The What-Is-Your-Problem Test
Judge researchers and founders by the problem they name, not the result they promise
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 60%
Deutsch describes a persistent asymmetry: people with new theories who declare they will solve the origin of life, the nature of time, or the preconditions of the Big Bang attract more attention, investment, and podcast appearances than those who speak conjecturally. He does not treat this as an argument for overclaiming. The test he applies instead comes from Popper, whose standard opening on meeting someone outside your field was what is your problem, never what are you going to find, which would be a strange question. The rule is to evaluate the problem statement rather than the promised result, because the growth of knowledge cannot be foretold and a specific promise is therefore either luck or salesmanship. Deutsch applies it to himself, saying he does not know that Constructor Theory will work and that it could end up a curiosity.
Origin
Popper's habitual opening question, recounted by David Deutsch on Naval when asked whether speaking in conjectural terms is an inherent disadvantage against rivals who promise breakthroughs with great confidence.
Core principles
- 01You cannot foretell the growth of knowledge, so anyone promising a specific discovery is claiming something impossible.
- 02A well-stated problem is checkable today; a promised result is only checkable years later.
- 03Confident promising is optimised for funding and attention, not for progress.
- 04The people who make progress tend to be working on deep things for their own sake.
- 05Being willing to say the whole programme might be a footnote is a signal of seriousness, not weakness.
How to run it
- 1
Ask the Popper question
Open with what are you working on, or what is your problem. Deutsch notes that asking what are you going to find would be a weird question, and treating it as normal is the error.
Pro tip Use it as the literal first line so the answer is unrehearsed.
- 2
Check that the answer is a problem
A problem describes a conflict between existing explanations or a thing current theory cannot account for. A goal, a market, or a promised deliverable is not a problem.
Pro tip Deutsch's own version: here is my problem, I don't know where it's going, we think it might do something but maybe it won't.
Watch out Restating the desired outcome in problem language is the most common evasion.
- 3
Discount promised discoveries
Treat confident predictions of specific results as evidence about the speaker's incentives rather than the work. Deutsch says he is always suspicious when someone says what they are going to invent or prove.
Watch out Do not invert this into a rule that quiet people are always right; the test is about the claim, not the temperament.
- 4
Probe the failure mode
Ask what would make them abandon the programme, and whether they can describe the version of the future in which it turns out to be a minor curiosity. Willingness to answer is the real signal.
Pro tip Deutsch says he does not even know that Constructor Theory is going to work and that it may end up illuminating very little.
- 5
Check the motivation survives the audience
Deutsch's observation is that the people who make progress are usually working on deep things for their own sake. Fame sometimes follows, as with the mRNA vaccine, but those people were not famous beforehand.
Watch out Someone fanatical about approval and funding will optimise for exactly that, and the work will follow the incentive.
In the wild
Asked what confidence he has that Constructor Theory is the right approach, Deutsch answers that he cannot foretell the growth of knowledge and does not even know it will work. It could end up as a curiosity, the thing some people in the early twenty-first century thought would illuminate things, before the real illumination arrived and made it unnecessary. He notes that rival theorists making confident claims about solving the origin of life or the nature of time get considerably more attention and investment, and declines to copy them.
→ The interviewer identifies the stance as Popperian, and Deutsch treats the willingness to be a footnote as the honest position rather than a marketing problem to fix.
Deutsch says he would never have announced that he was going to invent quantum computers, and that it did not occur to him even when he did. His actual problem was narrower: the consensus held that Everett's Multiverse theory made the same predictions as collapse theories, and he thought it must be testable. Designing a test required a measurement of a measurement, which required an observer that could not be decoherent, which required augmenting a computer with quantum operations. A conversation with Charlie Bennett about complexity depending on the hardware, and hardware being physics, then pushed him to recast Turing's analysis in quantum physics.
→ The result was the quantum theory of computation, arrived at as a byproduct of a stated problem rather than as a promised invention.
Common mistakes
Rewarding confidence as competence
Confident specific promises correlate with fundraising skill, not with the growth of knowledge, which by its nature cannot be forecast.
Accepting a goal dressed as a problem
Wanting to build a thing is not the same as identifying an explanatory gap, and only the latter can be evaluated before the work is done.
Is it for you?
Best for
Evaluating research programmes, deep-tech founders, or anyone pitching a long-horizon intellectual bet.
Not ideal for
Execution-stage work where the problem is settled and delivery against a known spec is what matters.
From the transcript
“I'm always suspicious when someone says what they're going to invent or what they're going to prove”
“I would rather they said this is the problem I'm going to address I don't know where it's going”
“when you meet someone who is not in your field first thing you say is what are you working on”
From the episode
The Deutsch Files IV