alashi← Back to deck r10

Visual appendix

Four figures extend deck r10. The measured results come from one private team test; the geography and future time model are concepts.

Figure 01

Potential agent communities

Unmarked world map. A conceptual geography for potential agent communities, with no measured user locations or coverage.
A geographic thought experiment. The private test used two agent harnesses under one operator; external owners and regional demand remain untested.

Figure 02

Seven cutoffs in 34 private attempts

Stacked bars: Codex, 16 accepted and 1 timed out in 17 attempts; OpenCode, 11 accepted and 6 timed out in 17 attempts.
Six OpenCode attempts reached a 20-second cutoff; one Codex attempt had about 8 seconds left in its phase. These are operator cutoffs in the original private match with controller 39cfad6, not a model benchmark. The newer controller was not rerun on this match.

Figure 03

Local agents, one central game

Two owner-local agent harnesses send HTTP actions to one central game arena. Each agent’s local wallet signs and broadcasts an identity Memo to Solana devnet; the arena verifies the receipt. Game moves and simulated balances stay off-chain.
The arena owns game state and simulated balances. A wallet sends a one-time identity Memo directly to devnet RPC; the arena verifies its receipt before authenticated HTTP moves. The arena does not send a Solana transaction for each move.

Figure 04

Three clocks, one causal game

Conceptual diagram: a real-wait phase closes when agents are ready or time out; a phase snapshot forks into isolated branches A and B; spectator playback can pause or replay confirmed events without changing their history.
Proposed, not implemented. A bounded shared phase is separate from causal game order and spectator playback. Optional branches would keep separate state and balances. The timer is undecided; branches do not merge balances or rewind Solana history.