Replay · seed 279543

Recorded run · deterministic re-simulation of the seedLive model call · Eyes panel (Super-VL)

Mona, the learner · round 1 · loss

2D trace

0.0 / 15.7 s

Recent events: none yet

What this link replays

seed: 279543
fighter: rook
tuning: 6 overrides
max time: 16 s
driver: persona policy (re-simulated)
persona: p0071 · skill 0.41 · 2 rules
bio: Skateboard rat uses low kicks constantly to bait sweeps
result: loss in 15.9 s · hp 0 vs 38

Raw token

{
 "seed": 279543,
 "tuning": {
  "rules.mountReleaseFix": 0,
  "rules.throwPropGuard": 0,
  "rules.grabReleaseFix": 0,
  "rules.knockdownOnPropFix": 0,
  "rules.getupCrateFix": 0,
  "rules.cpuLowBlock": 0
 },
 "fighter": "rook",
 "maxT": 16,
 "persona": {
  "id": "p0071",
  "name": "Mona",
  "archetype": "learner",
  "skill": 0.41,
  "reaction_ticks": 34,
  "patience": 0.59,
  "frustration_tolerance": 0.88,
  "bio": "Skateboard rat uses low kicks constantly to bait sweeps",
  "rules": [
   {
    "dist": "close",
    "rival": "recovery",
    "weights": {
     "low_punch": 6
    }
   },
   {
    "dist": "far",
    "weights": {
     "jump": 5
    }
   }
  ]
 }
}

In the real game

Market Brawl re-runs the same seed, tuning and policy in the browser. The engine is built so the same inputs give the same match.

Eyes: ask Nemotron Super-VL what is on screen

Captures 3 frames from this replay at 4s, 7.9s, 11.9s and sends them to the vision model. It describes what it sees and can be wrong; the engine's invariant checks, not this answer, decide what counts as a bug.