The Execution Gap
The Structural Discontinuity Between Intent and Action
Published · 2026
The Execution Gap identifies the structural discontinuity between an established upstream object—such as intent, decision, authorization, policy, or qualification—and the action and result that ultimately take effect in the real world. Through verified failures across software, finance, infrastructure, and autonomous systems, it separates local correctness from end-to-end execution correctness and defines the problem before proposing its control.
Lin Wang, Mengting Wu, Yong Zhang, Jiang Deng · 2026
Summary
The Execution Gap begins with a counterintuitive fact: intent, approval, authorization, policy, qualification, and other local judgments can each be valid within their own scope without establishing that the final real-world action is correct. The book names the structural discontinuity between those established upstream objects and the execution and result that ultimately acquire effect. It then separates execution process, execution act, and result; examines failures in binding, path, boundary, state, and evidence relations; and tests those distinctions against verified incidents spanning software, finance, infrastructure, industrial systems, and autonomous vehicles. The analysis deliberately resists turning these mechanisms into a mutually exclusive taxonomy. It also distinguishes change and amplification from direct causes, showing that AI agents can accelerate and scale an existing gap without creating a uniquely AI-specific mechanism. By defining both what counts as an Execution Gap and what does not, Book I establishes a bounded problem domain. It does not yet prescribe a complete control architecture. Instead, it concludes that execution effectiveness—the fact that a particular action can take effect in reality—must be treated as an independent engineering object before Execution Control can be designed responsibly.
About This Book
Book I of the Execution Engineering Trilogy defines the problem that precedes Execution Control. It develops its claims from engineering work and verified real-world failures, preserves explicit evidentiary limits, and leaves unanswered questions open rather than filling them with an assumed architecture.
Key Topics
- Intent, authorization, execution, and result
- Local correctness and end-to-end execution correctness
- Binding, path, boundary, state, and evidence relation failures
- Autonomous systems and AI-agent amplification
- Execution effectiveness as an independent engineering object