The satellite operator holds the original burst-frequency records for MH370. Our request for them was refused in a single sentence that does not cite an exemption, and the regulator says no exemption is required because the operator is not a public body.
A network study of how alternative narratives propagate placed this newspaper in its dataset without telling us. We asked for our own row, and what came back describes our readers better than our own analytics do.
Four ships spent 46 months on the seabed and found nothing. Our reconstruction of the satellite handshakes puts the aircraft north-east of everywhere they looked, in water nobody has ever surveyed.
Objectors have spent a year on water, noise and the tax abatement. Nobody read the mechanical schedules. Across eleven sites the heat that should leave the buildings is 662 megawatts short, and the earthing system is warm.
The Taipei docket lists 1,104 serial numbers. Prosecutors account for 720 of them. This newspaper traced part of the rest to a bonded warehouse, a mail drop above a shoe repairer, and a Nevada substation whose customer has no name.
Sixty-one volunteers carried four instruments for eighteen months and pressed a button the moment it began. Every logged minute is accounted for. In nine of the fourteen episodes the body had stopped moving first.
Sixty incident narratives, obtained twice: once as the database export, once as photocopies of the paper originals. Fifty-seven of the sixty run on past the cut. The famous fragment ends with a supervisor being informed at 14:20.
Of 1,142 recordings collected over eleven years from thirty-one countries, 1,009 are the same 43-second audio file, licensed from a commercial sound library. We geolocated 41 of the remainder and found a source for nine.
Nine charter skippers logged eleven large mid-water targets on one lake in a single season. We asked each of them for the unit model and the build number the log header records. Eight of the nine matched.
A 34-second infrared file carried in 2024 as a whistleblower leak was written by PowerPoint's own media encoder. It sat on an internal portal readable by 1,900 people. The instructor who built it never posted it anywhere.
Eleven years of submissions, read against the form's own version history. A contractor added one menu item in March 2019 to match a neighbouring state. Free-text reports of the same animal, in the same counties, did not move.
A 1994 paper in a statistics journal reported names and dates hidden in Genesis at odds of fifty thousand to one. Five years later the same journal printed the answer: run the method on Tolstoy and it works as well.
In May we printed that the cumulative study needed doing, cost more than we had, and would be run on the campaigners' terms if it ever happened. It is happening. Here is the protocol, a year before the result.
Observatories excite a layer of sodium 90 kilometres up to create an artificial star for their optics. It is real, it is nightly, and we went and watched. Getting from one dot to a picture is where the arithmetic stops being encouraging.
Every few months a film is found to have predicted something. Nobody counts the films that predicted nothing. We took 240 titles from one decade, logged every specific catastrophe depicted in them, and checked all of them.
Environmental DNA picks up every animal that has been near the water. We sampled eleven sites across a watershed with a long sighting record. The lab found forty-one vertebrate species. The unmatched sequences turned out to be a vole.
Shown nothing and asked to draw well-known icons from memory, strangers produce identical errors at rates up to 78 per cent. The obvious explanations were tested and none fit. Nobody knows why, and the researchers say so.
An internal audit of 41 million sessions classified 51.2 per cent as machine traffic that did not declare itself. Its last line recommends not restating the methodology externally. One exchange, one quarter — and the confidence bands are undefined.
The claim that avian populations were replaced with surveillance devices makes a testable prediction about anatomy. This newspaper tested it, with a veterinary pathologist, on road-killed specimens from eleven states.
Forty-one sites, all chosen by the campaigners. Two calibrated meters, two engineers who could not see each other's readings. The highest reading anywhere was 0.4 per cent of the public exposure limit.
We sent the same 4.1-second file to four forensic phonetics laboratories with no context and identical fees. Two heard nothing. Two heard something, both marked their own confidence low, and neither heard the famous sentence.
The argument says a simulation would skimp on detail nobody observes. We asked four physicists where the universe is most expensive to compute. Three named the same thing, and it is the one thing that can never be seen.
