Hold The Horde — Connecting The Station

Status. An early-reaction proposal for the owner's selected direction: certain actions join rooms into groups that stay connected through shifts, and connecting the entire station is one of the deepest and latest puzzles. Nothing here is canon — not the machinery, not the architecture, not any ending. Early shift uncertainty and permanent connections are preserved.

The proposal

Legality is a seam signature. Every coupling face carries a short signature in the builders' cut script. Two sections whose facing signatures match were built adjacent. At a matching seam the coupling can be latched; at a mismatch it visibly refuses.

Making the latch needs a live run. The coupling only engages while a service main runs live across the seam. Restoring that run is the work.

Once latched, it is permanent. This is the part that matters most for the owner's requirement: a latched coupling is mechanical, and nothing later un-does it. Power loss does not. De-energising the main does not. A circuit accident cannot erase progress, because after the latch the run's state is irrelevant.

What makes it a station-wide puzzle

If binds survive anything, why plan globally? Because the constraint is at the moment of latching, and source fittings are scarce and reusable.

A main carries one live source. To latch a seam you must energise that main from one fitting. Having latched, you can pull the source and use it elsewhere — the latch holds. So the puzzle is not maintaining power; it is sequencing: which main to energise, from which fitting, in what order, so that every seam gets its moment of live run.

That is a genuine global problem. Sources are few, mains overlap, and a fitting driving one main cannot drive another at the same time. Bind greedily near the ship and you can strand a distant region whose only usable source is the one you left committed.

A local example

Working. Blue Ward's east flange and Lift Service's west flange carry the same signature. The cable main between them is dead at a pulled contact. Re-seat the contact: the run goes live from Lift Service's source fitting, the coupling latches, and the pair is bound for good.

Refused, mismatch. Cold Storage's north flange carries a different signature from anything it currently touches. The lock refuses on contact. Cost: the trip, nothing else.

Refused, two sources. Recovery Stair's flange does match Lift Service's north face, but Recovery Stair sits on a stub already driven by its own fitting.

• Seeing it beforehand: a source fitting shows plainly whether it is driving, and a live main carries a flow tell along its length. Standing at the seam, the player can see tells running toward each other from both directions. Two feeds, one main.
• The action: pull the drive contact at one of the two fittings — the same class of work as re-seating one.
• The observed result: that side goes dead, the opposing tell stops, the run is live from a single source, and the coupling latches.

A failed attempt teaches the constraint and costs only the attempt. Nothing is destroyed, no bind is lost, and no penalty is proposed beyond time and oxygen.

Effects, stated exactly

Undo is unlatching, and it needs the run live again — the same requirement as latching. So undoing is symmetrical work rather than a free reset.

The later insight that changes a plan

Not "seams are unique" and not "there is a topology." Concretely: the same fitting can be the only viable source for two different mains at different points in the sequence. Realising that turns the endgame from a route problem into an ordering problem — the player stops asking which seam next and starts asking what must I latch before I can move this source. A plan already in progress gets reordered by that realisation, which is what makes it a late discovery rather than a fact to collect.

Reaching an undo — an access requirement, not a proof

A mistaken bind's release point sits inside the group it made, and a group's interior stays connected. But that does not show the player can get to it from the ship, and I no longer claim it does.

What the design actually requires: every coupling point must be reachable from the ship's point through some sequence of re-seats. That is an authored connectivity condition on the arrangement — something a designer must guarantee, not something the fiction proves. Its cost to the player is real: a group out of reach may take several expeditions and several deaths before it comes around, and the undo then still needs the run re-energised. Softlock is designed against, not demonstrated.

What I am leaving unestablished

My earlier claim that historical adjacency records stay predictive after binding does not hold up. The player knows the binds they made; they do not know the whole free set, because they have not seen every section. Any order among re-seats is an unestablished authored condition. If it exists, its legibility has to be designed, and nothing here shows the player could derive it. No forced death is proposed to prove any of it.

Completion, provisionally

Every room and region in one connected network. Not selected hubs — everything genuinely belongs. I am dropping the spanning-versus-every-seam question: extra latches beyond a connected whole would mostly duplicate geometry, and asking the owner a technical question with no player-visible difference is not worth his time.

Stabilisation is a proposed payoff, not an accepted rule. With nothing unbound, nothing re-seats, and the station holds one shape. Contents still vary — population, hazards, machine states — so it is stable rather than inert. This is a proposed change to every-death shifting and needs the owner's acceptance; it is separate from any ending, and completion is not required for one.

The simpler alternative, kept

Seams alone. One rule: matching signatures latch, mismatches refuse. It is markedly easier to learn and to author, and it fully satisfies the owner's direction.

What it loses is the endgame. Without the source sequence there is no ordering problem, so completion becomes a long collection task and patient brute force dominates it more completely. I prefer the combination for that one reason — the sequencing is the only thing in either proposal that makes the last stretch reasoning rather than accumulation — and I accept its real cost, which is a second system to teach.

If the owner wants a lighter game, seams alone is the honest choice and I would not argue hard against it.

Notes and remaining dissent

Registration is not a rival: a terminal could be the readable interface to the source plan, showing which fittings drive what and why a latch refused. That is worth exploring and does not add a system.

Nothing here requires the binding puzzle to carry both alien languages; that would be an invented obligation.

Remaining dissent: whether the one-source rule reads as reasoning or as arbitrary plumbing is unresolved and I cannot settle it on paper. It is the thing I would most want checked once observation is authorised.