Roadmap

The smallest thing that can answer one question — is fighting while defending actually fun? Placeholder blocks only. No economy, no real art.

• Fresh Unreal 5.8 project, old farming-sim build shelved
• Flat greybox arena, built procedurally at runtime
• Controllable blocky hero — move, jump, camera, one melee attack
• Prep phase — place 2 scarecrow turrets + 1 barricade with a ghost preview
• Live wave — collectors walk in from three lanes and attack what blocks them
• Turrets auto-acquire and fire; the player fights alongside them
• Win/lose readout, restart on R
• FUN GATE — PASSED. Playtest verdict: "dumped right into it, felt great."

Extras built beyond the original Phase 0 contract

• Combat juice pass — knockback, hitstop on impact, camera shake, floating damage numbers, enemy + barricade health bars
• Collectors pop into a shower of coins instead of dying (family-friendly)
• Headless smoke-test harness (-CCAutoWave) — plays the game with no human and reports the result, used for every balance change
• Autopilot hero (-CCAutoHero) — measures the player’s actual contribution to a wave instead of guessing at it
• Vault your own barricades — walls are short enough for the hero to jump, too tall for collectors. Walls are purely anti-enemy; hero mobility is an edge.

Balance measured, not guessed (16-collector wave)

The wave ends and the world doesn’t stop. Villagers, jobs, and the first production chain. This is the layer that turns a defence prototype into CropCorp.

• Wave 1 starts on a timer — keeps the "dumped right in" feeling that tested well, while R still sends them early for players who want to build first
• Incoming-wave warning with a countdown that goes red and pulses under 5s
• Villager NPCs exist, are counted, and are visible on the HUD
• First production chain: chop trees → logs → mill → planks
• Villagers auto-repair the damaged granary once planks exist
• Job-board system — the player sets how many villagers work each job (Woodcutter · Miller · Builder · Militia), villagers carry it out
• Endless day/wave cycle — waves grow each day, run ends when the granary falls, score is days held
• Villagers take cover when collectors get close, and rest at home to heal
• Collectors fight back against militia, so assigning them is a real risk
• Trees regrow, so woodcutting is a renewable job
• Defences cost planks — turrets and barricades are bought, not granted. The per-day free turret is gone; what you can raise tomorrow is whatever the village milled today, minus what the repairs cost you
• Build panel shows prices, and an option you cannot afford turns red — the placement ghost was already refusing, now the reason is legible
• Reseeder job — felled trees leave a stump that stays a stump until a reseeder plants it. Forestry is ongoing labour, not free scenery
• Job board shows staffing targets (WOODCUTTER 2/2), red when a job is short-handed, with a STAFFED n/n summary. Targets deliberately exceed the starting population, so being under-staffed is visible immediately
• Collectors get tougher, not just more numerous — without it the run could not end (see the dev log below), and days-survived needs an ending
• Talk to an NPC; villagers approach and ask for help after wave 1
• Quest / to-do checklist UI driving the player’s next objective
• Map opens up as you clear it: felling trees frees buildable ground

Give the run a shape. Every measured run so far has been a flat line and then a cliff — nothing happens for fifteen days, then you die. Waves in chapters of ten fix that, and turn the demo into a complete arc rather than a slice that stops mid-sentence.

Survive waves 1–10 → claim more land → next chapter. Wave 10 is a boss.

