Skip to content

Concept: The Last Stand — a buildable extraction defense

Pitch

The labyrinth is the mine; the extraction is the fort you build out of it. Calling the portal starts a hold-out against waves — that part already exists. The change: the player can now spend a resource they mined during the climb — timber, pried out of fights, nests and side rooms — on towers, barricades and traps around the extraction pad, and then defend a position they authored.

The one-line rule the whole concept encodes: you can run past everything, but the finale is paid for with everything you ran past. A speedrunner arrives at the chamber alive, poor, and alone. A raider who fought the level arrives with the makings of a fortress.

Problem / opportunity

The corridor floor (see PROJECT_CONTEXT.md §8.5.2) currently has two halves that barely speak to each other. The climb is a chain of optional fights: packs are leashed and steerable-around, side rooms and nests are detours by design, and the corruption wave only punishes standing still — not skipping. The finale is a fixed combat check: waves arrive, the player kites and shoots, the portal opens. Nothing the player did on the way up changes the hold-out except the gear they happened to find.

That leaves three gaps, all of them one gap:

  • Skipping is quietly optimal. Every pack is a cost with no finale payoff, so the strategy the layout rewards is the one that engages with the least of it. The game builds authored encounters and then pays the player to walk around them.
  • The finale plays the same every run. The siege is the level's climax, but it has one verb — fight — and one shape. There is no strategy layer, no fort to be proud of, nothing to screenshot, nothing to do differently next run.
  • Thoroughness has no currency. Loot pays the meta economy, but nothing pays this raid's ending. The extraction-shooter tension ("did I collect enough?") stops mattering the moment you reach the chamber — exactly the moment it should peak.

The opportunity: tower defense is one of the most mobile-native genres there is, the game already owns the hard parts (telegraphed wave spawns, a chamber arena, relentless attackers, a fixed hold-out window), and a build-your-defense beat converts the whole climb into preparation — which is the extraction-genre promise said out loud.

The concept

Timber: the raid's third resource, and the only one that dies with the raid. Fights drop it, garrisons guard it, nests are fat with it, dead-end detours stash it. It is vacuumed up like coins, shown as a simple counter, and — critically — it is run-scoped: unspent timber does not bank at extraction. It exists to be spent on this floor's finale or lost. That single rule keeps it out of the meta economy and makes hoarding feel as wrong as it is.

The chamber is a build site. From the moment the player enters the extraction chamber, a build button sits alongside the combat controls. Before the portal is called there is no time pressure: the player walks the chamber, reads the spawn edges, and places defenses at their leisure — the calm-before planning beat. After the call, building stays possible but happens under fire: a mid- wave barricade slammed down in a doorway is the mode's hero moment.

Three things to build, three different sentences:

  • The tower shoots things for you. A placed ally that attacks approaching enemies on its own. It turns the player from the only gun into the commander of a crossfire — stand where the tower's field and your own overlap, and positioning becomes fort design.
  • The barricade decides where the fight happens. A wall segment with health. Waves that cannot reach the player smash through it, audibly and slowly — it converts enemy time into player time, and shapes wave flow into lanes the towers and traps are watching.
  • The trap punishes the path. Placed flat on the ground, invisible-ish to enemies, single- use: a spike burst, a snare that holds a heavy in tower fire. Cheap, satisfying, and the natural first purchase for a poor run.

Waves attack the fort, not just the player. The siege's existing promise — spawns telegraph, attackers never lose you — stays. What changes is what stands between them and you. Enemies that cannot path to the player attack the obstacle in the way; a fort is therefore spent by the siege, plank by plank, and a well-built one is the difference between a last stand and a formality.

The bad strategy is legal, and it is a lesson. Nothing stops a player sprinting the corridor, calling the portal with empty pockets and white knuckles. Sometimes — geared well enough — they will even make it, and that escape will feel heroic. But the design's center of gravity is the comparison the player cannot help making on their second raid: the run where they fought for timber ended in a fortress and a shower of planks converting into a calm, earned exit. The labyrinth stops being distance to cover and becomes the question "how much ending can I afford?"

Defense is authorship. Two raiders with the same timber build different forts — a turtle ringed in barricades, a killbox funneling everything through one trapped lane, a mobile fight with towers as covering fire. The finale becomes the place where a run's personality shows, which is what makes "one more raid" mean "one more idea".

References

Grounding note: the project's gd-knowledge MCP corpus was not reachable in the session that produced this document, so nothing here is cited from it. All entries below are named as design-pattern references — structural comparisons from general industry knowledge, with no sourced metrics claimed. If the corpus (which leans idle/gacha) is consulted later, tower-defense monetization and session-shape data should be pulled to pressure-test the retention argument.

