When this site gets something wrong, the fix is logged here, in public, with a date. A site that hides its corrections is not more reliable, only quieter about its mistakes.
12 September 2026 (continued)
The Spin Drive ledger priced the tank's mass ratio of 21 against the 0.92 c commute's fare of 24 as two per cent under. That figure belongs to a different comparison: the cruise the tank buys, 0.909 c, is 1.2% under 0.92 c. As a fare it is 12% under, which is what the English claim row already said and the card and both Chinese passages now say. One inspector card in the same instrument put the schedule's total rapidity at 6.02; everywhere else, and in the bench, it is 6.03.
The English panels of the Hollow Moon ledger and the Mars Xenon ledger misstated their own benches' constants: the 620 Ma tidal rhythmites put the Moon at 58.16 Earth radii and the panel said 58.3 (the Chinese panel said 58.2), and ALH84001's ¹²⁹Xe/¹³²Xe ratio is 2.52 in the data while the English button said about 2.4 (the Chinese button said about 2.5). Both corrected, and both tune scripts now check the numbers on the panels.
Every Files briefing's source list printed the AARO UAP Records address as aaro.mil/UAP-Records. That hostname has no DNS record at all, so typing it in reaches nothing; the address that resolves is www.aaro.mil/UAP-Records. Corrected sitewide.
On the Chinese pages for the Avrocar, the Casimir effect, electrogravitics, exoplanet imaging, radiation dose and technosignature spectroscopy, the verdict chips in the claims table were empty and all took the first style, which reads as a green SETTLED, including claims the English page marks DID NOT SURVIVE, such as "the Avrocar explains the flying saucer reports". The Chinese tables now carry the English verdicts, and a new build check stops the build when the two disagree.
The Technosignature Spectroscopy bench gave the same gap two contradictory sizes: chapter 6 put oxygen eight orders of magnitude below the halocarbons, and chapter 9 said each halocarbon molecule absorbs about a million times more strongly than oxygen, which is six. The ladder the bench solves from Earth's own limb puts the halocarbons near 1.3×10⁻¹⁸ cm² against oxygen near 7×10⁻²⁶ cm², about seven orders of magnitude apart. Both sentences now say about seven orders of magnitude and about ten million times, the same claim is corrected on the intro page, the catalogue card and the share summary, and the calibration script now pins both wordings to what the bench returns.
The Near Light Speed sky bench gave a star's brightness as the fourth power of the Doppler factor. Surface brightness does rise as the fourth power, but a star is unresolved: its solid angle shrinks as the second power at the same time, and the flux you receive keeps only the square. The readouts, the prose, the story and both renderers now carry the square: the forward sky at 0.9 c reads ×19 rather than ×361, and the tens of thousands claimed for 0.99 c is about two hundred. On the same instrument the warp bubble preset showed an undistorted sky and a time-dilation toll at once; gamma now comes from the optical speed, which is what the copy describes.
The Drake Equation bench described its Green Bank 1961 preset as putting the nearest neighbour near a thousand light-years, with a round trip of two thousand years. The panel prints 830 light-years and 1,659 years, so the text now says about eight hundred light-years and near seventeen hundred. The same bench called our own radio bubble a fifth of a pixel at true scale; at the instrument's opening camera it is about two pixels, corrected everywhere. The Sun's seat moves from two-thirds of the way out an arm to just past halfway, which is the 53 per cent the bench models. The 13,000 to 33,000 light-year habitable annulus was credited to Lineweaver and colleagues 2004, who published the 7 to 9 kiloparsec zone, about 22,800 to 29,400 light-years; the text now says the modelled band is wider and contains theirs.
The Double Slit bench called a C60 molecule a thousand times heavier than an electron, where it is about 1.3 million. Raising the which-path detector was described as collapsing the stripes into two bands; in the far field one single-slit envelope remains, and the Chinese text already said the interference disappears. The baseball line put the fringe spacing 20 orders of magnitude below an atomic nucleus, which is the figure for an atom: one place now reads 15 orders below a nucleus, the other 20 orders finer than an atom.
The Light Bending bench priced a solar mass in an Earth-sized ball at about 85 arcseconds in its lens ladder, where 4GM/c²b gives about 191; the 85 belongs to a different white dwarf, near 0.6 solar masses at a radius of 8,800 kilometres. Jupiter's line moves from about 0.017 arcseconds to about 0.016. The story said Einstein's 1911 equivalence-principle calculation "got the same number: 0.87 arcseconds"; that paper published 0.83 with the constants of the day, so the sentence now says only that he arrived at the same Newtonian half.
The c-Wall ledger said its first tank of fuel buys 0.79 c and that by the tenth you are crawling through the ninth decimal of c. On the bench's own formula an equal-energy first tank buys 0.866 c, and the tenth leaves you at 0.996, the third decimal; the ninth decimal would take more than twenty thousand tanks. The same instrument said a year of ship time under one gravity carries you past 0.9 c, where it reaches 0.775 c: 0.9 c takes 1.4 years of ship time and 0.99 c takes 2.6. Both locales, the story chapters and the methods document now carry the bench's figures, and the step from 0.5 c to 0.9 c, called roughly eight times the first step, is roughly seven (exactly 7.37).
The Dyson swarm ledger described the swarm crossing one per cent coverage as some 10⁵ times humanity's consumption. One per cent of the Sun's luminosity is 3.83×10²⁴ W, which against the bench's own 18.5 TW is about 2.1×10¹¹ times; run the bench's doubling law to year 28 and it stands at 1.9 per cent and about 3.9×10¹¹ times. The line now says 4×10¹¹ and names the 1.9 per cent it stops at. The waste-heat evidence row read 148 to 633 K, but 633 K is the one-faced shell at Mercury's orbit, a figure no readout prints; it now reads 148 to 604 K, peaking at 4.8 to 20 microns, which is what the dial can reach. The coverage slider's caption had no time base and now names the minute it takes.
The Arecibo Message bench carried four slips. The fifty-thousand-year round trip was called longer than Homo sapiens has existed, where it is about a sixth of it. The 52 light-year wavefront was called past every star anyone has photographed a planet around, where Beta Pictoris and others lie beyond it. Proxima was called a quarter of the way to the read range, where it is 4.6 per cent. And the per-bit signal to noise was called one ten-thousandth of what reading needs, which is its absolute value: against the threshold of ten the shortfall is about seventy-three thousand.
The Render Engine bench said its scene holds about twelve thousand triangles while the counter prints 23,449. That total comes from a fixed-seed generator and does not vary. The intro now says about 23,000, and the paragraph that promises the counts stay honest prints the exact 23,449.
The Fermi Paradox map said, in both languages and everywhere it is mentioned, that a 0.1c wave fills the galaxy in about twenty million years. On Hart's own settings the bench divides the distance to its farthest star by the front speed and returns 14.2 million years, and the figures downstream were wrong with it: 0.15% of the galaxy's age is 0.10%, one part in seven hundred is one part in a thousand, the crawl scenario is five times slower and not thirty, the wave needs about 140 times the light-crossing time and not 200, and ten thousand years of refit fills the disk in under 150 million years, not 100. The same panels said the map draws twenty-two hundred stars, each standing for 10⁸ real ones; it draws sixty-eight thousand, each standing for about three million.
The Raptor and Merlin bench printed Merlin's sea-level thrust to weight as 184 to 1 in its comparison blocks and accuracy tables, beside a readout computing 183. 845 kN over 470 kg is 183.3. The published 184 stays in the sourced-data table that says it is published.