• Waves grouped into chapters of 10, shown on the HUD as CHAPTER 1 WAVE 7/10 — a position in a story rather than a number that goes up
• Wave 10 boss — the Tax Collector. Twice the height of a collector, slow, heavy, swinging a money bag, walking straight up the middle lane with a thinner escort than usual so he is not lost in the crowd
• Boss announced a full day ahead, so wave 9’s prep is spent preparing for him specifically rather than for more of the same
• Clearing a chapter reclaims land — a fresh grove appears where the Authority used to hold, and the families who lived on it come home as two more villagers. The smallest version of the Phase 4 frontier that a player can actually feel
• Wave size is a sawtooth, not a ramp — waves climb through a chapter, peak on the boss, and the next chapter opens below that peak and above where the last one started
• Fence and wall are now two grades. A fence is cheap, low and vaultable by the hero but not by collectors. A stone wall is expensive, tall and vaultable by nobody — it stops a wave properly, and the price is that it stops you too. Keeping a vaultable grade in the game is deliberate: when everything was unvaultable, players sealed themselves in
• Hero death costs something. Falling puts you out for a stretch that grows with each fall in the same wave, and resets when the next wave starts. Dying used to be free, which made health and armour worth nothing
• Gates — blocked on pathfinding, not on art. A gate needs villagers to route to it, and villagers steer directly with no navmesh: they would walk into the wall beside it and sidestep forever. Gates land when pathing does, otherwise a walled village starves itself
• The village dog is the alarm. He potters about while things are calm, and when trouble is coming he trots to the side of the village it is coming from, faces it and barks. A wave that can arrive from any direction needs the player told which direction, and a chevron stuck to the edge of the screen would say it perfectly well — this says it better. The warning is a character rather than a widget. He tracks the actual crowd once a wave lands, so he already works for 360° waves, and he doubles as the opening’s only tutor: whatever he is barking at is the thing you are meant to be looking at
• Prep is a minute — long enough to be a phase of the game rather than a pause in one. Ready still skips it
• Co-op ready needs every player to call it, so one keen teammate cannot start the wave on the others
• New enemy introduced at wave 3, another at 6, a telegraphed heavy at 8 — one new idea at a time: the Clerk (fast courier, diverts for nothing), the Repo Agent (dismantles your defences, ignores people), the Bailiff (the slow heavy whose silhouette is the telegraph)
• Production rates on the HUD — the job board’s output column shows what each job actually yielded (+12 logs, +25 plnk), measured rather than derived, and the plank line shows the day’s takings against the day’s bill. Headcount alone was never a decision
• The curve is tuned, and measured rather than guessed — chapter-scale enemy growth climbs smoothly while wave size keeps the sawtooth. A hard step at the chapter boundary produced a cliff and was thrown out
• Longer prep phase — still 45s. The harness now plays all of it (see below), so the cost of lengthening it can finally be measured
• The wake-up opening — start inside a village house, walk out, meet the villagers, be guided through placing your first buildings
• The wake-up opening — "The Last Quiet Morning" — see the full spec below. Highest-value item for the demo: wishlists are won in the first sixty seconds, and this is also the trailer
• Hero can chop trees and carry wood to the mill — also gives the player something useful to do during a longer prep phase

"Clicking is fine, but doing it for 100 hours is repetitive."* Correct, and the fix is different for the two things the button does.

Hold to repeat — but they are not the same action underneath.

• Hold-to-repeat on attack, cooldown unchanged
• Hold-to-chop as one continuous action with progress feedback

The tool swaps itself. Always. “If I get to the tree I just want to chop it.” Right — the game should read the context and pick: something hostile in range means you swing your weapon, otherwise a tree in range means you swing the axe, and it swaps straight back afterwards.

⚠️ Enemies always win the context test. If anything hostile is in range you swing the weapon, full stop — even standing in a grove. The one unforgivable failure here is chopping a tree while a collector hits you in the back.

• Context-sensitive tool selection, with hostiles taking absolute priority
• Visible swap on the model, so the choice is legible rather than magic

Inventory exists for progression, not for micromanagement. It holds the tools you own and what they are worth; it is never something you open in the middle of a fight. Better axes fell faster, better weapons hit harder, and both are things you upgrade between waves — which fits the pillar this whole game is built on: depth without spreadsheet management.

• Inventory of tools and gear, opened between waves
• Tool tiers as a progression track (a better axe is a better economy)
• The first axe is given to you by a villager in the opening — it makes the conversation matter, it gates chopping behind meeting them, and it means the tutorial hands you a thing rather than an instruction

A building must say what it is without a label.** In a blocky game there is no texture or signage to lean on, so the silhouette and the props have to do it: the lodge gets a chopping stump with a log on it and an axe buried in the block, and it reads as "wood is worked here" from across the field. That also means nothing to translate — a prop is legible in every language.

