Richard Feynman

Richard Feynman turned one of physics' hardest calculations into little drawings with arrows. That is why his name travels so well outside physics: he made understanding visible. The blackboard line found after his death in 1988, "What I cannot create, I do not understand," is not a motivational quote. It is a test.

Feynman was a theoretical physicist, 1918–1988, Caltech professor from 1950, Nobel laureate in 1965 with Julian Schwinger and Sin-Itiro Tomonaga for quantum electrodynamics. At Los Alamos, age 24, he helped manage hand calculations for the Manhattan Project using rooms of people and mechanical calculators. Before electronic computing became normal, he was already thinking in pipelines.

The artifact was the argument

Feynman's best work did not ask people to admire abstraction. It gave them an object to manipulate.

Feynman diagrams, introduced in 1948, made quantum electrodynamics computable by sight. Electrons became lines. Photons became wavy lines. Interactions became vertices. The diagrams did not replace the math; they organized it so a trained graduate student could track terms that had been buried in integrals.

Schwinger and Tomonaga reached equivalent physics by different routes. Freeman Dyson showed the equivalence in 1949. Feynman's version won daily use because it was teachable at a blackboard.

The same pattern shows up in The Feynman Lectures on Physics, delivered at Caltech from 1961 to 1963. The course tried to rebuild undergraduate physics from first principles, not from textbook order. Many students found it too hard; the books later became more famous than the class. The published lectures have sold more than 1.5 million copies and remain free online through Caltech.

Then came the Challenger hearing on February 11, 1986. Feynman placed a piece of O-ring rubber in ice water during the Rogers Commission proceedings and showed that it lost resilience in the cold. The demonstration did not explain the whole disaster, but it made one failure mode impossible to ignore.

The notebooks were not records

In an oral-history interview, Charles Weiner treated Feynman's notebooks as records of completed thinking. Feynman objected. The notebooks were the thinking.

That distinction matters. A derivation on paper is not a transcript of an answer already formed in the head. For hard problems, the page is part of the machine. Cross-outs, false starts, diagrams, and half-computed terms are not waste; they are how the mind extends itself.

That is why Feynman belongs near concept second brain, not as a productivity mascot, but as a severe example of external cognition. The point is not to archive everything. The point is to make a surface where thought can collide with itself.

What is contested

The Feynman myth is cleaner than the person. His QED work was not a lone conquest; the Nobel was shared for a reason. Schwinger's formalism was powerful, Tomonaga's work came from wartime Japan under worse conditions, and Dyson made the comparison legible.

His public persona also distorts the signal. The bongo drums, safe-cracking, and stories in Surely You're Joking, Mr. Feynman! made him culturally portable, but many popular references to Feynman quote the character rather than the physicist.

There is a harder accounting too: his own books describe behavior toward women that reads badly without needing present-day exaggeration. The method is worth studying. The man does not need polishing.

Why builders keep returning to him

Feynman gives engineers a clean rule: build the smallest version that still exposes the mechanism.

That rule maps directly onto person karpathy and small-code teaching artifacts like nanoGPT: a working transformer in a few hundred lines is not valuable because short code is cute. It is valuable because it forces every hidden assumption into view. The same instinct sits under clear explanations of concept rag, where the magic dissolves into retrieval, prompt construction, generation, and evaluation.

The danger is scale. A Feynman diagram works because the symbolic object stays tied to the physics. A toy model can lie if it keeps the easy part and drops the thing that matters. The construction test is not "small is true." It is "small enough to inspect, faithful enough to teach."

Why this has to do with other realms

Feynman's rule has an older cousin in overview vedanta: anubhava, direct knowing, ranked above knowledge taken only from testimony. Feynman would have rejected much of the metaphysical scenery, but the epistemic move rhymes. Being told is weaker than seeing. Seeing is weaker than doing.

That also links him to concept fermi paradox. In both cases, the honest position begins with a limit: you do not understand a civilization-scale silence, or a quantum calculation, by naming it. You need a model that can fail in your hands.

An open question

If the thing to understand is a 70-billion-parameter model that no individual can train, inspect, or mentally simulate, what counts as Feynman's "create" test: a toy model, an interpretability probe, a reproduction of one behavior, or something we have not named yet?

Key Sources

The Feynman Lectures on Physics by Richard Feynman, Robert Leighton, and Matthew Sands (1963–1965) — the teaching method in primary form.
QED: The Strange Theory of Light and Matter by Richard Feynman (1985) — Feynman explaining his diagrammatic view for non-specialists.
Genius: The Life and Science of Richard Feynman by James Gleick (1992) — the main biography for Los Alamos, Caltech, and the public persona.
Rogers Commission Report on the Space Shuttle Challenger Accident (1986), Appendix F by Feynman — the dissent on risk, O-rings, and NASA management.
American Institute of Physics oral-history interviews with Richard Feynman and Charles Weiner — source for the notebooks-as-work exchange.
Disturbing the Universe by Freeman Dyson (1979) — outside view on how Feynman, Schwinger, and Tomonaga converged.

Abhishek's take

I trust a model only after I can reduce it to one working sheet: input, rule, exception, output. On the floor, a size curve or a vendor promise looks clean until a single late fabric lot breaks the story. The tool I wrote earns its place when the buyer can point to that break before the buy is locked.

Abhishek