You and Your Research
Hamming’s question is why so few scientists do great work when so many have the ability. His answer is that first-class work is a managed outcome: luck is real but only “changes the odds,” and every other factor is under partial personal control. The talk is a working system for that management, drawn from watching Feynman, Fermi, Bethe, and Shannon up close and from auditing his own career against theirs.
The system
Compounding. Bode’s line — “knowledge and productivity are like compound interest” — is the engine. A steady extra hour of applied thought per day compounds over a career into a qualitative difference. Misapplied drive compounds nothing; Hamming saw harder-working peers with less to show.
Problem selection. An important problem is one with a reasonable attack, not one with important consequences — time travel and antigravity are unimportant by this definition. Most scientists never work on important problems in this sense. His enforcement mechanism was ritual: Friday “Great Thoughts Time,” 10% of his week reserved for what the field is becoming and where the door-opening opportunities are. Great scientists carry 10–20 attackable important problems and pounce when a new idea makes one tractable (the Berkeley fission near-miss is the cautionary tale).
Exposure. Open doors, the “what are the important problems in your field?” lunch campaign — deliberately trading short-term focus for ambient awareness of what matters. He can’t prove causation, but the closed-door / minor-work correlation was strong across decades.
Ambiguity tolerance. Believe your theory enough to push forward, doubt it enough to log every misfit (Darwin writing down each contradicting observation). Great contributions live in the flaws you kept track of.
Subconscious management. Deep, exclusive commitment keeps the background processor on your problem; a divided attention starves it. His creativity model is unsentimental: dreams rework the day’s inputs, so control the inputs.
Problem transformation. Repeatedly, the move that turned his good work into important work was reframing: the “defect” of having no programmers became automatic programming; one-off answers became “mass production of a variable product” — never solve an isolated problem except as the characteristic of a class. Generalizing often simplifies, and work shaped this way is what others can stand on.
Selling. Work unread is work undone: clear writing, prepared talks, and — the skill people skip — the informal hallway version. In the Q&A he goes further: at least 50% of time on presentation. The talk itself, still read forty years on, is its own evidence.
System judo. Learn what the system will do for you rather than fighting it (educating bosses via the BSTJ programmer-acknowledgment survey; Schelkunoff and the Friday deadline). Ego assertion — clothes, “I’ll do it my way” — is a small steady tax that compounds against you, the exact inverse of the extra hour. Fight the system only for amusement, in measured doses; the person who reforms the system and the person who does first-class science are almost never the same person.
Self-management. No alibis, honest inventory of faults, and conversion of faults into assets (his egotism became deadline courage). Emotional commitment is a necessary condition, and it costs — he had incipient ulcers; “nice guys end last” is quoted approvingly.
Failure modes he catalogs
Fame sterilizes: the Nobel effect (Brattain resisting it and losing anyway), Shannon after information theory, the Institute for Advanced Study as the institution that “ruined more good scientists than any other.” Fields exhaust: change subfields roughly every seven years, before your originality runs out. Perfect working conditions aren’t — the best conditions are the ones that force reframing (Cambridge shacks; his own machine scarcity, which in the Q&A he credits, half-seriously, for UNIX).
Assessment
The advice is a coherent control system, not a list of virtues: allocate attention (10% meta-work), compound the rest, keep the problem queue warm, and minimize steady-state ego taxes.
- Durable: the core claim that problem selection and problem reframing are trained, scheduled activities, not traits.
- Era-bound: the mechanics (open doors, secretaries, Bell Labs abundance) translate but don’t copy.
- Caveats (flagged by Hamming himself): the sample is survivors (“I have never dared to go out and ask those who didn’t do great work”), and luck’s counterfactuals are unknowable — his strongest claim is only that Feynman-grade people visibly would have done something great regardless of which something.