Development covered 25 Jul 2026 to 27 Jul 2026 (commit dates).
Three small activities landed this stretch: swinging a pick at a spirit vein, holding your concentration through a sweeping meter, and sorting a tray of goods by eye. Each took a couple of days to build and considerably longer to balance, and they all ran into the same wall, which turns out to be the central problem with putting an active pastime inside a game about not being active.
The reward curve has to bend down
The tension is easy to state. If a clicking activity pays well, the optimal way to play an idle game is to not idle, which insults the entire premise. If it pays badly, nobody touches it and you built a room nobody enters.
What I landed on is a reward that is generous at the very start and fades as you advance. At the first realm, a few minutes at the vein is worth a meaningful fraction of what a quest pays. Six realms later it is worth a rounding error. The activity stays available and stays pleasant, and it stops being an answer to the question of how to make progress.
That shape does two useful things at once. Early on, when a new player has nothing to spend and no quests unlocked, there is something to do with their hands that pays enough to feel worth it. Later, when the economy is real, the main loop is unambiguously where the value is. Nobody has to be told to stop mining. The numbers say it politely.
A floor for newcomers made the early half work: a guaranteed minimum for the first several swings, scaled by realm, so the first experience of the activity is never a run of nothing.
Seven payouts in a row of exactly one
Real play immediately found something the maths had hidden. A tester reported seven swings in a row paying exactly one stone.
The average was correct. The distribution was flat, because the payout was drawn from a pool in a way that collapsed to the same value under common conditions. Average correct, variance nearly zero, and the result feels less like a game than a subscription. A reward with no spread does not read as a reward at all; it reads as a tax refund.
The lesson I would state generally is that the average is the least interesting property of a random reward. What a player experiences is the run: the streak of nothing that makes the next hit land, the occasional result well above the others. Design that shape deliberately and the mean will follow, because the mean is a consequence and the shape is the experience.
Two related legibility fixes came out of the same session. The rock's health had to be readable as a quantity, so a swing visibly takes a chunk out of it, and the reward had to be visible at the moment it is won rather than quietly added to a total somewhere. An unseen reward is not a reward.
The sorting game had to teach itself
The third activity is a sorting puzzle built from goods drawn out of your own realm, which gives it a quiet second job: it is how a player learns what the objects in their world are, without a tutorial screen.
Two things came out of testing it on a phone. The first level has to be a demonstration rather than a challenge, because the rules are learned by winning once. And a board on a small screen needs its own layout, not the desktop layout scaled down, which is obvious and was still not what I built first.
One more thing about that puzzle, more of a warning. Three genuine faults in it were found by rereading the code rather than by playing, and none of them would have shown up in a playthrough. Some classes of bug are invisible to testing because they are about what the code can produce, not about what it did produce this time. Sitting down and reasoning about the whole space of boards is not old-fashioned; it is the only method that covers it.
The frame was chosen by its name, not by what it draws
Last one, because I made the same mistake three separate times and want it written down.
Artwork in this project is looked up by name, and it is very easy to pick a piece because the name matches the concept. A place called the Frozen Stone Port should stand on snow. A forest of whispering bamboo should use the frame called forest wall.
Except that particular frame draws a dark green ring, so four of them made a forest of hoops. A tile named for farmland turned out to paint almost none of its own square, so ten of them read as loose grey blocks on grass. A piece named for a mountain peak was in fact a horizon backdrop with opaque edges, which put a rectangular card on the hillside.
Names describe intent. Only the picture describes the picture. The only reliable fix has been to look at every frame I choose, at the size it will be seen, before believing what it is called.
Everything here ships in Cultivation Game.
Top comments (0)