Path Expansion began as a decision principle: an action matters partly because of the desirable future states it unlocks. The Choseverse Trajectory Engine turns that principle into an executable ranking kernel.
The public v0.1 record separates the August conceptual architecture from a September 12 implementation. That distinction matters. A framework can exist before code; code can exist before validated real-world performance.
State before score
The engine starts with current state: time, budget, energy, travel radius, stuckness, and dimension-specific needs. Candidate actions can span experience, build, career, money, create, learn, social, reset, and admin modes.
Before ranking, an action has to be feasible. The current kernel can reject actions that exceed time, budget, travel, or energy boundaries rather than allowing an attractive score to hide an impossible move.
Ten dimensions
The implementation evaluates money, network, visibility, risk, confidence, future options, reversibility, information gain, alignment, and Path Expansion. Each action effect is combined with personal weights and current need. Hybrid actions can receive a bounded bonus for shifting several useful dimensions at once.
Path Expansion receives special treatment in the default kernel, including tie-breaking, because the engine is explicitly designed to value future option creation rather than immediate reward alone.
The learning loop
The deeper promise is not one permanent scoring formula. It is a system that can compare predicted outcomes with observed outcomes and update personal weights over time. In other words, the engine can become less generic as it learns which kinds of actions actually change one user’s state.
The code does not prove the model is optimal. It does establish something more concrete than a diagram: the state-transition philosophy now has an implemented decision kernel that can be tested.