The Idiothesis
Generate two stupid hypotheses a day and stay unattached to every one of them.
- Difficulty
- Starter
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 90%
The idiothesis is Eagleman's term, coined mid-conversation, for an idiotic hypothesis — a really stupid hypothesis that might just be right, but could just as well be wrong. He defines it precisely: the kind of hypothesis he is not attached to enough to fight and die for, but which is interesting enough that he will keep an eye on it and see if he can collect data on it. He generates at least two a day. The example that produced the term is his own guess that we might see a tiny decline in dementia some decades from now, because the pandemic knocked everyone off their hamster wheels and forced people to rethink supply chains, haircuts, and how the world works in the middle of their lives. The practice sits directly downstream of Francis Crick's advice, which Eagleman quotes approvingly: the dangerous man is one who has only one idea, because then he'll fight and die for it, whereas the way real science goes is that you come up with lots of ideas and most of them will be wrong. Eagleman notes he was already this way — he rejected an engineering professor's lumberjack advice to pick one tree and hit that tree, saying it sounded wise and reasonable but was just not who he is. The naming does real work: calling an idea an idiothesis pre-declares that you will not defend it, which lowers the cost of generating and of abandoning it.
Origin
Eagleman coined the word live on the podcast while floating a speculative guess about pandemic-driven cognitive resilience, formalising a habit he already had of producing at least two low-attachment hypotheses a day.
Core principles
- 01The dangerous person is the one with only one idea, because they will fight and die for it.
- 02Real science runs on many hypotheses, most of which will turn out to be wrong.
- 03Labelling an idea as stupid up front removes the ego cost of it being wrong, which raises your generation rate.
- 04An idiothesis is defined by low attachment plus enough interest to keep watching for data.
- 05Volume matters more than initial quality — Eagleman produces at least two a day.
- 06The practice is a portfolio hedge: with many live hypotheses, no single refutation costs you your identity.
How to run it
- 1
Notice the anomaly
Catch something that does not fit your model — an odd correlation, a surprising exception, an assumption everyone shares. Eagleman's trigger was how thoroughly the pandemic disrupted everyone's internal models.
Pro tip Anomalies you stumble on incidentally are richer than ones you go looking for.
Watch out Waiting for a well-formed research question kills the practice before it starts.
- 2
Write the first stupid explanation
Put down whatever occurs to you immediately, without filtering for plausibility or reputational safety.
Pro tip Speed is the point — Eagleman generates at least two a day, which is only possible without a quality gate.
Watch out Editing at the generation stage collapses volume, which is the whole mechanism.
- 3
Label it an idiothesis
Explicitly tag the idea as one you are not attached to. This is the load-bearing move: the label pre-commits you to abandoning it cheaply.
Pro tip Say the label out loud when sharing it, so listeners calibrate too.
Watch out An unlabelled speculation gets treated by others as your position, and then you start defending it.
- 4
Generate a rival
Produce at least one competing explanation for the same anomaly, in line with the hypothesis-testing method Eagleman teaches his own children.
Pro tip Two candidates turn a belief into a comparison, which is the unit science actually works on.
Watch out A single hypothesis with no rival is exactly the one-idea position Crick warned about.
- 5
Name the distinguishing observation
State what you would need to see to tell the candidates apart. Eagleman's dream hypothesis qualified as serious precisely because it made quantitative predictions across species.
Pro tip If a cheap existing dataset could settle it, promote the idiothesis to a real project.
Watch out An idiothesis that cannot in principle be distinguished from its rival is entertainment, not a hypothesis.
- 6
Keep an eye out and let go
Log it, watch passively for data over months, and drop it without ceremony when evidence goes the other way. Eagleman's own test of an elephant counterexample to his dream theory is the model — he went looking for the thing that would kill it.
Pro tip Actively hunting for counterexamples to your own idiothesis is the fastest way to promote or kill it.
Watch out Any reluctance to discard it means it has stopped being an idiothesis and become a position.
In the wild
Eagleman floats the guess that we might see a tiny decline in dementia decades from now, on the grounds that the pandemic knocked everyone off their hamster wheels mid-life and forced them to rethink supply chains, whether the barber shop would open, and how the world actually works. He immediately frames it as an idiotic hypothesis he is not attached to.
→ The idea got aired and named without Eagleman having to defend it, and produced the term idiothesis on the spot.
An electrical engineering professor Eagleman liked told him life is like being a lumberjack: you go into the forest with your axe, pick one tree, and hit that tree, because taking a whack at all the trees gets you nowhere. Eagleman calls it the worst piece of advice he ever received — it sounded wise and reasonable, but he is just not the pick-one-tree guy.
→ He kept the many-hypotheses posture that later produced eight books, 70 sensory projects, and the dream theory.
Common mistakes
Letting a favourite idiothesis quietly become a position
The definition requires low attachment. Once you find yourself arguing for it, you have become Crick's dangerous man with one idea who will fight and die for it. The observable tell is reluctance to look for the counterexample that would kill it.
Filtering at the generation stage
The rate — at least two a day — only survives if nothing is screened for plausibility on the way in. Applying scaffolding, constraints and filters up front, as Tim notes, is exactly the pressure the label is designed to remove, and it collapses output to the safe ideas everyone already has.
Publishing idiotheses without the label
An untagged speculation is read as a claim, and once it circulates as your claim, defending it becomes socially costly to abandon. Eagleman states the label explicitly every time, which is what buys him the freedom to be wrong in public two times a day.
Is it for you?
Best for
Researchers, founders, and anyone whose edge comes from noticing anomalies and generating many candidate explanations cheaply.
Not ideal for
Contexts where floating a half-formed public claim carries real reputational or safety cost, such as clinical or regulatory advice.
From the transcript
“Here's my idiothesis, which is an idiotic hypothesis.”
“really really stupid hypotheses that might just be right, but they could just as well be wrong.”
“interesting enough that I'm going to keep an eye on it and see if I can collect data on it.”
“the dangerous man is one who has only one idea because then he'll fight and die for it. The way real science goes is that…”
From the episode
#674: Neuroscientist David Eagleman — Exploring Consciousness, Sensory Augmentation, The Lazy Susan Method of Extraordinary Productivity, Dreaming, Improving Hearing with a Wristband, Synesthesia, Stretching Time with Novelty, Lessons from Titans of Science, and Much More