python-agenda.md

Python Agenda

Objective

Use Python to complement MATLAB in three fronts:

• fast validation of propositions;
• automated consistency tests for minimal models;
• generation of figures, parametric sweeps, and auditable computational appendices.

Canonical scope already implemented

1. Bounded-optionality model

• two policies: exploit and preserve;
• discounted-value calculations;
• numerical verification of Theorem 1 in the o_t in [0,1] regime;
• explicit exit cost incorporated into the core.

2. Reinforcement-based capture model

• logistic stay function implemented;
• sensitivity of p_stay to local reward measured;
• expected duration in the bad game estimated and tested.

3. Competence-capture model

• local competence q represented;
• demonstration that more effective agents may remain longer in the bad game;
• comparison between short horizon and total structural value.

Current canonical tests

1. case in which exploit wins locally and loses structurally;
2. case in which the inequality of Theorem 1 fails and exploit ceases to be dominated;
3. monotonicity of stay probability with respect to reward;
4. monotonicity of expected duration in the bad game with respect to local competence;
5. invariance of optionality state inside the interval [0,1];
6. effect of explicit exit cost on the boundary;
7. monotonicity of phase-map slices.

Canonical regime-analysis layer

• pure-Python sweep over parameter space;
• CSV export with regime classification;
• SVG export with phase map;
• automatic Markdown report with quick reading;
• slices by structural weight and erosion factor;
• metadata.md with model version, timestamp, and fixed parameters.

Legitimate next expansion

• notebooks with parameter sweeps;
• richer graphics and multiple sections of parameter space;
• calibration of case studies.