A colony sim withholds the one thing every other strategy game hands you: direct control. You rank priorities, designate work and draw the plans, and your colonists choose who does what, in which order, and whether the mood they are in permits any of it. The base is not the output. The account of how the base went wrong is the output.
The interface is a grid — colonists down one side, kinds of work across the top, a ranking in each cell saying how eagerly that person picks up that job. You never issue an instruction. You edit the table, mark a wall for building or a corpse for burial, and the colony re-sorts itself around your edits. A surgeon who also cooks will keep cooking while someone bleeds, if that is how you ranked her.
Then you unpause and watch your org chart collide with reality. Someone is hungry, someone is grieving, the harvest came in early and there is nowhere dry to put it. The rhythm is plan, run, interrupt: pause is your most-used control, because the sim raises a crisis and you freeze the world to re-plan around it. Needs pile up faster than hands, and that gap is the game. The one exception is combat — when raiders arrive and you draft colonists to hold a doorway, the game briefly hands you direct control and borrows its verbs from turn-based tactics.
Indirect control only produces stories if consequences decompose into reasons. A RimWorld colonist who breaks down and starts smashing the furniture has an itemised list behind the outburst: slept on the floor, ate without a table, saw a corpse, spent a week in a room with no window. Read the list and the disaster becomes something you helped write. Hide it and the same event is arbitrary, and nobody retells an arbitrary event.
That is also why the good ones simulate fewer things, connected more tightly. Temperature, food, mood and labour sharing one world produce anecdotes; a hundred systems in their own silos produce a dashboard. Dwarf Fortress, the wellspring, was played for years through an ASCII interface because what it modelled mattered far more than what it showed, and its motto — losing is fun — is still the genre's founding attitude.
Left to itself an emergent sim paces badly: long placid stretches, then three catastrophes at once. RimWorld's answer is the storyteller, a pacing AI that watches your colony and decides when you have been comfortable for too long. The default is Cassandra Classic, who builds threats towards a climax and then eases off; Phoebe Chillax spaces the same events further apart, and Randy Random ignores pacing altogether. Choosing one is choosing a difficulty curve and a narrative shape in the same click.
The neighbours need care here, because the storefront tags blur them. A management sim — a hospital, a theme park, a coffee shop — shares the indirect control, the priority tables and the staff with moods of their own, but not the total-loss state: going broke embarrasses you, whereas a colony that runs out of food is over. That missing ending changes what every other system is for. It is also why a city builder can abstract its people into population graphs and a colony sim cannot, and why the cozy farming sim reads as this same daily stewardship with catastrophe switched off.
Every colony sim asks you to learn a small enterprise-resource-planning system before it will show you its best material. Stockpile rules, work tabs, bills that repeat until a count is met, filters deciding which meat goes in which freezer: the tutorial for the genre is a spreadsheet, and the doctor who slept through the fire is on the far side of it. Most people who bounce off these games bounce off the tables, not the stories.
Scale makes it worse. A large colony has more of everything except interesting decisions — the crises repeat, the babysitting thickens, and managing thirty people is not three times the game that managing ten was. The genre's most useful lever is the one it reaches for least: letting you delegate away the work you have already proved you understand.
Put the control scheme in the first line, because it is the line an AI collaborator will otherwise get wrong:
Type: Colony sim (indirect control; a management sim is this without the total-loss state)
Core loop: Rank priorities, designate work, pause on the crisis, re-plan, unpause
Perspective: Top-down or isometric, zoomable from the whole colony to one bedroom
Win / lose: Open-ended survival; a wiped colony is the only real ending
Session: Hours — the story needs time to accumulate
Progression: Research and expansion, plus the debts your decisions leave in people
Then judge every system you add by one test: can a colonist's worst week be explained by it, out loud, to somebody who does not play these games? A number that only ever moves another number will never be retold.