Milestone 27 didn’t add any new gameplay systems to Project Nightfall. Instead, it built something arguably just as important for an in-development strategy game: the first real map to run everything built so far — economy, AI, and combat — on the same battlefield at the same time.
Until now, each system had its own isolated test scene. This milestone combines them: two mirrored home bases, three resource types (Gold, Lumber, and Moonstone), a wall obstacle splitting the map down the middle, two matching expansion resource clusters, and a neutral cluster sitting exactly between both bases. The map is symmetric by design — whichever side the AI or the human plays, the terrain and resources are identical, so neither side gets a built-in advantage. A toggle even lets the AI take either corner, or lets both corners be AI-controlled at once for unattended playtesting.
Two real bugs, caught two different ways
Putting all these systems on one map for the first time immediately surfaced a genuine problem that no isolated test scene could have caught: matches didn’t actually end when one side’s entire army was destroyed. The cause was architectural — buildings can’t currently be destroyed at all in Project Nightfall, and the match’s “is a player still in the game” check counted a surviving building as presence, even with zero units left. Since the AI reliably builds at least one structure early on, wiping out its army alone could never end the game.
Rather than quietly patch around it, the fix was scoped specifically to this map: a units-only win condition, used only for this prototype, so a full army wipeout now ends the match. The existing building-based logic used elsewhere in the project is untouched. This is an explicitly temporary trade-off — a base that loses every unit for a brief instant before its next recruit finishes could, in theory, be called defeated a moment early — and it goes away entirely once building destruction is implemented down the line.
A second issue was caught earlier, before the map was ever manually played: an early version of the resource nodes reused old capacity values too small to fund a full match, which would have run dry mid-buildout. The map now ships with its own larger resource pools sized for an actual game.
Why this matters
A prototype map like this is a stress test disguised as content. It’s the first place bugs in pathfinding, AI behavior, and win conditions can surface under realistic conditions rather than narrow unit tests — and this milestone already caught two real problems, one during actual play and one before it ever launched. The known gaps are tracked openly: buildings still can’t be destroyed, and the AI doesn’t yet expand into unclaimed resource areas. Both are expected limitations at this stage, not surprises.
With this map in place, the project has, for the first time, one battlefield where a complete 1v1 skirmish can actually be played start to finish.