Language between: predictions

Two eight-position rings exchange words and replies while their clocks turn at different rates.

Protocol fixed before running

Material study

What would change the question

Protocol fixed before running

A world has eight possible light positions. At each world change a seeded generator picks a position independently and uniformly. A samples it on one clock, encodes its observation as one of eight tokens, and sends a message. B reads on another clock, chooses a position through its dictionary, and echoes that choice. A can reply with one bit: the choice fitted or did not fit A's original observation. Message identifiers associate the reply with the earlier choice. B may eliminate rejected interpretations and retain accepted ones. No policy receives the auditor's true world position. A has its sensor reading; B has the received token and returned bit.

Transport fidelity counts uncorrupted delivered token messages. Reference accuracy counts actions matching the actual position at the time A sampled it. Present success counts actions matching the actual position when B acts. These denominators differ and will be reported. Coarse sensing merges adjacent pairs; its feedback accepts either member of the observed pair. A durable channel keeps the latest unread message; an ephemeral channel forgets it after its lifetime. Neither is an infinite queue.

Use seeds 1–12, 24,000 integer ticks per run, and score actions from tick 6,000 onward. Base settings: world changes every 37 ticks, A samples every 5, B reads every 7, one-way delay 3, eight sensory bins, zero noise, durable messages, shared dictionary, no repair. A tick is an abstract unit, not a claim about a nervous system or relativistic time. All changes below are relative to this base. Means pool events across the 12 seeds. Report failed predictions without changing thresholds.

ID	Prediction	Confidence	Fails if
P1	With a shared dictionary, transport fidelity and reference accuracy are both exactly 1. Present success is lower than reference accuracy.	99%	Either exact score differs from 1, or present success is 1.
P2	Rotate B's interpretation by three positions, keep repair off: transport fidelity remains 1 and reference accuracy is 0.	99%	Either equality fails.
P3	With that rotation and repair on, reference accuracy after warm-up exceeds .95 and exceeds the frozen P2 condition by .90.	90%	Either threshold fails.
P4	Equal clocks do not suffice: A and B both every 5 ticks, world every 7, one-way delay 80, shared dictionary. Reference accuracy is 1, present success is below .20.	95%	Either criterion fails.
P5	Unequal clocks can work: A every 4, B every 40, world every 1,000, shared dictionary. Reference accuracy is 1 and present success exceeds .93.	95%	Either criterion fails.
P6	A every 7, B every 41, world every 1,000, lifetime 1: ephemeral uptake (actions per B reading opportunity) is below .35; durable uptake is above .95.	95%	Either threshold fails.
P7	World every 1,000: reducing A to four bins gives present success below .60, while eight bins give above .95, with the shared dictionary. Perfect transport cannot restore a distinction the sensor discarded.	90%	Either threshold fails.
P8	Multiply every clock, delay, expiry and run duration by 7, including the warm-up boundary: P3's event counts and all three scores are identical. Ratios and event order, rather than the chosen tick unit, determine this model.	99%	Any scored event count or score differs.

Language between: predictions?

Protocol fixed before running?

Material study?

What would change the question?

Language between: predictions

Protocol fixed before running

Material study

What would change the question