Quick refresher in case it's been a while. A Boltzmann brain is a hypothetical observer that fluctuates into existence by random chance in a high-entropy universe — a momentary configuration of matter that happens to contain the experience of being a conscious person, complete with memories, perceptions, and a sense of continuity, all fabricated. It exists for an instant and then dissolves. The unsettling part: in some cosmological models, such fluctuations vastly outnumber normal embodied observers like us, which means a randomly selected "observer-moment" is more likely to be a Boltzmann brain than a real person. So when you ask "am I a real person or a fleeting fluctuation that just happens to feel like one?" — that's a live question, not just a stoner thought.
The standard reply is that you can't test your way out from the inside. Any evidence you marshal could itself be part of the fluctuation. I want to propose a test that I think gives you genuine traction anyway, with caveats.
The setup:
Next time the thought of Boltzmann brains crosses your mind, write a number 1–10 on a piece of paper, memorise it, fold it into your wallet. Next time the thought crosses your mind again — could be hours, days, weeks later — recall the number first, *then* unfold the paper and check. Tick if you matched, cross if you didn't. Fold it back. Repeat indefinitely.
The logic:
If you're a normal embodied person with working memory, you'll match nearly every time — call it 99%. If you're a Boltzmann brain, there's no causal link between the number you "remember" and the number actually written down, because neither has a real history. The match is 50/50 by indifference. (The 1–10 range is a red herring — it could be letters, Roman gods, anything. What matters is the binary outcome that gets recorded: tick or cross.)
The objection I expected:
A Boltzmann brain can fluctuate into existence already holding the memory of N consecutive ticks. So accumulating ticks isn't really accumulating evidence — any given moment of "checking my paper" might be a single isolated fluctuation whose contents include a false memory of having played before. The tick marks aren't evidence of past trials; they're evidence of present perceptions of tick marks, and a fluctuation can produce those directly.
Why the test still works:
Among all Boltzmann fluctuations that contain the experience of "I just played N rounds," only 1 in 2^N will contain all-tick memories. Coherent runs are exponentially rare among possible fluctuation contents — most fluctuations holding "I just played 20 rounds" will hold a roughly even mix of ticks and crosses, because that's what random pairing produces. So finding yourself looking at a paper covered in ticks is genuine Bayesian evidence against being a Boltzmann brain. Not because such Boltzmanns can't exist, but because they're a vanishing fraction of the total.
The math:
Probability of N ticks under the Boltzmann hypothesis is (0.5)^N.
- 10 games: ~1 in 1,000
- 20 games: ~1 in 1,000,000
- 30 games: ~1 in 1,000,000,000
You don't need a likelihood ratio against the normal-observer hypothesis. You're just asking how surprised you should be by your result *if* you were a Boltzmann. Pick a threshold that satisfies you personally and play until (0.5)^N crosses it. Twenty games gets you past one-in-a-million.
What this actually accomplishes:
It doesn't disprove that Boltzmann brains exist or that they dominate the universe by measure. It gives the individual observer a personal Bayesian instrument for their own situation. The question "am I a Boltzmann right now" is about a token, not a type, and tokens admit updates from observed coherence.
Caveats:
There's a mild circularity. You're using your memory of past ticks to argue against being the kind of thing whose memories are unreliable. You have to trust the coherent experience you're currently having to even run the test. This is the standard Cartesian floor — not fatal, but real.
In cosmologies where Boltzmann brains vastly outnumber normal observers, even all-tick Boltzmanns might outnumber all-tick normal observers, and the test loses traction by measure. But those cosmologies are arguably self-undermining anyway (Sean Carroll's argument that they predict our evidence is untrustworthy, including the evidence for the cosmology itself).
The test really just formalises and quantifies the intuition "things seem coherent, so probably I'm real." It doesn't escape the skeptical loop. It puts a number on where you were already standing.
Within those limits: carry a folded paper in your wallet. After twenty matches you've earned the right to feel real to one part in a million.