cluster-village-10

Relations

Rows give, columns receive. 80 exchanges between 30 of 30 villagers. The most connected is smith8, with 2 partners. In this engine the one who knows everybody pays for it: belief grows with meeting people rather than with population.

Economy

Every completed hand-over, in time. The first was at t7 (smith8 to digger8) and the last at t133. A chain that closes once is a demonstration; a chain that keeps closing is an economy.

• nugget8 x 1
• nugget9 x 1
• nugget10 x 1
• nugget1 x 1
• nugget2 x 1
• nugget3 x 1
• nugget4 x 1
• nugget5 x 1
• nugget6 x 1
• nugget7 x 1
• berries135 x 1
• berries136 x 1
• berries137 x 1
• berries138 x 1
• berries139 x 1
• berries140 x 1
• berries131 x 1
• berries132 x 1
• berries133 x 1
• berries134 x 1
• berries144 x 1
• berries145 x 1
• berries146 x 1
• berries147 x 1
• berries148 x 1
• berries149 x 1
• berries150 x 1
• berries141 x 1
• berries142 x 1
• berries143 x 1
• berries164 x 1
• berries165 x 1
• berries166 x 1
• berries167 x 1
• berries168 x 1
• berries169 x 1
• berries170 x 1
• berries161 x 1
• berries162 x 1
• berries163 x 1
• berries224 x 1
• berries225 x 1
• berries226 x 1
• berries227 x 1
• berries228 x 1
• berries229 x 1
• berries230 x 1
• berries221 x 1
• berries222 x 1
• berries223 x 1
• berries234 x 1
• berries235 x 1
• berries236 x 1
• berries237 x 1
• berries238 x 1
• berries239 x 1
• berries240 x 1
• berries231 x 1
• berries232 x 1
• berries233 x 1
• berries284 x 1
• berries285 x 1
• berries286 x 1
• berries287 x 1
• berries288 x 1
• berries289 x 1
• berries290 x 1
• berries281 x 1
• berries282 x 1
• berries283 x 1
• berries294 x 1
• berries295 x 1
• berries296 x 1
• berries297 x 1
• berries298 x 1
• berries299 x 1
• berries300 x 1
• berries291 x 1
• berries292 x 1
• berries293 x 1

Clustering

The most crowded place, tick by tick. Density is what this engine charges for. Population is nearly free when the work is spread, and an agent with thirty neighbours costs about five times one with none. The peak here is 1 in one place at once.

• square1 held 1
• square2 held 1
• square3 held 1
• square4 held 1
• square5 held 1
• square6 held 1

Everyone

The spread a mean would hide. In one run of the winter the best hunger rate belonged to the strategy that also had a corpse.

Thinking

What 2740 funded deliberations came to. content is having what you want; exhausted is a proof that nothing achieves it. They used to share a word.

Minds ran short of avoids (130). Doing less than asked is sayable here on purpose. Every silent truncation this project has had to dig out had the same shape: a cap that did less and reported success.

Scale

The question the design has to answer before share means anything: at 10 clusters (30 agents), how much of the smith’s 64-bit working memory does one cluster cost on its own, and what happens when share clusters are told to meet at one square? Measured directly against compileFrame and a full start() run of the scenario above — not estimated.

Frame width at clusters=10, by share — the smith’s frame, the tightest of the three roles in the baseline. Two implementations of the same knob, because the naive reading of “generated by index” and a deliberately scoped one land in different places:

Requested at the top of this run: share = 1, 3, 6. At 3 and 6, both implementations overflow — the naive one immediately (share=2), the deliberately-scoped one by share=5. share is not free headroom to dial: it is a cliff, not a slope, because grounding reads objects within kHops of the shared square whether or not anyone is physically standing there this tick. A believed workplace two hops from a shared square is still two hops away — which means the survey’s own “an acquaintance not co-located costs 0 bits” finding does not transfer here: co-located and within-kHops are not the same test, and a share-group puts every member’s workplace inside kHops=2 by construction, every tick, whether anyone trades that tick or not.

Population, on its own, is exactly as cheap as the households anchor predicts. clusters from 3 to 100 at share=1 left the smith’s frame unchanged — still 29/64 bits, 24 ops — because ownMap plus a scenery() scoped to one cluster means a mind never hears about a cluster it doesn’t belong to. Wall time over a real start() run tracked the same story: 332.6 ms/agent at 3 clusters, 307.1 at 10, 303.2 at 30 — flat, the households shape, not the fort’s. This design sits squarely on the households side of the ladder: share is the fort’s hub risk reintroduced on purpose, and clusters alone is free the way isolated corners always are.

Frame numbers from a probe against the real compileFrame/KB/schema machinery (not the ECS loop); population and wall-clock numbers from a full start() run of the scenario built above. Neither edited village.ts or any other repo file.