spiderpilot89/ocean-failure-boundaries-n2500

Ocean failure boundaries (n=2500)

First-principles Monte Carlo at the clocks where marine assumptions die: slam, thermocline, crush, dead-reckon.

Each config is 2,500 trajectories. Unique SHA-256 proof_hash per row. Dual-regime flags (warn / hard) are booleans. One row = one trajectory.

Custom OEM displacement, GM, crush depth, battery, and ping Hz recompile the same binaries: 48-Hour Forge.

Hull

Config	Warn / hard
hull_slam	foil lag 16.2% / submarined 40.4% · held 55.4%
boat_roll	speed-limited 71.6% / capsized 17.0% · held 11.4%
diesel	overtemp 49.7% / melted 4.6% · overtemp-held 45.2%
wake	in-band 23.0% / snapped 14.8% · both 10.0%

Column

Config	Warn / hard
thermocline	miss first ping 81.5% / collision 55.5% · late detect 26.0%
cavitation	cavitating 76.0% / deafened 46.8% · cav-quiet 29.3% · flow-mask 0%
stiction	seal held 70.1% / tumble 15.6% · clean 14.2% (exclusive)
crush	power-starved 34.0% / hull crushed 33.9% · live 32.1% (exclusive)

Reckon

Config	Warn / hard
dead_reckoning	lock-loss 38.1% / current-dominated 52.5% · both 19.5%
sea_clutter	CFAR buried 16.4% / impact 36.7% · late track 20.3% · held 63.3%

Load

from datasets import load_dataset

slam = load_dataset("spiderpilot89/ocean-failure-boundaries-n2500", "hull_slam", split="train")
print(len(slam), slam[0]["is_submarined"], slam[0]["is_foil_phase_lagged"])

Engine

• Forge: johnkruze/genesis-core
• Reflex: johnkruze/ztp-runtime
• Site: zerotrustphysics.com/ocean
• Other sealed banks: doe-genesis-sealed-n2500 · humanoid-failure-boundaries-n2500

License

MIT.

@dataset{kruze_ocean_failure_boundaries_n2500_2026,
  author = {Kruze, John},
  title = {Ocean failure boundaries (n=2500 dual-regime)},
  year = {2026},
  url = {https://huggingface.co/datasets/spiderpilot89/ocean-failure-boundaries-n2500}
}