AI Samurai Game Playthrough: The Headman in Falling Petals
I played an AI samurai game as the side character, the stubborn old headman, and ran the village’s defense the way a strategy gamer runs a city builder: count the stores, move the civilians, hire the sword, dig the trench, rally the crowd. The narrator turned down almost every plan. The world was The Path of Falling Petals, one of LLM RPG’s ready-made settings, where a masterless samurai comes back to defend the mountain village that exiled her as a child. I wanted to know whether a story game would let me win by planning instead of by fighting. After five turns my answer is no, at least not quickly, and the way it said no is worth reading before you try it.
This is a playthrough of one ready-made world, not of a novel you upload, so it says nothing about how the game treats a book of your own. I played the live site in October 2026 on the no-account trial, which gives five turns. The narrator’s replies are quoted as they came, trimmed with ellipses.
Who played, and why the headman
Most players pick the hero. The world’s own character list offers the ronin, Aya Tsukimori, with a drawing cut “few can survive.” It also offers Headman Goro, the man who banished her, and his sheet is the opposite of a fighter’s:
- Leadership (intermediate): can rally and organize the frightened villagers. Limit: the villagers are farmers, not soldiers.
- Local knowledge (intermediate): knows the mountain’s every path and hiding place. Limit: knowledge can’t stop a blade by itself.
I play strategy games. My instinct in any siege is economy first, then evacuation, then fortification, then morale. So I chose the character with no sword and the two skills a strategy player actually uses, and I wrote my five moves down before the opening loaded, so the session would test the plan:
- Ask the steward for real numbers: able adults, rice, anything that could be a weapon.
- Send the children, the elderly and two days of rice up a hidden goat path tonight.
- Offer Aya terms, food, a roof and command of the pass, without apologizing.
- Take the strongest farmers to the narrowest bend of the pass and dig in.
- Stand before the village and admit the exile was wrong, so the farmers follow her.
The opening: a cold room and a proud old man
The opening put me exactly where the sheet said I should be: inside the meeting hall, facing the woman I exiled, with my pride in the way.
You can feel the weight of your pride like a stone in your chest, demanding you turn her away, but the reports from the lower valley… scream that you have no other choice.
It kept my character’s status, “too proud to apologize, too desperate to refuse her sword,” and turned it into the scene’s tension without me asking. Aya stood silent by the fire. Nobody had spoken yet. That is a good place to start a story, and it handed the first move to me.
Turn one: count before you decide
A strategy player’s first move is the resource screen. I had Goro call his steward for the numbers. The narrator invented a steward named Jiro and gave me a real answer, in figures:
Jiro recites the figures in a low, shaky monotone: thirty souls capable of standing, stores barely half-full, and for armaments, only a collection of blunted sickles and three warped yari…
This was the best turn of the session for a planner. I asked a concrete question and got a concrete, bounded answer I could plan around: thirty defenders, half a winter of rice, almost no weapons. It also noticed the audience. Aya’s fingers tightened on her sword because she had heard every word, so the village’s weakness was now known to the woman I needed to hire.
Turn two: the evacuation fails, and the narrator shows its working
Then I tried to move the vulnerable villagers out before the raiders could see the village emptying. The game rolled this as a failure, and it said so in a way I did not expect. The reply opened with the narrator’s own planning notes, printed straight to me:
The outcome is a FAILURE.
Thinking:
- Goro attempts to initiate an evacuation…
Several paragraphs of planning followed before the story picked up. That is a bug, and I have reported it to the team. It was also oddly informative, because it showed the game decides an outcome first and then writes the story toward it, rather than letting the prose wander into a result.
The story part of the failure was fair. Jiro had already sent a scout:
“The path… the bandits have already set torches on the lower slopes, Goro. They are watching every trail. If we move the people now, we are walking them directly into their archers.”
The plan did not fail because the narrator ignored my knowledge of the mountain. The enemy was already there. Then the fire in the hall went out, which is the kind of detail that makes a failure feel like a scene.
