LittleWorld / v0

State passes into state.

Twelve little systems live in one world. They sense locally, keep resident state, emit tiny marked events, change one another's future responses, move, and thereby change what happens next. There is no designed word vocabulary.

sender history→sender state→3-number event→receiver susceptibility→behavior→changed world

0resource captures

0events emitted

0.00selected private confidence

0.00selected receiver susceptibility

Closed-loop microscope

This animation is deliberately a playground, not the scientific receipt. The Python gate below is the matched comparison.

Communication condition Simulation speed

resource; relocates after capture agent with current local sight selected agent agent relying on resident/social state

mark = Q · [private_x, private_y, confidence]\nreceiver Δsocial ∝ susceptibility(state) · Q⁻¹ mark

Gate 0 — matched hidden-world prediction

The evidence layer removes movement and uses one fixed hidden-world tape so every arm sees the same source state and event schedule. Score: MSE only on agent-ticks with no direct observation right now.

Canonical seeds 200–211 were run after the protocol freeze. The machine-readable receipt is results/v0.json.

What v0 actually attacks

BINARY: same events, no continuous state mark.
YOKED: same event schedule, marks from an independent world.
SCRAMBLED: exact live mark multiset, wrong emission times.
RESET: live marks, receiver social state erased each tick.

The sender's private tape is generated before the arm is chosen, so a better result cannot come from giving LIVE different source observations.

What v0 does not establish

It does not show that biological spikes use these exact marks, that a receiver reconstructs a neuron's complete hidden state, that language emerges, or that evolution selected this circuit. The encoder/decoder compatibility is supplied. Later versions can make that relation learned and generational.

The supplied Gemini waveform demo is kept as motivation, not proof: its synthetic waveform features were explicitly written from the latent drive and then decoded.