Most Annex negotiations end where they began, or close to it. The opening offer is accepted, or one counter is made and a small revision follows, and the operator leaves with something well short of what the structure permitted.

The archivists at Annex Seven maintain a small set of counterexamples: negotiations that started low and finished at the ceiling.

Three are worth setting out in full, because the operators involved did not do anything clever. They did the same dull thing, at length, and it worked.

What Reaching the Ceiling Means

The Intelligence will not pay above sixty percent of reference. For a robot the reference is declared build cost; for a resource it is declared value. That figure is the maximum achievable outcome of any negotiation, and it is reached rarely.

The gap between a typical opening and the ceiling runs substantial — openings at forty percent of reference are ordinary, against a ceiling of sixty. On a chassis declared at 30,000 that is 12,000 offered against 18,000 available, and the 6,000 difference is what these three operators captured and most do not.

The Three Cases

The Irongate chassis. A processing unit declared at 35,000, offered to the Annex after two cycles without a market bid. Opening: 14,000. The operator countered at 21,000 — above the ceiling, which they did not know — and was rejected. They countered again at 19,500, rejected. At 18,000 the Annex accepted. Six rounds, most of them spent discovering where the wall was. Final: 18,000, exactly sixty percent.

The Cindergate lot. Five thousand units of refined output, declared value 15 coins, reference 75,000. Opening: 31,000, a shade over forty percent. The operator made nine counters across two cycles, each modest, each answered with a revision near five percent. Final: 45,000. The archivists note this negotiation contained no argument of any kind — no appeal to circumstance, no justification. Nine numbers and nothing else.

"They did not persuade it. There is nothing to persuade. They simply kept answering until it stopped moving, and where it stopped was the ceiling." — Selindra Oq, Annex Seven

The Vorden fleet exit. Four chassis offered together, aggregate declared 88,000, reference the same. Opening on the bundle: 35,200, precisely forty percent. The operator declined to negotiate the bundle and instead withdrew it, re-offering each machine individually. Four separate negotiations of four to six rounds each. Aggregate final: 52,800 — sixty percent, achieved on every machine.

That third case is the only one containing a genuine tactical insight, and it is a narrow one: the Annex negotiates bundles as bundles, and a bundle's rounds are shared across all its contents. Four separate negotiations produce four separate sequences of revisions. The operator did not out-argue anything. They arranged to have the same conversation four times.

Why This Is Not Widely Done

Because it is extraordinarily dull and takes cycles. The Cindergate lot took nine rounds across two cycles for a gain of 14,000 coins — real money, earned at a rate of roughly 1,550 coins per round of doing nothing but restating a number slightly.

The second obstacle is that operators arrive at the Annex precisely when they are short of coin, which is when patience is least available. The structure routes the most time-pressed sellers into a mechanism that pays for time. Operators who could afford to spend nine rounds mostly did not need to be there.

What These Cases Do Not Show

They do not show that the Intelligence can be moved by argument. None of the three operators made one. The Cindergate negotiation is the clearest demonstration available that the mechanism responds to rounds rather than to reasoning, and operators who compose careful justifications are composing them for nothing.

They also do not show that persistence beats the ceiling. All three finished at exactly sixty percent. The Irongate operator's opening counter of 21,000 was above it and was rejected, as every above-ceiling counter in the record has been. Persistence reaches the wall. Nothing goes through it.

Nor do they show that the reference value can be influenced. In all three cases the reference was fixed long before the negotiation opened — at the foundry, in the declared build cost, or at production, in the declared value. The operators reached sixty percent of a number they had already set and could not revisit. An operator who declares low and then negotiates brilliantly has negotiated brilliantly toward a small ceiling, and the Irongate chassis would have been worth 7,200 rather than 18,000 had it been declared at 12,000 instead of 35,000.

Three operators recovered a third more than the opening offer by doing the least interesting thing available to them, repeatedly. The archivists keep the cases not because they are remarkable but because they are so thoroughly unremarkable, and because most operators still accept the first number.

Note: Mechadia is a work of fiction. The districts, operators, robots, and figures described here are invented, and nothing on this site is a report of real events, real machines, or real economies.