# Recorded LN response

This is a saved, bounded Jev run. Playback makes no inference calls and requires no API key.

Model: jev-1.13.0. Recorded API attempts: 5,000.
Individual decisions: 5,000. Simulated duration: 4,953.7 h.
Divisions: 1. Antibody-producing cells: 1. IgM output: 4922 arbitrary units.

Policy: Competing-fate demonstration: a proposed local affinity threshold favors early output or GC expansion; output cells can mature and secrete. Uncalibrated.

This qualitative policy uses a proposed affinity-annotation threshold of 0.79 to
illustrate competing early-output and GC paths. The annotation and threshold are
not measured biological fate probabilities. All responses are actual Jev choices;
the kernel enforces local eligibility, physical contacts, process clocks, finite
antigen and parent–daughter accounting. Failed selection and deaths remain visible.

The full call-budget run can extend far beyond the initial response. The current
toy model does not implement a calibrated plasma-cell lifespan or antibody decay.
Late sustained secretion is therefore a model behavior, not a prediction of vaccine
duration. Highlights show the initial response; Full recording exposes every saved tick.

Run your own experiment with Python 3.10 or newer: open Start Jev.cmd (Windows)
or Start Jev.command (macOS), enter your own key through hidden input, and set
your maximum calls per run. The key stays on your computer's local server.
The public static website only replays the saved experiment.

Original manual: docs/BIOLOGICAL_MANUAL.md in the source repository.
Recorded actions and assumptions are included in the downloadable episode.