respects it. Quiet means every signal is explained, not every trace deleted.

An observation surface becomes a work tool only when approved judgment reaches recorded reconciliation. Use candidate snapshot, typed verdicts, and before/after proof without giving the detector authority.

Code

Build a before and after scoreboard·python

before = {"clusters": 3, "orphans": 2, "drift": 4}
after = {"clusters": 0, "orphans": 1, "drift": 0}
dispositions = {"orphans.keep-as-is": 1}
scoreboard = {
    key: f"{before[key]} -> {after[key]}"
    for key in before
}
assert scoreboard["clusters"] == "3 -> 0"
assert dispositions["orphans.keep-as-is"] == 1

Display alone is cosmetic

The candidate snapshot becomes the work order

Lens exclusions are product decisions

Before and after prove completion

Code

External links

Exercise

Progress

Comments 0