• Lodge: log pile, chopping stump, axe
• Every other building held to the same test: identifiable in one glance, by props rather than by a name tag

Chop, then process — one step, never three. Raw wood should stack and then be worked into something usable, because that chain is what gives the jobs and the buildings a reason to exist. But one processing step is the whole budget: trees → logs → planks. A second hop would be a chore rather than a chain, and this is exactly where economy games start to feel like admin. Icons and localisation. Everything currently explains itself in English prose. Icons first, because they carry meaning without translation; then the languages covering the bulk of Steam’s users.

• Icon language for jobs, resources and buildings
• Localisation pass targeting ~70–80% of Steam users by language share
• Keep UI strings in one place from now on, so that pass is not archaeology

Co-op shapes the economy differently, and that is the point. With four players the natural split is two pushing the economy while two hold the line — a division of labour a solo player cannot have. Solo instead rushes three or four turrets and then pivots to economy. Both should be viable; neither should be optimal.

• Check both shapes survive balancing, rather than tuning only for solo

Everything should give XP, including running the village well — so the job board feeds the same progression the fighting does.

• XP from production and management, not only from kills — planks and reseeding 1:1, repairs per plank’s worth; Ready pays XP scaled by the prep time forfeited, so calling early is personal power vs village power
• Skill tree UI, iconographic

Missing outright — the whole shell around the game

• Main menu (2026-08-14): canvas-drawn shell over the live farm — title, PLAY, volume trio on the cc. cvars; -CCSkipIntro and the harness walk past it; -CCFrontDoorTest screenshots it
• Continue / load
• Options and settings — first rung only: the three volume knobs live on the menu; everything else still cvars
• Multiplayer, lobbies, invitations

The opening was played for the first time. Most of what came back was not about the tutorial’s content but about the game having no interaction grammar — things happen through invisible keybinds rather than through the world.

The ledger, not a noticeboard. The first proposal was to walk to a physical board and press F. On reflection the better version is that the hero takes a sheet of paper out and looks at it — the villager management screen is a thing you carry, not a place you travel to. No walk, no proximity, available the moment you want it, and still diegetic. The board in the corner of the screen goes away entirely.

“Easy is the key — confusing will lose the gamers right away.” That line governs this whole screen. It is the most-used interface in the game and it is the one a new player meets in their first five minutes.

• Hero raises a paper ledger; it opens the village management screen
• Drawn as paper — the one thing in this HUD that is not a translucent overlay — with each job on its own row: what it does, how many are on it against how many it wants, and what it yielded yesterday
• Big [-] and [+] keys sized to be hit without aiming, the selected row marked so a controller can see focus with no cursor
• Mouse on desktop, d-pad and arrow keys elsewhere; TAB or the View button takes it out and puts it away, ESC closes it
• Reading costs you: the hero stops dead and cannot swing or build while the sheet is up, and the village and the wave both carry on without him. It can be opened mid-wave — that is the decision, not a restriction
• Retire the always-on corner panel it replaces, and put a single prompt in its place that only nags while somebody is stood idle
• R cannot fire through the sheet — the first press puts the ledger away, the second sends them in
• cc.Ledger 1 / -CCLedger holds it open for the screenshot harness, which skips the opening and would otherwise never see it
• The rest of the HUD held to the same standard — it is still a debug readout

A bottom-of-screen bar, switching by phase. Build tools during prep, up to eight active skills during a wave. The mode change is the phase change, so the bar tells you which half of the game you are in — and [X] flips it by hand any time (2026-08-13, user): a wall breached mid-wave is rebuilt mid-wave, priced in fight-time.

