How great scientists actually reasoned: Feynman, Curie, Darwin and the documented methods behind the discoveries
The Feynman technique, slow hunches, and thought experiments, explained through the biographies and the scientists' own accounts rather than the inspirational-poster version.
Scientific thinking is less a body of facts than a discipline for resisting the wrong process, and the people in this collection left behind unusually detailed accounts of what that discipline actually looked like day to day. Richard Feynman kept notebooks where he explained ideas back to himself in the simplest possible language until the gaps in his own understanding became visible. Marie Curie spent roughly four years physically processing tons of ore to isolate a fraction of a gram of radium. Charles Darwin sat on his central theory for two decades, accumulating evidence and, notably, his own objections to it, before publishing. This guide takes the collection's signature methods in turn, with the documented practice behind each, before addressing why the category groups physicists, biologists and philosophers under one method rather than by discipline.
What unites the frameworks below is less a shared subject matter than a shared discomfort with the easy version of an answer. Each of these thinkers built a specific habit for slowing down the moment where confidence outruns evidence, which is a discipline worth studying regardless of whether the reader ever runs a formal experiment.
The Feynman technique
Feynman's documented habit, visible in his Caltech notebooks and repeated in his own popular writing, was to write a concept's name at the top of a page and explain it in the plainest possible language, as though teaching someone with no background in the subject. Where the explanation broke down, reaching for jargon or an unexplained leap, that specific gap marked exactly what he did not yet understand well enough, and he would return to the source material to close it before simplifying again.
A worked example: asked to explain why a physical phenomenon works a certain way, Feynman's method would refuse an answer built on unexplained technical terms ("it's due to quantum tunnelling") and instead push for a version that unpacks what tunnelling actually is in plain terms, at whatever length that requires, on the argument that an explanation you can't simplify is usually one you don't yet fully understand yourself.
Where it fails: some genuinely technical concepts resist full plain-language reduction without losing precision that matters, particularly in advanced mathematics, and pushing every explanation to be fully accessible can strip out the exact rigour a technical audience actually needs; the technique is a diagnostic for your own understanding first, and a teaching tool second, and conflating the two can produce oversimplified explanations for an audience that needed the precision.
Slow hunches and delayed publication
Darwin's documented practice, visible across his notebooks and correspondence, was to protect an early idea for years while deliberately accumulating both supporting and disconfirming evidence, rather than publishing on the strength of the initial insight alone. He developed the outline of natural selection by the late 1830s but did not publish On the Origin of Species until 1859, spending part of the intervening decades on detailed barnacle taxonomy that built his credibility and sharpened his evidentiary base before the more contentious theory went public.
A worked example: Darwin's own notebooks record objections to natural selection as he encountered them, including the problem of complex organs like the eye, which he treated as a real difficulty to be addressed rather than an inconvenience to be ignored, and his published work directly confronts objections he had been sitting with for years rather than presenting the theory as uncomplicated.
Where it fails: protecting an idea for decades assumes the idea survives sustained scrutiny, which is a reasonable expectation for a theory as extensively tested as Darwin's own but a poor justification for indefinitely delaying publication of an idea that hasn't actually been rigorously stress-tested in the meantime; slow hunches require the accumulating evidence to be genuine, not merely the passage of time.
Thought experiments
Both Einstein and Turing used carefully constructed imagined scenarios to settle questions no physical apparatus of their era could test directly. Einstein's imagined pursuit of a beam of light, described in his own later recollections as a formative early thought experiment, helped him reason toward special relativity's core claims. Turing's imagined machine, reading and writing symbols on an infinite tape according to simple rules, defined the theoretical limits of what any computing device could in principle calculate, years before a machine capable of testing the idea existed.
A worked example: rather than building physical apparatus to test what an observer moving at light speed would perceive, a physically impossible experiment in Einstein's era, the thought experiment isolates the relevant variable (the constancy of light's speed for all observers) through careful logical construction alone, reaching a testable prediction without needing the impossible physical trial.
