Owster LabsChaos Theory — Phase 1
Methodology
Three scenarios. 1.8 million simulations. One question: what happens when human strategic judgment meets AI precision?
Total Simulations
1.8M+
Scenarios
3
Strategies Tested
11
The Napoleon-Berthier Thesis
Napoleon chose where to fight.
Berthier calculated how to get there.
Human strategic architecture — the instinct to flank, sacrifice, deceive, concentrate — combined with AI route optimization and probability calculation consistently outperforms either alone. This holds across three independent scenarios with different structures, confirmed at publication-grade statistical significance.
Human strategy + AI logistics
over pure AI (Scenario 002)
Human flanking + same AI route
over pure AI (Scenario 001)
Pure AI synthesis (Claude+GPT)
beaten by solo human (Scenario 003)
Scenarios
Each scenario is an independent experiment. Click to see full data, strategies, and analysis.
The Dogfight
3× Rafale vs 4× F-16 Block 70
Flanking geometry beats volume. Human's multi-phase Napoleonic Divide outperformed GPT's single-phase BVR ambush across 100,000 runs.
Simulations
200,000
Human +2.5pp over AI
The Cargo Escort
2× UCAV + Cargo vs 5× Su-30MKI
Human tactics on AI's route = +80% over pure AI. Sacrificial decoy at the right moment beats symmetric escort — geometry over symmetry.
Simulations
400,000
Napoleon-Berthier 4.44% vs GPT 2.46%
The Hunter Strike
3× Ghatak UCAV vs HVT + 4× Su-30MKI CAP
Claude's math maximizes expected value but accepts a 0% blind spot. Human+GPT covers all states symmetrically. Two AIs without human input scored second-worst.
Simulations
1,200,000
Human+GPT 11.09% — full coverage wins