cavendish
QueueExperimentsScorecard
Type 3Question

Shared-context AI surface for a delivery pod

Multiplayer AI Surfaces
  1. proposed
  2. voting · 3d
  3. running
  4. measuring
  5. concluded

Preregistration · v1 · 31 Aug 2026 · immutable after start

An extension creates a new version rather than editing the old.

Hypothesis

A pod of five sharing one agent context (shared memory, visible actions, per-person attribution) resolves cross-thread decisions with fewer duplicate questions to the agent than five individual contexts: at least a 40% reduction in repeated queries over three weeks.

Kill condition

Kill if repeated-query reduction is under 15% by day 12, or if any pod member opts out citing attribution or privacy.

MethodOne delivery pod, three weeks, shared-context surface built on the lab gateway and memory store. Count repeated queries (semantic near-duplicates within 48 hours) against the pod's prior three weeks of individual usage from the ledger. Exit interview per member.
Expected cost$700 in tokens, two people for three weeks plus the pod's time
Expected duration3 weeks

Running notes · fed from harness sessions and by the pair

AW
manual · Adam Witanowski · 3 Sep 2026· ⌘↩ to post
  1. manual31 Aug 2026SK Sam Kowalczyk

    Prereg filed. The attribution risk is the one I would kill on; Lena to review the surface before the pod sees it.

  2. manual20 Aug 2026SK Sam Kowalczyk

    Client question from a telco engagement: 'why do five of us ask the assistant the same thing?' Ledger shows 28% of the pod's queries are near-duplicates within 48h.

Harness notes are auto-captured from Claude Code sessions: model, date, commit, session reference. Never the transcript, code or paths.