TWEAK is a planner that simplifies and proves correct the nonlinear conjunctive planning approach. It is based on constraint posting, which defines plans by successively specifying partial descriptions. TWEAK has three layers: plan representation, goal achievement, and top-level control. The plan representation uses constraints to define partial orders on steps, allowing for flexibility in execution order. The goal achievement layer transforms plans to achieve new goals, while the top-level control structure manages nondeterminism through search. TWEAK's algorithm is proven correct and complete, and it addresses the limitations of previous planners by focusing on the representation of actions as the key constraint. The thesis analyzes past planning research, showing that all conjunctive planners work similarly, and suggests that the traditional action representation is inadequate for real-world planning. The Sussman anomaly illustrates the difficulty of conjunctive planning, where the order of steps affects the outcome. TWEAK's approach allows for efficient computation of truth criteria, enabling the planner to determine when a proposition is necessarily true in a situation. The thesis concludes that future research should focus on developing truth criteria that can handle more expressive action representations, allowing conjunctive planning to apply to real-world problems.TWEAK is a planner that simplifies and proves correct the nonlinear conjunctive planning approach. It is based on constraint posting, which defines plans by successively specifying partial descriptions. TWEAK has three layers: plan representation, goal achievement, and top-level control. The plan representation uses constraints to define partial orders on steps, allowing for flexibility in execution order. The goal achievement layer transforms plans to achieve new goals, while the top-level control structure manages nondeterminism through search. TWEAK's algorithm is proven correct and complete, and it addresses the limitations of previous planners by focusing on the representation of actions as the key constraint. The thesis analyzes past planning research, showing that all conjunctive planners work similarly, and suggests that the traditional action representation is inadequate for real-world planning. The Sussman anomaly illustrates the difficulty of conjunctive planning, where the order of steps affects the outcome. TWEAK's approach allows for efficient computation of truth criteria, enabling the planner to determine when a proposition is necessarily true in a situation. The thesis concludes that future research should focus on developing truth criteria that can handle more expressive action representations, allowing conjunctive planning to apply to real-world problems.