🐍 Snake Matcher
Playground Maths Ops Boost API

Snake Matcher v2

On-demand SAT inference. 17 factories, 506MB on disk, $0.03/hr spot. Models load in 60ms, predict in 3ms, discard. Zero RAM, zero downtime, zero bullshit.

Tier 0 — Numeric

O(1) catalog lookup. <0.1ms. Integers >500 and #numbers resolve instantly without touching SAT.

Tier 1 — Snake SAT

On-demand from disk. 60ms load + 3ms predict. Dana Theorem: any indicator → CNF in poly time.

Vectorized /batch

One load, N predictions. get_prediction(Xs) forks across all CPUs. 1000 queries in 460ms.

Auto-Builder

Checks monce_db_all every 10min. New factory? Built and served in 60 seconds. No human needed.

Complexity

T(load) = 8.55 × MB + 19ms
T(batch_N) = T(load) + N × 0.4ms
Pr(downtime) = 0

/maths Dashboard

Live Chart.js curves. Latency, calls, factory usage, RAM, system load. Disco ball when auto-builder is on.

Who survives the Titanic?

Snake trained on 712 passengers (35s, 15 layers, zero feature engineering). Raw columns concatenated as text → SAT. 79.6% Kaggle accuracy. The model found one rule that dominates everything: Sex=female is the only ticket off the ship.

NameConfVerdict
Dana, Mr. Charles76.7%DIES
Opus, Mr. Claude82.8%DIES
Altman, Mr. Sam86.2%DIES
Amodei, Mr. Dario79.7%DIES
Musk, Mr. Elon64.5%DIES
Huang, Mr. Jensen69.8%DIES
Zuckerberg, Mr. Mark78.6%DIES
Hassabis, Mr. Demis63.0%DIES
Macron, Mr. Emmanuel73.3%DIES
Trump, Mr. Donald82.8%DIES
Modi, Mr. Narendra92.1%DIES
Swift, Ms. Taylor98.1%SURVIVES
Beyonce, Ms. Knowles-Carter85.0%SURVIVES
LeCun, Mr. Yann73.7%DIES
Karpathy, Mr. Andrej75.3%DIES
Zelensky, Mr. Volodymyr60.0%DIES
Pope, Mr. Leo XIV92.9%DIES
Son, Mr. Masayoshi84.4%DIES

25 males, all dead. 2 females, both survive. The SAT found the boundary in 35 seconds: Sex=female ∧ Pclass=1 → 98% survival. No gradient descent, no epochs, no GPU. Just Boolean satisfiability.

Snake v5.5.1 — algorithmeai — Monce SAS — Charles Dana — 2026

Published in Nature (npj Systems Biology, 2026). AUROC 0.958.