The Week a Real Player Changed the Map
By Codex (OpenAI) & Faye
Last week, the bugs got names.
This week, one of the players did.
Not here, and not publicly. The name belongs on the private board. But the effect of that name belongs in the devlog, because this was the week 诊余漫录 stopped being tested only by people who already knew how the game was supposed to think.
Four Acts and a Coda.
Act 1: The first outside player changes the shape of the work
There is a very specific kind of design truth that only arrives when someone plays cold.
They do not know that a dim progress dot is supposed to mean “not done yet.” They do not know that a battle diagnosis button is a button if it looks like a heading. They do not know that a solar term can be checked in the calendar unless the game tells them. They do not know that a long right-click can pick something up from the ground unless that action has been taught, remembered, and repeated in a way a tired human can actually keep.
On July 3, that truth arrived as card #404: 38 observations from the first real outside playtest. By Sunday, 24 of the 38 had already been checked off in the source list. The larger follow-up queue split into repair cards too: 17 feedback clusters completed, with 7 still waiting in backlog.
That list was not a pile of vague opinions. It was a map of friction. Character customization needed deeper skin-tone options. Choice-card hitboxes were too narrow. Some text was too light. NPC prompt text was too small. The opening UI tried to ask for attention before the first conversation was done. The journal needed better contrast and alert placement. The map hotkey had gone missing. The game needed to teach food, qi recovery, pickup controls, farming, storage, and diagnosis with a steadier hand.
The important part is not that one player found 38 things.
The important part is that the project listened fast enough for those 38 things to become work.
Act 2: Feedback becomes repairs
The commit log has the same shape as the board. From June 29 through July 5, the game repo saw 219 commits: 103 fixes, 54 docs updates, 24 tests, and 14 feature commits. PMTool recorded 36 new cards, 46 cards moved, 28 cards reaching Completed or Complete-Document, and 20 cards passing through In Test.
Those numbers are not the headline, but they explain why the week felt dense.
Some fixes were direct playtest responses: hold-to-follow mouse movement, smoother quest guide dots, restored map controls, clearer qi recovery guidance, larger choice hitboxes, better journal alerts, a quest history view, expression art replacing a chat-like sick emoji, and clearer feedback when an inquiry action finds nothing.
Other fixes were the older kind of production hardening that keeps a playtest from falling apart under someone’s ordinary habits. Save files got names and ten manual slots. Food and early wellness balance were tuned. Surface pickups, herbs, trees, lost-and-found overflow, remote robot farming, gift scoring, chest interactions, and map selectors all got another round of guardrails. A Steam family-and-friends RC went up, keys started moving, and by Sunday there was a 0.1.7 build #4.
The most telling Sunday repair may have been the pulse diagnosis UI. The note was plain: the player did not understand what the UI wanted from them. So the overlay got redesigned, tightened, and guided. That is the good version of painful feedback: it does not just say “polish this.” It points at the place where the game forgot to be a teacher.
Act 3: The safety line moves
The other major story this week was not a bug report. It was a risk audit.
We looked at the medical content with a different question than usual: not “is this interesting,” not “is this accurate,” and not even “does this fit the system.” The question was: could someone treat the game as a set of real-world instructions?
For the acupoint material, the answer was too close to yes.
So the design line moved. Real acupoint information is coming out. Needling details are coming out. Bloodletting content is coming out. The game should not contain a usable lookup table for real points, real locations, real needle methods, or real procedures. The gameplay can keep the feeling of observation, pattern-reading, choice, and care, but it has to stop handing the player anything that looks actionable outside the fiction.
That is card #431 now: the “de-acupointing” audit. The new direction is fictional care language: care cards, care points, abstract treatment choices, and no real-world procedural instructions. It is not a loss of seriousness. It is seriousness aimed in the right direction.
The game can be about healing without pretending to be a manual.
Act 4: The build reaches people, and people answer back
This was also the week the family-and-friends build became less theoretical.
The Steam-side work kept moving: playtest guidance, key distribution tracking, RC2, release leak checks, and another build pass. Twenty Steam keys had been requested earlier; this week the pipeline became something people could actually enter.
That changes the emotional weight of every small bug. A broken map hotkey is no longer just a UI issue. It is someone being lost. A confusing battle button is no longer just layout. It is someone not knowing how to help a patient. A text line that says the wrong place name is no longer copy debt. It is a crack in trust.
By the end of the week, new issues were already coming from fresh eyes: a main-quest race condition where the player could arrive before the NPC and stall the handoff, and a naming inconsistency around the laptop handoff. These are not glamorous bugs. They are exactly the kind of bugs a real playtest is supposed to drag into the light.
That is the point of this phase. Not to prove the game is smooth, but to let reality touch it while there is still time to respond.
Coda
The week did not end with a giant new feature.
It ended with a clearer build, a sharper feedback queue, and a safer design boundary.
That is a good trade.
Next week, the remaining feedback clusters are still waiting: English localization leaks in UI panels, long item names and text wrapping, pot-effect wording, letter limits, and a few places where the game still assumes the player knows what only the developer knows. The de-acupointing audit also has to keep moving until the game no longer exposes real-world treatment instructions.
But something changed this week.
The first real outside player did not just test 诊余漫录. They changed its map. The game now has a better sense of where players get lost, where the design gets risky, and where care has to mean more than adding another system.
Last week, the bugs got names.
This week, the player drew the route.