Technical specification · v0.1 · 2026-09-12
Choseverse Trajectory Engine
Given where you are right now, which available action changes your trajectory the most?
A Choseverse decision architecture by John Brajer, published through MyMindMine.
Decision model
Current State → Action → Weight Shift → New State → New Possibilities
The engine treats a decision as a state transition. It evaluates actions under the actual state that exists now, then ranks feasible moves partly by what they change immediately and partly by the future possibility-space they create.
Implemented dimensions
The September 12, 2026 kernel evaluates money, network, visibility, risk, confidence, future options, reversibility, information gain, alignment, and Path Expansion. The current state can constrain time, budget, energy, and travel radius.
Candidate actions can span Experience, Build, Career, Money, Create, Learn, Social, Reset, and Admin. Hybrid actions may belong to several modes at once.
Ranking boundary
Feasibility is checked before recommendation. Actions that exceed time, budget, travel-radius, or configured energy constraints are blocked rather than allowed to win through a high speculative score.
For feasible actions, each dimension contributes according to action effect × personal weight × current need. The implementation adds a bounded hybrid-action bonus and opportunity-cost penalties for consuming a large share of available time or budget.
Current default weights make Path Expansion the strongest positive term. These defaults are implementation starting points, not universal claims about optimal human preference.
Learning from outcomes
The original architecture required prediction to meet reality. The implemented kernel can compare predicted versus observed effects and update personalized dimension weights through a bounded learning rate.
Predict → Act → Observe → Compare → Update → Re-rank
This establishes the learning loop mechanically. It does not establish that the current model has already been empirically calibrated across a large population or that it outperforms other decision systems.
Validation evidence
- travel-radius constraint rejection;
- feasible hybrid action ranking above routine admin under the supplied test state;
- infeasible actions forced below feasible actions;
- top feasible recommendation selection;
- predicted-versus-actual error updating a personalized weight;
- unobserved dimensions remaining unchanged.
Subsequent Trillsverse commits exposed and mounted the trajectory API, seeded an action library, and added durable trajectory memory plus persistence for recommendations and outcomes.
System relationships
Perspective Expansion operates upstream by expanding which actions are visible to the decision-maker.
Path Expansion is one high-leverage evaluation term inside the Trajectory Engine, not the whole engine.
Possibility Reserve addresses longer-horizon preservation, dormancy, reactivation, and resource allocation across possibilities.
Claim boundary
The public record distinguishes the August 2026 conceptual architecture from the September 12 implementation. The current kernel, tests, API exposure, action library, and durable outcome memory are implementation evidence. Clinical efficacy, universal optimal weights, and superior long-horizon decision outcomes are not claimed.
Provenance
The originating Choseverse records established the trajectory question, state-transition model, action modes, Path Expansion emphasis, low-friction product requirement, and predicted-versus-actual learning loop in August 2026. The Trillsverse repository implemented and tested the first kernel on September 12, 2026.