Before the counting begins, the prisoners get one meeting in the courtyard. They can agree on a plan, then they return to solitary cells, unable to see who enters the central room or what the light is doing. The bulb starts off. Each day, the warden chooses one prisoner at random to visit it. Somewhere inside that small room, a shared memory must take shape.
The rule is severe: one prisoner may announce that everyone has visited, but a false announcement means all 100 are shot. A true one frees them. So the puzzle is not asking for a good guess, or even a guess that will probably be right. It asks how separate people can build proof from a signal no one else can watch.
The bulb is the only record they share. It cannot name the person who changed it, and no prisoner sees the room on anyone else’s turn. A plan has to make a small change carry a precise meaning, and keep that meaning intact over visits that may be separated by many days.
That makes the puzzle feel less like a trick with a switch than an exercise in keeping an agreement. The prisoners do not need to remember every visit. They need a rule that turns some visits into countable evidence, and a way to know when the count is enough.
The setup and the rule that makes it hard
There are 100 prisoners, each in a separate cell, and a central living room with one bulb. No one can see the bulb from a cell. The warden picks a prisoner at random each day, with every prisoner equally likely to be chosen. The visitor may change the bulb’s state, but cannot leave a note or speak to anyone.
The group can meet once before the process starts and agree on what the signal means. Afterward, they cannot compare observations. Any prisoner who enters the room must follow the plan using only what they can see then: whether the bulb is on or off, and whatever they remember about their own visits.
This is why “everyone has probably been there by now” is not an acceptable conclusion. The puzzle attaches a fatal cost to a false claim. A long wait may make it more likely that all 100 have visited, but likelihood is not proof. The question is whether the plan gives someone grounds to be certain.
A leader’s count, built from one-time signals
One cautious plan, attributed to Denotkine on the puzzle page, appoints one prisoner as leader. The other 99 prisoners each turn the light on exactly once, the first time they enter and find it off. After using that one signal, they never turn it on again. The leader is the only prisoner who turns the light off.
When the leader enters and finds the bulb on, the leader turns it off and adds one to the count. Each such count corresponds to a signal from a non-leader. Because every non-leader is allowed to send only one signal, the leader cannot count the same person twice. Once the leader has turned the bulb off 99 times, all 99 other prisoners must have sent their signal. The leader has also visited, so all 100 have been in the room.
The plan relies on a rule simple enough to survive a long stretch of silence: a non-leader uses an off bulb to send one signal, and the leader clears that signal by turning the bulb off. The count is not a record of every visit. It is a record of distinct prisoners who have managed to report at least one visit.
A second scheme and a larger threshold
The same page gives another scheme, attributed to Winkler. Here, Alice is the leader. Each other prisoner turns the light off the first two times they find it on, then leaves it alone. Alice turns it on when she finds it off and counts those visits. The quoted threshold is 2n - 3, which equals 197 when n is 100.
That threshold is striking beside the 99 signals in the first plan. The two methods ask the prisoners to behave differently, and Alice’s count must account for the possibility that some prisoners have sent more than one signal while others have sent none. The number is presented as the point at which she can claim certainty.
But the page does not provide the reasoning behind 2n - 3, so the formula alone is not a proof for a reader trying to check the scheme. It is worth keeping that distinction visible: a proposed rule can sound exact, and a stated threshold can look authoritative, while the explanation that makes it trustworthy is still missing.
Why certainty changes the question
A probabilistic answer mentioned on the page suggests waiting three years and then announcing, with great probability, that everyone has visited. That may sound sensible if the goal is to make a very good prediction. It does not meet this puzzle’s stated condition. Even a small chance of being wrong remains a chance that the entire group dies.
So the clock is not the evidence. More days give more opportunities for every prisoner to be selected, but the passage of time does not tell the prisoners which individuals have visited. A guaranteed method must make the bulb’s history rule out every possible case in which someone has never entered.
This is the harder standard the puzzle asks us to keep in view. A plan should not merely make the desired outcome likely. Its count, signal, and stopping point must together make a false announcement impossible. In the first scheme, the one-time signal is what lets the leader interpret 99 counted offs as 99 different people.
What to check before trusting a solution
For any proposed plan, begin with the individual rule. What does each non-leader do on a visit? Can that person accidentally send the same signal twice? Can one prisoner’s action be mistaken for another’s? Then ask what the leader actually counts and why the count excludes every situation in which an unvisited prisoner remains.
The first scheme offers a clear answer to those questions: each non-leader can signal once, and the leader counts each signal by turning the bulb off. The second scheme has a different set of behaviors and a stated threshold, but the source excerpt does not show the proof. It also cuts off before the author’s correction of an earlier solution and fuller discussion of the optimal approach.
Nor does the available account give an expected number of days before anyone can announce success. The warden’s random selection makes the wait uncertain, and the certainty of the final claim is a separate matter from how soon it arrives. Those are useful limits to leave on the page rather than fill with a confident-sounding guess.
Return to the bulb, still a small object in a room no one can watch continuously. Its glow does not preserve a list of names. Under an agreed rule, though, a change can stand for a distinct act, and a count can become a shared memory. The prisoners are not remembering the room together. They are making a record that none of them could make alone.
Sources
- 100 Prisoners and a Light Bulb
- https://arxiv.org/pdf/2208.00771
- 5 Star Logic Problem (100 prisoners and a light bulb) • Physics Forums
- https://arxiv.org/pdf/1708.02654
- 100 Prisoners and One Light Bulb
- http://www.quanquancliu.com/slides/Eureka-64.pdf
- Puzzle: 100 Prisoners and a Light Bulb | Hacker News
- https://staff.fnwi.uva.nl/d.j.n.vaneijck2/papers/10/pdfs/FinalLightKR.pdf
An exceptional writer at Deeperpectives, she covers public memory, family archives, and the histories held in ordinary objects. Before joining the desk, she cataloged oral histories for a neighborhood museum, where she learned to treat a small detail as a door, not a conclusion.