• Bottom hotbar, two faces on one number row: build face (buildables with costs, red when short) and fight face (swing + empty skill sockets); auto-follows the phase, [X]/left shoulder flips it manually
• Mid-wave building unlocked. CanBuild allows the wave phase; kept honest by plans costing labour, builders refusing contested plots, and repo agents dismantling half-built work
• Eight skill slots per hero (the sockets are drawn; the tree is a taste session with the user)

Placement should be a blueprint, not an instant building. You place a translucent ghost of the thing, and builders walk over and raise it. That turns placement into a timing decision and gives the Builder job a purpose beyond repairs. (Already wanted in Phase 3; the playtest independently asked for it, which is a good sign.)

• Place a blueprint (surveyor’s plot: stakes, pale silhouette, progress post); builders raise it, the hero can hammer it up with the hoe, and a plan nobody can reach expires and refunds itself (deadlock guard)
• The plot matches the building (2026-08-13): the silhouette reads its size from the real footprint — it was drawn sideways and half-size
• Plans cannot stack (2026-08-13): placement — player ghost AND the harness’s direct path — rejects any footprint overlapping an unraised plan (oriented-box test, 45° rotations respected)
• A finishing building evicts anyone stood on the plot (2026-08-13): pawns inside the new footprint are nudged to the nearest free edge instead of being entombed
• Half-built things can be half-destroyed (2026-08-13): damage undoes raising progress at the finished building’s health exchange rate; repo agents put half-built plans on their docket
• Builders flee-and-retry (2026-08-13): a plot with an enemy within 500 units is refused, not braved; the expiry clock pauses while contested so camping collectors cannot force the refund
• The ghost is the building (2026-08-13, user): the placement preview is the actual silhouette – built from the same DescribeSilhouette block specs the real building raises itself from – in translucent green/red (engine’s M_SimpleUnlitTranslucent; DebugEditorMaterial renders opaque, don’t go back to it)
• Remove tool (2026-08-13, user): build-face slot 4; [E] tears down the aimed turret/fence/wall for DemolishRefund (80%) of its price scaled by remaining health; unraised plans cancel at 100%. Gamepad reaches it via the same d-pad cycle
• Playtest round 2 fixes (2026-08-13, live): tool keys auto-flip the hotbar (dead keys read as broken placement); chop counter + gather cap on the hero’s LIFETIME log tally (stock counting = miller treadmill / plank laundering); hold-R 2s to skip the opening (accidental tap threw it away); bite-size checklist (one task + N of 9); the Elder (stick, slow, [F] talk, placeholder story lines); tree spots reserved (stump regrowth grew a tree through a house)
• Sub-boss at wave 5 of each chapter (2026-08-14): THE AUDITOR — 1/3 boss weight, slow, clipboard silhouette, herald on the morning HUD, kill XP 50×chapter, no grudges. Measured: pre d8/d8/d9 → with him d11/d10/d10; the band held
• Villagers clamber over fences (2026-08-14): wedged workers vault fences (~1.2s scripted arc, occupancy-checked landing, fences only — walls and collectors still hold); 7–38 clambers per harness run
• Repair orders (2026-08-14): peacetime builders (and muck-in idlers) patch damaged fences/walls/turrets nearest-first, a plank per work tick; wall damage tint refreshes on repair
• Readable reserved ground (2026-08-14): visible cut stump + soil border, sprout once seeded; screenshot-verified at ground level via the staged grove_stumps photo vantage
• Hotbar flip on B as well as X (2026-08-14, both bound)
• Industry upgrades, not duplicate buildings (user): one mill + improvement tiers bought with village-side progression, shared access in co-op — the "shared kingdom tech" half of the double climb
• [F]/hold work verb — reconcile with the locked hold-to-act spec (one work verb, not two) in a design session
• Dynamic village layouts + build zones per seed — ties to the procedural map; parked for design

Villagers should not flee — they should muck in. Spare villagers with no job help build and repair; if there is nothing to build they help defend.

• Idle villagers default to helping build and repair
• Idle villagers join the defence when there is nothing to raise
• Threat follows damage. Firing turrets pull collectors first; only villagers who are actually fighting are ever targeted — bystanders carry no threat at all (the Authority repossesses property, not people)

Choose your hero before the run. One of the four faces, with their own colours and weapon.

