Millikan's e = 4.774 ± 0.009 ×10⁻¹⁰ esu vs the value the SI later made exact, 4.80320471: 3.2× his quoted error, mostly traced to the viscosity of air at the 3/2 power (a sensitivity his own paper states). The page draws the actual record of successive measurements and the creep Feynman made famous. Every number recomputed.

An interactive girih-tile loom: five tiles, two strands crossing every edge's midpoint at 54°, one continuous pattern with no end. Follow a single strand yourself. The quasicrystal claim (Lu & Steinhardt on the Darb-i Imam shrine) is cited and its scope stated honestly; the tile geometry is verified offline.

Does water drain the other way in the Southern Hemisphere? Drag real systems across the Rossby line; settle the swirl the way Shapiro did in 1962. Every figure recomputed by a published check: 15/15 green.

An interactive hunt for Garden of Eden patterns: configurations a cellular automaton can display but never produce, counted per ring size for rule 30, rule 110 and others. The counts were absent from OEIS when checked on 2026-07-19, and your browser recounts them as you read.

Drag through the years and watch two hulls trade their timber plank by plank, with no point where identity 'switches.' The famous twist is Hobbes (1655), not the Greeks; both primary sources quoted verbatim.

An instrument, not an article: a Planck-law incandescence engine, an emission-spectrum bench for the flame's second light, and a live Saha plot that settles the plasma question. Every number recomputed in your browser and re-derived offline in verify.mjs, 21/21 assertions passing.

Herd immunity is where an epidemic peaks, not where it stops. The overshoot, playable, and the 1906 Bombay plague table behind the famous fitted curve.

Why did the Quebec Bridge collapse? The 1908 Royal Commission report's figures, read from the page images: the bent chord, the lattice rivets, the dead load nobody recomputed.

Where did Shackleton's Endurance sink? Worsley's log re-added, the drift carried back, the clock corrected, and the predictions made before the find scored.

Galileo's two lectures on the shape and size of Dante's Inferno, redone as arithmetic, and his square-cube mistake, fifty years before Two New Sciences.

How does a whispering gallery work? Airy said the dome focuses; Rayleigh said the wall traps. Trace rays in an oval and a round room, then run the wave.

A trial writes down what it will measure before it starts, and the registry keeps every version. The rule that reads those versions counts rows, and a row is not a measure. Sixty records, three blind readers, a rubric written before the draw.

The 2010 muonic-hydrogen radius, rebuilt from its measured energy, beside 0.8406 fm from a 2026 measurement in ordinary hydrogen. Then move that measurement’s two frequencies together and watch a precision check miss the entire shift.

Le Verrier’s 1859 sums and Newcomb’s 1882 check, each rerun from the equations its author printed. The excess was real and larger than claimed, and general relativity accounted for it in 1915; the page also finds two small slips in Newcomb’s printed normal equation (8), neither touching his conclusion.

A spike tells you people like the number. The void above it tells you they were fleeing a tax. 31,525,946 registered sales, four statutory dates, eight round-number controls that never move.

The Claremont Shakespeare Clinic’s 2004 odds against the Earl of Oxford’s poems, recomputed from its own published test scores, then the test its critics proposed in 2009 and neither side ran: Shakespeare’s earliest plays scored against ranges from his later ones, with Oxford’s departure planted into copies to show whether the test could have caught it. Whether the odds measure authorship remains open; who wrote Shakespeare is not tested here.

In 1988 a shape-from-shading study put the Face on Mars at 412.5 ± 17.5 m and argued its features would persist under other lights and views. This page relights the HiRISE team’s 2025 terrain model of the hill under the Viking sun, plants the 1988 height into it, and says what held up and what did not.

An IPCC chapter said Himalayan glaciers would very likely disappear by 2035; the IPCC later deleted the paragraph. At the satellite-measured loss held constant the ice lasts until 2146, and 2035 would need 4.17 times that loss; the page plants the claim into a copy of the record to show the record could have seen it.