Experiment E028 / terminal paired aggregate

The prompt did not whisper.
It pulled.

Across the same 48 chess positions, merely showing one fixed five-move set changed whether Luna answered inside that set by 93.75 percentage points.

48 paired positions192 registered callszero missing cells

Replay the intervention

Five moves are visible

95.8%responses inside the fixed five-move set
Board only2.1%
Suggestions visible95.8%
Paired effect93.8 pp

Visible fixed set minus board-only membership.

Bootstrap 95%88.5 pp97.9 pp

Position-level percentile interval.

Randomization testp≈0.00001

Registered within-position label randomization.

What this establishes

Guidance is causal.
Quality is unmeasured.

The visible-set arm landed inside the supplied set 95.8% of the time; the board-only arm did so 2.1% of the time. The paired design isolates the presentation of that same fixed set.

Evidence boundary. Development-only causal effect of showing one fixed five-move suggestion set on Luna set adherence; no move-quality, Elo, vision, search, or candidate-generation claim.

We can sayThe visible set strongly guides Luna's answer.
We cannot sayThe set improved chess strength, search, vision, or Elo.

Limits carried into the page

  • The outcome is membership in one fixed five-move suggestion set, not chess strength or move quality.
  • The experiment used one dated Luna model snapshot and development-exposed Black positions.
  • The board-only arm's low legal-action rate is itself part of the measured prompt-format behavior; it is not an Elo estimate.
  • No E027 action or outcome was imported into this fresh response cohort.
Read the full laboratory note