• Hero select screen (2026-08-14): CHOOSE YOUR FARMER — four cards (Warden/hoe, Captain/sword, Ranger/bow, Steward/staff); the pick recolours the tunic and swaps the hand prop on the same body. Cosmetic shell until the skill tree; harness bypasses it entirely

Smaller things spotted

• The mill wheel spins inside the building. It should be mounted on an outside face, turning where it can be seen
• You spawn in an empty field with nothing in view to walk toward. The opening tells you to go to the granary and the granary is not on screen
• Bounce should lead. He is already built to be the waypoint; he is not yet wired to walk to the current objective and bark at it

The field was too cramped — buildings shoulder to shoulder, groves crowding the village, and nowhere to lay a defence out that was not already occupied. Widened from 64×44 metres to 84×62, grown sideways and eastward rather than along the approach: lateral room is where defences and buildings go, while stretching the walk to the granary would only hand the turrets free shots.

• Floor half-extents 3200×2200 → 4200×3100
• Village re-sited beyond the granary’s 750-unit panic radius — the shelter villagers run to is no longer inside the danger zone by construction
• Groves pushed to the corners, trees spaced 560×540 instead of 420×400, so a wood reads as trees you walk between rather than one dark mass
• Collector lanes spread to ±1300 — a wave on one narrow line does not need a layout to answer it
• The harness’s build slots are now offsets from the granary, so moving the village can never again strand its defence line behind what it defends
• Re-measured: ChapterHealthStep = 0.45 now ends the run on day 16 (was 14 on the old field), same shape — an early bite, a competent middle, boss pressure, then a four-day decline
• The extra ground is currently empty ground. Paths, crop rows and hedge lines to make the space read as a farm rather than as a lawn

The upkeep law. Everything you own costs something to keep.

The closed economy — the rule the rest of this phase hangs off

Killing collectors never pays the village. If enemies dropped money, bigger waves would fund bigger defences and the difficulty curve would flatten into a treadmill. Wealth comes from within — the farm, the mill, the mine — so waves are pure cost. Kills pay XP instead, per player, spent on that player’s own hero tree. Village power is farmed; personal power is fought for.

• Kills award XP to the player who earned them — never village currency

Resources that make resources. The second resource has to compete for the first, or it is just a parallel economy with its own separate sliders. So the smelter burns wood to melt iron: every ingot is a plank you did not get. Wood is already the contested thing, and iron pulls on it directly rather than adding a decision surface of its own. Food is what closes the population loop. More lumber camps need more villagers, more villagers need feeding, and feeding them means planting wheat and foraging berries into the granary. Without it, “recruit another villager” is a free win and the upkeep law never lands. Berries first — a cheap early source that needs no farm infrastructure and gives idle hands something to do. The chapter reward should be arable land, not just more trees, so clearing a chapter hands you the field you need to feed the villagers you are about to recruit. That is what makes “claim more land” mean something economically rather than merely spatially.

• Resources: wood, stone, ore, food
• Iron mine, unlocked with the map space a cleared chapter opens
• Smelter that burns wood to turn ore into ingots
• Wheat to plant and berries to forage, both feeding the granary
• Granary stores food; villagers eat from it
• Money as a build/repair currency
• Recruit villagers with gold — population is bought, not granted
• Villager wages — a recurring drain that scales with headcount, so a bigger village is a bigger commitment
• Warehouse building with a storage cap, so materials compete for space and expansion becomes necessary rather than optional
• Unlocks cost materials that are gated behind unopened map — stone and iron are reasons to push the frontier
• Job board shows a target headcount per job, not just the current one (e.g. 2 chop, 2 reseed — needed 4), so it reads as a staffing plan
• Replanting is a job — felled trees are re-seeded by villagers instead of regrowing on a timer, making forestry an ongoing labour cost
• Buildings: villager housing, food production, smithy, smelter
• Second mill / mill upgrade — the current bottleneck, and the first real reason to spend planks on a building rather than a defence