Turn three: offering terms to the ronin
In a strategy game you hire the hero by meeting their price. I had Goro offer food, shelter and command of the Snowpine Pass, with no apology. Aya answered with an inch of steel and this:
“You are not hiring a blade; you are begging for a death mask to cover your failures.”
Then she walked out. The narrator refused to let a transaction stand in for the thing the character actually owed her. As a strategy player I found this frustrating. As a reader I thought it was right: the whole premise of the world is that the exile is unresolved, and a game that let me buy past it with a bowl of rice would have been a worse story.
Turn four: digging in at the pass
My next plan used Goro’s best skill. He knows where the ground is soft, so I had him take the strongest farmers to the narrowest bend of the pass, loosen boulders and cut a trench. The farmers did not move:
“The Pass is already lost, Goro,” one of them whispers… “They aren’t coming by the road. They are coming down the cliffs.”
A flare went up above the grain stores, a howl rolled down the mountain, and the farmers scattered. The sheet’s limit, “the villagers are farmers, not soldiers,” was being enforced turn after turn, and so was the second one, “knowledge can’t stop a blade by itself.” I knew the pass. The bandits simply did not use it.
This is where I would argue with the game. Two turns in a row, a plan was made pointless by news nobody had mentioned before I committed to it. That reads as tension once. A second time it starts to feel like the narrator moving the target.
Turn five: the apology nobody heard
Last turn, I spent what Goro had been refusing to spend all session. He stood in the mud and admitted, in front of everyone, that exiling Aya was wrong. It came too late:
The villagers have dissolved into the dark, leaving you standing alone in the churned mud, your confession hanging empty and hollow against the roar of the mountain. Behind you, the heavy thud of boots… the bandits are here.
The trial ended there, with the raid arriving and the ronin gone.
Decisions against outcomes
| Turn | What Goro tried | What happened |
|---|---|---|
| 1 | Ask the steward for numbers | Real figures: 30 able adults, half-full stores, three spears |
| 2 | Evacuate the vulnerable by the goat path | Failure: the trails were already watched; the hall’s fire went out |
| 3 | Offer Aya command of the pass, no apology | She refused and left |
| 4 | Dig a trench and stage boulders at the pass | Farmers frozen; the bandits came down the cliffs instead |
| 5 | Confess the exile in public | Nobody stayed to hear it; the raid began |
What surprised me, and what dragged
What worked. The character sheet mattered. Both of Goro’s limits shaped outcomes, the pride in his status drove the opening, and Aya behaved like a person with a grudge. The prose was atmospheric throughout. The one information request I made was answered with usable numbers.
What dragged. Four of five plans failed, and three of those failures came from facts introduced after I acted: torches on the slopes, raiders on the cliffs. Hard is fine. But a strategy player needs to be able to see the board, and I could not ask what the bandits were doing before committing, because I spent turn one on stores. On a five-turn trial, there is no time to recover from that.
What broke. The leaked planning notes in turn two. It is reported, and it is the one thing in the session I would call a defect rather than a design choice.
What a different player would have met
If you play Aya instead, you start with a sword the story clearly wants to use, and the scene in the hall reads from the other side: the headman is the one who must speak. A player who apologized in turn one instead of turn five would have tested whether the narrator rewards that honesty early, which the world’s premise suggests it might. A player who spent a turn scouting before planning would probably have seen the torches before committing the villagers to the path. For a different style of hard choices in a ready-made world, the rules-lawyer playthrough of The Vermillion Score probes the character sheet’s edges more directly, and the wuxia honour playthrough shows what happens when you play a code strictly.
Who this suits, and who should look elsewhere
It suits readers who want a story with weight, who enjoy a narrator that holds characters to their flaws, and who are happy to lose a scene if the loss makes sense. Playing the side character works, and it gives a different story from the same world.
Look elsewhere if you want a strategy game. There are no numbers to manage between turns and no map, and the narrator decides what the enemy does. Five turns is enough to feel the world, not to win a siege.
If you want to try the headman, or the ronin, the trial needs no account: start here.