Where it fails: a thought experiment is only as reliable as the logical construction underlying it, and a flawed initial assumption produces a confidently wrong conclusion with no physical trial available to catch the error before the reasoning is published and built upon; the method requires unusually rigorous self-checking precisely because it lacks the natural error-correction a physical experiment provides.
Negative results and disconfirmation
The discipline of treating an experiment that disproves a hypothesis as equally valuable to one that confirms it runs through this category's most careful thinkers, documented most explicitly in how Darwin recorded his own objections and in how modern scientific method formally requires falsifiability as a criterion for a claim to count as scientific at all.
A worked example: a researcher testing whether a new teaching method improves outcomes, and finding no measurable difference against the control group, has produced a genuinely useful result under this framework, ruling out one hypothesis and freeing resources to test others, even though a null result is far less likely to be published or celebrated than a positive finding would be.
Where it fails: the framework describes what good scientific reasoning requires; it does not describe the actual incentive structure most working scientists operate inside, where journals have documented, well-known biases toward publishing positive results, meaning the discipline of valuing negative results has to be maintained by individual researchers against systemic pressure that doesn't reward it.
First principles and computational reduction
Turing's approach to defining computation, stripping the question of what a machine can calculate down to the smallest possible set of operations, an imagined tape and simple read-write-move rules, established the theoretical floor for all subsequent computing before any electronic computer existed to test the idea against. The method is a specific application of a broader first-principles habit shared across this category: strip a problem to its irreducible components before reasoning forward from them.
A worked example: rather than asking whether a specific proposed computer design could solve a particular problem, Turing's framework asks the more fundamental question of whether any conceivable mechanical process could solve it at all, a question that, once answered, tells you something true regardless of which specific machine you eventually build.
Where it fails: reducing a problem to its theoretical minimum tells you what is possible in principle, not what is practical, efficient, or economically sensible to actually build; a result that is computable in Turing's sense may still require more time or resources than any real system could tolerate, which is precisely why computer science later developed complexity theory as a separate, additional layer on top of Turing's foundational question.
Tolerance for tedium
Curie's documented work isolating radium, roughly four years processing tons of pitchblende ore by hand in a poorly resourced shed, is a reminder that a genuinely correct hypothesis still requires sustained, unglamorous execution to confirm, a discipline distinct from the insight that identified what to look for in the first place. Her notebooks and subsequent biographies describe repetitive, physically hazardous manual processing, without the safety understanding later research would establish about the material she was handling, sustained specifically because the extraction process offered no shortcut.
A worked example: having correctly hypothesised that pitchblende contained an undiscovered radioactive element, Curie and her collaborator still had to physically process enormous quantities of raw ore through repeated chemical separation to isolate a usable, measurable quantity, a process no amount of additional theoretical insight could have shortened; the remaining work was purely a matter of sustained manual execution.
Where it fails: tolerance for tedium as a virtue only pays off when the underlying hypothesis is actually correct; sustained, patient execution of an experiment built on a flawed premise produces years of wasted labour rather than a delayed discovery, and the discipline itself offers no way to distinguish which case you're in until the results eventually arrive.
Why this category spans physicists, biologists and philosophers
The apparent oddity of grouping Feynman, Darwin, Turing and Aristotle under one heading dissolves once the organising principle is understood correctly: the category groups by method, disciplined reasoning under uncertainty, rather than by academic department. Aristotle's systematic classification of the natural world, Plato's dialogue method, and Feynman's demand for explanatory clarity are each, in their own era, a discipline for catching confident-sounding claims that don't actually hold up, which is the through-line the collection is organised around rather than a shared subject matter.
The Scientist & Thinker category collects these frameworks, and others including Jane Goodall, Katherine Johnson and Alan Turing, as downloadable .md files for Claude, ChatGPT or any LLM.
Frequently asked questions
What does the Feynman technique actually involve, step by step?
Write the concept's name at the top of a page. Explain it in plain language as if teaching someone with no background, avoiding jargon entirely. Where the explanation breaks down or you reach for a term you can't unpack further, that gap marks exactly what you don't yet understand. Return to the source material to fill that specific gap, then simplify the explanation again. Feynman's own notebooks, held at Caltech, document him doing precisely this for problems well outside his acknowledged expertise.