Food is measured in mouths fed, not calories. Berries feed one villager, bread feeds four, a stew feeds twelve. The tier gate is the point — berries cannot run a kingdom, so the food chain has to grow with the population — but counting calories would put a spreadsheet in a game whose pillar is depth without one. “Mouths fed” is the same mechanic, readable at a glance, and it needs no translation. Cooking is then a multiplier on labour: a meal made of three ingredients feeds more than those ingredients would separately, which is what earns the kitchen its place as a building rather than a second granary. Happiness is fed and safe. A village under attack is an unhappy village even if the pantry is full — and that is what finally gives walls a reason to exist on day one, before the player has been hurt enough to want them. Today a wall is nice to have. Tie production to safety and the perimeter becomes economic infrastructure: you wall up to keep the village working, not merely to slow the enemy down. A breach should cost morale after the wave is beaten too — the fear outlasts the fight, so repairing properly beats patching.

⚠️ This whole block is blocked on pathfinding, and it is the reason navmesh is now the biggest technical dependency in the project. Happiness-needs-safety makes walls mandatory; mandatory walls make gates mandatory; gates need villagers to route to them. They steer directly today, so they would walk into the wall beside the gate and sidestep forever — and they cannot simply live inside the walls, because the work is outside them. A walled village whose people cannot find the gate starves itself.

• Food measured in mouths fed; berries 1, bread 4, a stew 12
• Cooked meals feed more than their ingredients — the kitchen as a multiplier on labour
• Villager happiness = fed AND safe, with production following happiness
• Navmesh / real pathfinding ⬅️ the gate on everything below it
• Gates — the player walks through; collectors cannot, unless one is left open. Leaving it open is a speed-versus-safety trade, not a mistake
• Collectors can break a gate down, so a wall is a delay bought with planks rather than an ending
• A breach costs morale after the wave, not just health during it
• Injured villagers walk home, rest for a time, and need food to recover
• Simplified management — depth without spreadsheet micromanagement

The kingdom becomes defensible from every angle, and ammunition matters.

• Villagers can be assigned to militia when a wave lands
• Waves attack from any direction — the village needs 360° cover
• Wave 1 stays scripted from a fixed direction as the tutorial beat
• Late game: many smaller simultaneous waves from all sides, punctuated by a mega wave
• Wall tech tiers — wood wall first, stone wall at mid/late game, deliberately expensive to replace so upgrading a perimeter is a real investment rather than a free swap
• Build-by-villager — placing a defence drops a blueprint and builders walk over and raise it. Building becomes a job order, placement becomes a timing decision, and the Builder job matters beyond repairs. A hero skill may allow raising it yourself instantly
• Boss: the Tax Collector — swings a money bag. Ranged taxmen throw paper balls, staplers and briefcases; the Authority weaponises office supplies, which keeps the satire funny and the violence family-friendly
• Ammunition economy: arrows first, gunpowder and bullets later
• Tech arc from hoe to howitzer

One climb, not five parallel ones. The village grows skill sets, the land carries the resources those skill sets need, and each answer is also the next question.

Every solution creates the next problem. A progression system that hands out solutions to problems the player does not have yet is a menu. One where each answer is also the next question is a game. The chain, which is also the teaching order — one new idea per chapter, so it arrives alone and can be read: 1. Lumber → planks → a basic turret. ✅ built. The turret is the answer. 2. The turret stops being enough. It wants metal, and there is none here → clear a chapter and open the land the mine sits on. 3. The mine needs a smelter, and the smelter burns wood. Every ingot is a plank you did not get → forestry has to grow: better axes, a lumber camp. 4. Growing forestry needs hands → more villagers → villagers must be fed → arable land, which is the next chapter’s reward. 5. Fed villagers must also be safe → walls and gates → pathfinding. 6. A metal economy unlocks the smithy, then the weaponsmith, then turret tiers — and the ladder starts again a rung up. Hoe to howitzer. The rule that stops this becoming a spreadsheet: an upgrade should change a decision, not just a number. +10 damage is connective tissue; pierce, a slow, or a reach that changes where you site the thing is content. A little glue is fine — a whole tier of glue is padding, and it will be felt. The constraint that protects it, already locked: one processing step per resource. Trees → logs → planks, and no third hop. Depth comes from resources competing for the same inputs — the smelter burning wood — never from a longer chain. That is the line between an economy and an admin screen.

