Early PoC exploring how to make distributed capacity markets (DePIN, restaking, AI inference slots, etc.) provably non-overcommitting — even under network partitions and adversarial excitation.Core idea: enforce strict global conservation (Σ allocations ≤ CAP) via:Contractive dynamics (Banach fixed-point convergence)
Lattice join merges (⊔ = max, never additive)
Budget-token gating + partition-local caps (≤ CAP/n)
Try the guided 5-step interactive simulator (single HTML file, no install): it demonstrates normal operation, stress to cap, partition attack, adversarial injection, and safe reconnection/merge with invariants always held.Full technical report (proofs of conservation, contractivity, partition/merge safety) in DOCS.md: https://github.com/Distributed-markets-overcommit/Distribute...: https://github.com/Distributed-markets-overcommit/Distribute... welcome on the math, sim behavior, or potential DePIN/restaking applications. Early stage — honest critique appreciated.