March 12, 2026
Schrödinger’s wallet, Reddit’s war
Newcomb's Paradox Needs a Demon
Internet meltdown over the million‑dollar box — bring the demon
TLDR: Sam Estep argues Newcomb’s paradox only works with a near-supernatural predictor; otherwise coin flips break the game. Comments erupt: determinists say free will vanishes, pragmatists push one-box for the million, skeptics two-box, and jokesters demand a demon to settle it.
Veritasium dropped a mind-bender on Newcomb’s paradox: two boxes, one opaque, maybe hiding $1M if a predictor guessed you’d only take it. Writer Sam Estep clapped back, saying the setup “needs a demon” — a perfect knower like Laplace’s demon. Without a supernatural brain, he says, coin‑flip trolls nuke the magic, and probability talk falls apart. Cue comments chaos. The determinists showed up first, insisting that a flawless predictor means free will is canceled: “you do not have a choice,” thundered danbruc. The pragmatists brought street sense: jack_pp asked who would gamble for $1000 when the million is more likely if you commit to one box. Meanwhile, contrarians like vidarh shrugged, saying without testing the predictor it’s “Pascal’s Wager married to the Halting Problem,” so just pick both and move on. Confused readers chimed in (“why not take both?”) while the rules‑lawyers begged everyone to accept the hypothetical, as gegtik put it. Jokes flew fast: Team Demon vs Team Dr. Phil, Schrodinger’s wallet memes, and coin‑flip flexes. Estep’s twist about Nozick’s paper—punishing “conscious randomizers”—only fueled the drama. Verdict? The box is opaque, but the internet’s feelings are anything but.
Key Points
- •The article argues that in Newcomb’s paradox the predictor’s identity matters: a demon-like perfect predictor differs from computers or psychologists.
- •It claims computers/psychologists cannot reliably predict deliberate randomization (e.g., coin flips), challenging the video’s equivalence of predictors.
- •Veritasium’s “almost always correct” predictor changes the problem versus a flawless predictor by allowing future errors.
- •A probability argument in the video (one-box if accuracy C>0.5005) is criticized for assuming predictor accuracy is independent of the chosen decision strategy.
- •Robert Nozick’s 1969 formulation is cited: a never-wrong predictor that sets $0 if it predicts conscious randomization, raising definitional issues.