Richard Feynman · Chapter 4

The Safecracker of Los Alamos

How Mechanical Tolerances Reduced a Million Combinations to Twenty Minutes

A standard Mosler three-wheel combination lock fitted to the steel filing cabinets of the Manhattan Project featured a rotating dial numbered from zero to ninety-nine. In textbook combinatorics, dialing three independent numbers in sequence yields $100 \times 100 \times 100 = 1{,}000{,}000$ distinct combinations. At ten seconds per trial, opening a single locked cabinet by exhaustive trial and error would require nearly one hundred and sixteen days of continuous spinning. Army security officers at Los Alamos, New Mexico, between 1943 and 1945 relied on that million-to-one arithmetic. They believed the locks guarded the blueprints for uranium-235 gun assembly and plutonium implosion lenses.

Young theoretical division group leader Richard Feynman looked at those same steel cabinets and saw a physical machine governed by friction, lever slack, and human habits. When the laboratory upgraded from simple key padlocks to three-wheel Mosler filing cabinets, Feynman resented the bureaucratic theater. Security rules locked technical reports away from the scientists who needed them to build the bomb. He ordered a commercial lock-repair manual, dismantled his own office lock at night to inspect its brass tumblers, and mapped the mechanical shortcuts that ruined its theoretical security.

By combining mechanical tolerances with behavioral observation, he collapsed one million possibilities into a twenty-minute routine:

He later explained that whenever a colleague left a safe open during an office conversation, he would casually rest his hand on the cabinet handle and turn the dial just enough to read off the last two numbers of the combination. Back in his dormitory room, Feynman wrote the last two numbers of each colleague's safe in a private notebook. If he ever needed a document after hours, he already knew two of the three wheels. With the five-unit tolerance on the first wheel, he only had to test 20 numbers. The drawer pulled open in less than four minutes.

To keep his skill sharp, he also learned from a professional locksmith at Los Alamos how to feel the tiny bounce of the friction latch when the third wheel slot passed beneath the drop lever. Soon Feynman could crack an unknown safe from scratch in half an hour while sitting alone in an office.

The famous test came when Feynman needed classified cross-section reports stored in the office of Frederic de Hoffmann while de Hoffmann was away for the weekend. Opening all three of de Hoffmann's filing cabinets in succession, Feynman removed the needed calculation sheets. Inside the drawers, he left paper notes mocking the locks. When an Army officer named Captain Karl later installed a heavy Mosler vault door and boasted that no scientist could breach it, Feynman discovered that the captain had never changed the factory shipping code from $50\text{-}25\text{-}50$. When Feynman pulled open Captain Karl's newly installed safe to borrow classified uranium-enrichment reports, he left a yellow slip inside reading that he had picked the lock because the cabinet was no safer than a wooden drawer.