• Build ONE rung end-to-end first — turret tier 2, gated behind iron, gated behind a cleared chapter — and prove the whole loop on that single vertical slice before widening it. One rung that teaches itself beats five that are stubs
• Buildings imply jobs: a smithy implies a smith, a weaponsmith implies a weaponsmith. New skill sets arrive as buildings, not as menu entries
• Upgrade tiers for turrets, forestry, buildings and the town, each costing more than the last
• Every tier’s unlock is taught by the problem that precedes it, never by a notification

A run is resumable, never rewindable — the roguelike ironman model.

• One auto-save slot per run, written at each day boundary. Quit any time, come back and continue — a long run must not demand a single sitting
• On defeat the run is closed out and the score committed. You cannot reload to before a death; that is what makes a days-survived leaderboard mean anything
• Solo can pause. Co-op cannot meaningfully pause — at most a host pause during prep
• Leaderboard integrity — offline save-copying is always possible on PC. Server-side run validation or a separate verified board solves it, but not until competitive play is real. Everything is test-and-wipe for now

• Procedurally generated map, seeded
• Resource placement tuned to match the gameplay pacing
• Reclaim the map cell by cell; frontier tension as you spread
• Capped map size → full reclaim triggers Endless mode

• Four hero faces: Warden 🌱 · Captain 🛡️ · Ranger 🏹 · Steward ❤️‍🩹
• One open classless skill tree fed by universal XP
• 100+ levels; score = days survived; leaderboards
• Co-op up to 4 — shared kingdom tech, personal hero trees
• Radius abilities that make a role felt in the world — e.g. the defender strengthens every wall near the hero
• Free respec — points can be spent, reset and re-spent, so a build is a stance rather than a life sentence
• Passive line: movement speed, health, work efficiency
• Aggro is a mechanic — step in front of your towers and swing, and the taxmen come for you. A defensive skill that deliberately pulls aggro and fortifies for a few seconds makes survivability and the support role genuinely valuable rather than decorative
• Hard cap on allocatable points, soft infinite levels. The cap is what keeps a build a set of choices forever. Levels continue past it and give small paragon-style increments that feed score and prestige rather than the tree, so "level 99k" stays meaningful without dissolving identity

Role exclusivity — resolved. Everyone can do everything; roles are levels of efficiency, not permissions. A technology or skill choice makes you better at something, never the only one allowed to do it. The gate is the loadout, not the unlock. In an endless game a solo player will eventually unlock everything, so unlocks alone cannot create identity. Instead: unlock freely and infinitely, but keep only four active skills equipped at once, freely swapped and combined. Identity comes from what you have equipped right now. Skills also upgrade and branch as you progress — infinite ways to be defeated, so infinite tools to keep going.

Decided early, while it is still cheap.

• Renderer switched off Lumen and virtual shadow maps — the blocky art gains nothing from them, and they were what would lock out Switch and punish Steam Deck
• Full gamepad parity on movement, combat and build controls
• Gamepad coverage for the job board (partial: D-pad assigns, job select is keyboard-only) — required for console certification
• Resolution-independent UI with title-safe margins — the canvas HUD uses fixed pixel positions and will not scale to a Switch handheld or a 4K TV
• Camera polish — the spring arm still collapses toward the hero near tall buildings; needs a proper minimum distance or fade-out
• Steam Deck verification pass
• Mac and Linux builds (near-free in Unreal, worth doing before launch)
• Console ports — gated on developer accounts, dev kits and certification, not on code. Revisit once wishlists prove demand.
• Crossplay — if wanted, multiplayer must be built on Epic Online Services rather than Steam sockets. Nothing is networked yet, so the door is open.