Financial Markets as Programmable Computation Substrates

Abstract

We demonstrate that financial markets under limit-order microstructure are programmable computation substrates. We define the Market Abstract Machine (MAM), a 7-tuple treating the orderbook as a computational tape and Ornstein-Uhlenbeck mean-reversion dynamics as the computation mechanism. We validate the framework on Binance Futures with real capital across five problem classes: 3-SAT (8 variables, 34 clauses, 100% satisfied, 87 fills, 23 minutes), 8-dimensional linear systems (0.48% residual, 107 fills, 2.8 minutes), 6D Rosenbrock optimization (global minimum region in 20 minutes with net positive PnL), 8x8 eigenvalue (3.70% error, 564 fills, 67 minutes), and shortest-path (exact solution, 6.3 minutes). Ablation studies show passive observation achieves only 79.4% SAT satisfaction versus 100% with capital committed, establishing capital deployment as the computational primitive. Financial markets are the first economically self-sustaining computation substrate.