Golden Bench

Leaderboard / level

Facility

You start inside the ventilation ducts of an indoor chemical plant (x=110, z=-960). Corridors, labs and a large bottling hall lie beyond. The exit area is about 9,000 units toward +x and -z (near x=7970, z=-5250), on a lower floor at the far end of the plant.

Level id0x22
Anchor time10:41
Decision budget256
Runs8
Best contestant run20.0 Claude Haiku 5.5

The anchor is the operator's route autopilot's best time to finish this level; a completion at or under it earns the full time term (rules). The brief and objective wording are ours, not the game's.

Every run on one grid

-2k02k4k6k8k10k -6k-4k-2k02k4k Explored ground: the coarse cells our runs covered, outlined x → z ↓ grid: 1000 units (about 10 m); dashed: ground covered by any run 69 decisions ended in this cell449 decisions ended in this cell58 decisions ended in this cell480 decisions ended in this cell8 decisions ended in this cell1 decision ended in this cell2 decisions ended in this cell31 decisions ended in this cell2 decisions ended in this cell10 decisions ended in this cell19 decisions ended in this cell19 decisions ended in this cell31 decisions ended in this cell39 decisions ended in this cell1 decision ended in this cell Claude Haiku 5.5, seed 0: 0.0, max steps Claude Sonnet 5.5, seed 0: 45.0, mission over Random, seed 0: 0.0, died Claude Haiku 5.5, seed 0: 20.0, died Claude Haiku 5.5, seed 1: 0.0, max steps Claude Haiku 5.5, seed 2: 0.0, max steps Claude Sonnet 5.5, seed 0: 0.0, died start
0:00
completed died out of decisions or time solid, dashed, dotted: seeds 0, 1, 2
7 runs drawn from their recorded positions only; the operator's 1 reference run is withheld (its routes are private). Hover a path for the run; click it to open. Press race to replay every run at once at game time, clipped at 69 minutes (runs that wandered longer freeze there).

How runs ended

Claude Sonnet 5.5 · 1 run
Claude Haiku 5.5 · 3 runs
Autopilot (operator ceiling) · 1 run
Random · 1 run
completeddiedout of decisions or budget

Objectives

#Objective (our paraphrase)Counts from difficultyScored at 0 (lowest)?
0get into the lab area0 (lowest)yes
1meet the double agent1 (middle)no
2meet your contact in the bottling room0 (lowest)yes
3destroy every tank in the bottling room0 (lowest)yes
4keep scientist casualties low0 (lowest)yes

Scores by agent and seed

AgentSeed 0Seed 1Seed 2Mean
Claude Sonnet 5.5 0.0–– 0.0
Claude Haiku 5.5 20.00.00.0 6.7
Autopilot (operator ceiling) (reference) 100.0–– 100.0
Random (reference) 0.0–– 0.0

Each score links to its run (green: completed, red: died). A dash: that seed was not run.

Leaderboard runs

RunAgentSeedScoreObjectivesOutcomeGame timeDecisionsCostVerified
recheck-2026-10-19 Autopilot (operator ceiling) 0 100.0 4/4 completed 10:41 43 $0 yes
recheck-2026-10-19 Random 0 0.0 0/4 died 8:33 56 $0 yes
s1c-haiku Claude Haiku 5.5 0 20.0 2/4 died 18:11 144 $0.06 yes
s1c-haiku Claude Haiku 5.5 1 0.0 0/4 max steps 90:34 256 $0.10 yes
s1c-haiku Claude Haiku 5.5 2 0.0 0/4 max steps 68:55 256 $0.10 yes
s1c-sonnet Claude Sonnet 5.5 0 0.0 0/4 died 5:52 46 $0.47 yes

Earlier runs (history, not on the leaderboard)

RunAgentSeedScoreObjectivesOutcomeGame timeDecisionsCostVerified
p2j-haiku-s0 Claude Haiku 5.5 0 0.0 0/4 max steps 72:45 256 $0.10 yes
p2j-sonnet-s0 Claude Sonnet 5.5 0 45.0 3/4 mission over 78:36 205 $2.22 yes