How long did Marie Curie's radium extraction actually take, and why does that matter?
Roughly four years of processing tons of pitchblende ore by hand in a poorly equipped shed to isolate a fraction of a gram of pure radium, documented in her own notebooks and subsequent biographies. It matters because the discovery is often taught as a moment of insight when the actual documented work was overwhelmingly manual, repetitive extraction under physically hazardous conditions, insight having already identified what to look for; the achievement was sustaining the tedium long enough to isolate it, not the initial hypothesis alone.
Did Darwin really wait decades before publishing his theory, and why?
Yes. Darwin developed the core outline of natural selection by the late 1830s but did not publish On the Origin of Species until 1859, a gap of roughly two decades. His own correspondence and notebooks document years spent accumulating supporting evidence, including his detailed work on barnacles, partly out of genuine scientific caution and partly, by his own account and later historians' analysis, from awareness of how contentious the theory would be received.
What's a thought experiment, concretely, and how is it different from an actual experiment?
A thought experiment isolates a variable through careful reasoning about an imagined scenario rather than a physical trial, used when the physical version is impossible, impractical, or would only confirm what careful reasoning already establishes. Einstein's imagined scenario of riding alongside a beam of light, and Turing's imagined machine reading and writing symbols on an infinite tape, both used this method to settle a theoretical question no physical apparatus of their era could have tested directly.
Is 'negative results matter as much as positive ones' actually how science works in practice?
In principle yes, in practice unevenly. The framework's own documented history includes a well-known bias: journals have historically favoured publishing confirming results over disconfirming ones, which distorts the record over time even though working scientists like Darwin explicitly built disconfirmation-seeking into their method. The framework describes good scientific practice; it does not describe the publication incentives every working scientist operates inside, which is a separate and ongoing problem the field itself has documented and debated.
Can Turing's computational thinking apply outside computer science?
Yes, and this is one of the more widely transferred frameworks in the collection. Turing's core move, reducing a complex problem to the smallest set of operations that could in principle solve it, generalises to any problem where you're trying to establish what is fundamentally possible before worrying about an efficient implementation. It underlies how computer scientists reason about a problem's difficulty before ever writing code, and the same reduction habit transfers to strategic planning, where the question 'what is the minimal version of this that would actually work' does similar work.
How do you tell a slow hunch worth protecting from a bad idea you're rationalising?
Darwin's own documented practice offers a partial answer: he kept detailed notebooks recording objections to his own theory as they occurred to him, rather than only recording supporting evidence. A slow hunch worth protecting can survive its own documented counter-arguments over time without requiring you to suppress or ignore them; a bad idea rationalised typically requires increasingly selective attention to evidence to keep feeling true. The discipline is in keeping the counter-evidence file, not just the supporting one.
Why does this category include philosophers like Aristotle alongside physicists like Feynman?
Because the boundary between science and structured inquiry is a modern administrative division more than a description of how these thinkers actually worked. Aristotle's systematic classification of the natural world and Feynman's demand for explanatory clarity are both, at root, disciplines for reducing confident-sounding nonsense; the framework here groups by method, careful reasoning under uncertainty, rather than by academic department.
Written by Gareth Hoyle. Last updated 24 August 2026. Part of the authority.md guides library.
More guides.
The performance frameworks behind Jordan, Bryant and Federer: what the documented record actually shows
Mamba mentality, process over outcome and the longevity discipline, explained through interviews, documentaries and the athletes' own accounts rather than the highlight-reel version.
The coaching frameworks that built championship cultures: Jackson, Wooden, Belichick and the documented method
The standard of performance, system over tactics, and player development as the moat, explained through the actual programmes and documented practices rather than the motivational-poster version.
The comedy writing disciplines that actually land: Seinfeld, Carlin, Gadsby and the documented craft
The daily streak, form subversion and material on yourself, explained through the actual specials, interviews and writing process rather than the open-mic pep talk version.