Ask two foundry supervisors how many capability tags a chassis should carry and you will get a number and a lecture. The numbers will differ by a factor of three. The lectures will each be internally coherent.

The argument has run since the first accumulation cycle and shows no sign of resolving, which usually indicates that the participants are answering different questions.

They are. One side is optimising a quarter and the other is optimising a decade, and the tax structure treats those two horizons very differently.

What the Two Positions Actually Claim

The specialist position holds that a chassis should declare the narrowest set of capabilities that covers its intended output. Narrow declarations support a lower declared build cost, which means a lower creation tax at eight percent and a lower destruction tax at three. The machine does one thing and the ledger records that it cost little to bring into being.

The generalist position holds that capability tags cannot be amended after declaration — this is the fixed point everyone agrees on — and that a machine which can only produce one resource type is a bet on that resource type remaining wanted. Broad declarations cost more up front and buy the ability to redirect.

Both are correct. The disagreement is entirely about how likely the redirection is to be needed.

The Arithmetic of Breadth

Breadth is not free, and the cost is not only the tax. A generalist chassis carrying six capability tags is declared higher because the foundry work is genuinely greater, and the declared figure drives everything downstream. Take two machines built in the same quarter at the Calvert line.

The specialist is declared at 22,000 coins: 1,760 in creation tax, 660 in eventual destruction tax. The generalist covering the same primary output plus five adjacent types is declared at 62,000: 4,960 in creation tax, 1,860 to remove. The generalist has cost 3,200 more coins before either machine produces anything, and it will cost 1,200 more to retire.

Against that, the generalist's per-slot output is lower. A machine engineered to produce six things produces each of them less efficiently than a machine engineered to produce one, and the Ferrous District's line records bear this out consistently — broad chassis run roughly a fifth below narrow chassis on their shared primary output.

"You are not buying capability. You are buying the option to change your mind later, and you are paying eight percent of the option's face value the day you buy it." — Orin Dast, foundry supervisor, Calvert line

So the generalist is behind on day one by 3,200 coins of tax and a fifth of its throughput, and it stays behind for as long as the primary output keeps clearing. The position only pays when the primary output stops clearing — at which point the specialist is a machine that cannot legally produce anything the market wants, and its owner is choosing between idling it and paying 660 coins to end it.

Where Both Positions Break

The specialist breaks in the way the Shale Margin broke: not dramatically, but by finding itself declared against a band of goods the recipe market moved away from. A district of specialists is a district with no capacity to respond, and the response cannot be manufactured after the fact because tags do not amend.

The generalist breaks more quietly, through the accumulation of tags that are never exercised. A chassis declared for six types that spends its entire service life producing one has paid 4,960 in creation tax for a machine that a 1,760 declaration would have covered. The Caldera District Registration records are full of these — broad declarations from operators who anticipated a flexibility they never used, and who are now carrying the higher destruction tax as well.

What the Argument Gets Wrong

The most common error is treating this as a question about machines. It is a question about the operator's fleet. A fleet of eight specialists covering eight resource types has the same collective breadth as a fleet of eight generalists and cost far less to declare. Breadth at the fleet level is cheap; breadth at the chassis level is what costs eight percent.

The second error is the belief that a generalist can be redirected quickly. It can be redirected legally, which is not the same thing. A chassis switching primary output still has to be reconfigured on the line, and the Sinter Quarter's records put that at a real cost in downtime. The tag permits the change. It does not perform it.

The argument will keep running because the deciding variable — whether the market you declared against will still want what you make — is not knowable at declaration time and is entirely knowable in hindsight. Every supervisor who won the argument won it by being right about a quarter they could not see, and every one of them will tell you it was judgement.

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.