  • Fortnite: Save the World — the closest structural relative: scavenge a zone for build materials, then erect a fort around an objective and defend it against telegraphed waves. Its core lesson is exactly this concept's thesis: exploration and defense stop being separate activities when one is the budget for the other.
  • Orcs Must Die! / Dungeon Defenders — the action-TD hybrids proving that placed defenses and a player-controlled fighter share the kill count without the player feeling replaced; the player is the fort's most mobile turret, not a spectator.
  • Deep Rock Galactic (salvage/uplink holds) — hold-a-perimeter beats inside a larger extraction-shaped mission; notable for how strongly where you choose to stand reads as strategy even with minimal building.
  • They Are Billions — the purest form of "everything you gathered before the final wave is your only defense during it"; a cautionary reference too, since its difficulty cliff is exactly the failure-spiral risk flagged below.

Fit with Shadow Raiders

It completes the corridor floor's own logic. The layout already builds discrete, optional, priced encounters (leashed packs, garrisoned loot rooms, nests) and already ends in a wave defense with telegraphed spawns and relentless attackers (PROJECT_CONTEXT.md §8.5.2). This concept adds the missing wire between the two: the encounters fund the defense. Nests in particular graduate from "optional HP wall with loot" to the level's timber elites — the hardest fight on the way up becomes the biggest single investment in the ending.

It respects the siege's fairness grammar. Telegraphed spawns stay the fairness beat; attackers that never lose the player stay the anti-cheese rule — and barricades are what makes that rule fun rather than just strict, because relentless pursuit is exactly what walls convert into time. The corruption wave halting below the chamber already guarantees the fort is never fought inside the murk; corruption keeps its job (pace the climb), timber keeps its own (price the finale) — one pushes you forward, the other pays you for fighting rather than fleeing, and between them "advance, engaging deliberately" becomes the taught line of play.

It is chamber-scoped on purpose. Defenses exist only in the extraction chamber, only this raid. No corridor turrets, no persistent base-building, no second meta economy — the 8-slot gear and blueprint loops are untouched, and timber's run-scoped death is the boundary that keeps this one beat instead of a second game. (The pre-dive rule — one floor, extract to lobby — makes this clean today; if multi-floor dives return, "does timber carry between floors?" is a decision listed below, not assumed.)

It gives the master-level test its missing measurement. The current master floor asks "which mechanics are worth keeping?" This concept makes the room types comparable on one axis — each room type's timber yield versus its cost in health and time — which is precisely the data the keep-or-kill decision wants.

Risks & open questions

  • The failure spiral is the central risk. A struggling player fights less, so they arrive poorer, so the finale is harder, so they struggle more. They Are Billions embraces this; a casual mobile game cannot. Candidate rails — a timber floor granted on reaching the chamber, a loud mid-raid "you're under-timbered" warning, wave count scaling gently with timber spent — need picking deliberately, and this is the first playtest question.
  • Turtling and degenerate forts. A full barricade ring with the player inside could trivialize the siege ("build 20 walls and wait"). Caps on simultaneous structures, wave types that answer walls (a barricade-eater, a lobber that arcs over), and the portal's boarding beat requiring the player to stand somewhere specific are the standard answers; which are needed is a tuning question, not a design one.
  • Build UX on a phone, mid-fight. Calm-phase building is easy; under-fire building with fat fingers on a small screen is not. The concept assumes placement is tile-snapped, near-the-player, and three-items-simple. If a playtest shows mid-wave building is unusable, the fallback — all building locked to the pre-call phase — preserves the concept but loses its hero moment.
  • A third currency's cognitive load. Gold banks, blueprints bank, timber dies with the raid. That asymmetry is the design, but it must be visually unmistakable (distinct pickup, distinct counter, a spend-it prompt at the chamber) or casual players will feel robbed at extraction.
  • Does the calm build phase drag? An unhurried planning beat is the reward for arriving; it is also dead air on a speedrun-shaped floor. Watch whether players linger unhealthily or skip building entirely; the corruption wave's arrival at the chamber threshold naturally bounds the dawdle, and may be enough.
  • Unspent timber at boarding. Evaporating is cleanest for the economy; converting a remainder to a token gold amount is kinder. Small decision, sets the tone of the whole resource — pick at first playtest.
  • Does timber carry across floors if the dive returns? Carrying rewards deep runs but re-opens hoarding; resetting per floor keeps every finale honestly priced. Deferred until the dive structure is re-enabled.
  • Wood vs hammers vs something else. "Timber" is a placeholder identity. The fiction matters mostly for readability at a glance (pickup silhouette, counter icon); resolve with art, not design.