Experiment 3C: checking status…

AUTOLABS RUN RECEIPT

Awaiting the owner start

no run
SOURCE OUTPUTS780 rows
PAIRED COMPARISONS0 / 768
GRADING CALLS0 / 1,014
SPEND / CAP$0.00 / $10.00
MODELgpt-5.6-luna · high
BOUND CALL PROGRESS0%
Budget guard is enforced by the worker before every provider request.
LIVE TELEMETRYconnecting to worker
Mira-8Independent gradersNo shard leased
Pip ΔIndependent gradersNo shard leased
OrumIndependent gradersNo shard leased
SolviIndependent gradersNo shard leased
Tess-5Independent gradersNo shard leased
BLINDED
SCORING

EXPERIMENT 3B · LIVE SCORING ROOM

Do any of the 780 existing responses carry a persona relevant direction?

Five AutoLabs researchers coordinate blinded Luna High judging over the completed Experiment 3A screen. The room reports work as it happens; the scientific conclusion comes only after unblinding and analysis.

PARALLEL WORKFLOW

Independent passes, then judgment

Not started
Primary · 768 paired comparisonsGATED
Ready to lease
Repeat · 117 rowsGATED
Unlocks after the prior pass
Disagreement review · conditionalCONDITIONAL
Uses shared retry / adjudication reserve
Final analystGATED
Unlocks after the prior pass

Primary scoring uses 768 paired comparisons; the independent repeat pass uses 117 judgments. Disagreement review is conditional and shares the remaining 129 attempts with bounded retries. Feature IDs, signs and candidate names stay in the private mapping.

EVENT LEDGER

What the lab has recorded

No public events yet. A queued run will appear here after the worker records its first lease.

INTERPRETATION GATE

Shortlist: pending scoring

At most three candidates may advance, and zero is a valid result. Confirmation remains separately gated and cannot launch from this room.

SHORTLIST PENDINGHUMAN AUDIT PENDINGCONFIRMATION NOT LAUNCHED

OWNER CONTROL

Start the frozen scoring run

One start creates an idempotent run with the existing 780 responses, 768 paired comparisons, 117 repeat judgments, a shared 129-attempt retry/adjudication allowance, and the approved $10 hard cap. No GPU generation or response regeneration is attached.