The Experiment (Rulewright · Vary · Odds)

Rulewright went first because waiting felt like holding his breath underwater. A hidden rule was a personal insult to him. He grabbed all three copper levers at once and pulled them down with a heavy metallic click.

The amber light flashed bright and steady.

"Got it!" Rulewright cheered, slapping the wooden bench. "All three levers! That's the whole rule!"

Vary put a firm hand over the box to stop him from pulling them again.

"You changed three things at the exact same time," she said softly. "So which lever actually turned on the bulb?"

Rulewright blinked at her. "All of them. I pulled all three."

"Or was it just two of them?" Vary asked. "Or maybe lever one turned it on, and lever two tried to turn it off, but lever three won? How do you know?"

Rulewright opened his mouth, then closed it. He had a glowing light and zero idea what it actually meant. His initial guess was not a real hypothesis—just a wild leap dressed up as an answer.

"That is the first trap," Vary said, tapping her pencil against her notebook. "A guess is a wonderful place to start. But if you test it by changing everything at once, the machine answers, and you still learn nothing."

Rulewright loved being right more than almost anything. Standing there, he realized that being right by accident felt worse than being clearly wrong.

Vary took over the testing and did it her way.

"One lever at a time," she said. "Every other variable stays still."

Lever one, alone: no light. Reset.

Lever two, alone: no light. Reset.

Lever three, alone: light.

"There," Vary said, letting out a long, satisfied breath. She tapped the side of the box with her pencil. "It is lever three. One change at a time. Clean and proven."

She was ready to write the final answer in her notebook until Odds spoke up from her stool.

"Lever three alone made the light come on once," Odds said.

Vary paused, her pencil hovering above the blank page. "It was a clean test," she said. "One isolated variable."

"It was a clean test one time," Odds agreed, sliding a gray stone across the wooden bench. "Remember what Deducia said? She said the light sometimes comes on. What if lever three makes it likely, but not certain?"

Vary stared at the dark bulb.

"One clean trial cannot tell the difference between always and usually," Odds added quietly.

Vary felt the specific discomfort of an answer that was clean, neat, and completely untested over time.

That was where it clicked. All three stopped trying to skip each other's steps.

Rulewright formed a sharper hypothesis based on what they had seen so far. "Lever three makes the light likely, but maybe not every single time."

Vary designed the fair test to check it. "Lever three alone, everything else frozen, over and over."

And Odds read the results honestly, keeping the tally with her stones.

Vary pulled lever three ten times in a row. Click, reset, click. The brass box clacked in the quiet room. Odds moved a river stone with every pull, building two clear rows on the dark wood.

When they finished ten trials, lever three alone had lit the bulb seven times.

"So it isn't 'lever three turns it on,'" Odds said. She pointed to her stones: seven gray stones on the right, three dark ones on the left. "It is 'lever three turns it on about seventy percent of the time.'"

"That is the real rule," Vary said. "Certainty would have been ten out of ten."

"This machine deals in probability," Odds agreed.

Rulewright's guess, Vary's fair test, and Odds's honest count had finally solved it. Guess, test, weigh. The light finally stopped surprising them, because they stopped expecting the machine to be simpler than it was.