On 27 April 2020 the Department of Defense released three cockpit videos and said one thing about all of them. FLIR1, GO FAST and GIMBAL had leaked years earlier, and the release was made “to clear up any misconceptions about whether the footage was real.” That sentence settled provenance for all three at once. It settled nothing else.
This site has built a separate instrument for each of them, and all three were built the same way: take the ordinary explanation seriously enough to compute it, then see what the computation costs. Run that method three times and it returns three different answers. This page is about why, and about why the spread is the interesting part.
Where the tapes come from
FLIR1 comes from the Nimitz carrier strike group off southern California in November 2004. The USS Princeton’s SPY-1B crew, the Nimitz and multiple aircrews describe the same two-week period of intermittent tracks descending from high altitude. Multiple witnesses agree with each other. No radar recordings have ever been released, so the episode around the video rests on testimony.
GO FAST and GIMBAL both come from the Theodore Roosevelt strike group off the US East Coast, during the 2014-15 period when crews reported recurring unexplained radar tracks. The two clips come from the same 24 hours. That is context, reported, and it is not an identification of anything.
Provenance is the one thing all three share, and it is the one thing that is finished.
GO FAST: the case the reported numbers close
The margins of the clip carry ALT 25,010, CAS 250, M 0.61 and a range closing from 4.4 to 3.4 nautical miles, alongside about 15° of bank in the attitude bars and about 26° of camera depression at range lock. Those are what the pod displayed, and the pod’s own accuracy is left open.
One line of trigonometry on the printed values, h = ALT − RNG·sin(EL), puts the object near 13,000 feet rather than skimming the sea. Mick West published it within days of the 2018 leak; the NASA UAP Independent Study Team re-derived it in 2023. String the per-frame solutions together and the track moves at roughly 40 mph over the video’s 22 seconds. The winds aloft that day were never published, so the airspeed floats in an honest band of about 40 to 115 mph, and the instrument hands that uncertainty to you on a dial instead of hiding it.
Size rides on the same assumption as everything else. A roughly 10-pixel blob in the 640-pixel, 0.7° narrow field subtends about 0.011°: bird-sized at the reported range, bus-sized down at the waves. The size slider and the altitude slider are one dial wearing two costumes.
No public identification of the object exists, and the instrument endorses none. A drifting balloon fits NASA’s lean, a large bird fits the speed, a small drone fits the fleet context. What the clip does not leave open is the reading it is named for. The sea-skimmer is retired by the video’s own margins. The video was real. The speed was parallax. Those are two separate sentences, and the whole value of the case is in keeping them apart.
GIMBAL: the case the missing number keeps open
GIMBAL’s symbology prints ALT 25,000, CAS around 240, black-hot polarity, an azimuth counting down from 54 left through zero at about 30 seconds, an elevation of −2°, and a 0.35° field of view. The bank arcs roll from 22.6° to a 38.2° peak. What it never prints is RNG. The pod tracked passively, and the missing range is the axis of the entire debate.
Even the date is contested. It is officially given as 20 or 21 January 2015, while the released file’s own metadata encodes the evening of the 24th, and reanalysis weather data fit that evening’s 120-knot westerly jet better. The instrument therefore says “January 2015” and nothing more.
The rotation that made the clip famous has a mechanical account. A roll-over-nod gimbal’s image-up is the projection of its roll axis, and holding the horizon level costs a derotation. That axis rides the nose, pitched up by about 3.6°/cos(bank); as the recorded azimuth crosses zero at −2° elevation, the line of sight passes under the pole and the roll whips through most of its swing in the final ten seconds. The measured rotation moves in about five discrete jumps with one long hold, each landing on a visible jolt in the background clouds, which is what a heavy roll gimbal parking and catching up looks like. The counter-argument comes from the pod’s own patent: a stepped roll would itself be anomalous. Both positions stay on screen.
Without RNG, every distance is a candidate world. Near 30 nautical miles, a straight, level, constant-heading object threads the recorded fan of sight-lines with an ordinary flying-away speed profile. Anchored at 8 nautical miles, the best straight world has to brake from about 300 knots to a stop. The published close reconstruction, which had radar and size inputs the fan does not, adds a vertical reversal on top.
The clouds stayed level. Something rotated. The glare reading explains the rotation mechanically and asks you to accept an unidentified jet inside an instrumented exercise range. The close reading fits the crew’s testimony and asks you to accept unprecedented kinematics on classified radar. Both are priced. Neither is endorsed.
FLIR1: the case with no decisive toggle
The 2004 encounter has the most testimony behind it and the least data underneath it. Four aviators describe a roughly 40-foot white lozenge with no wings and no exhaust, mirroring them and then outrunning them; the Princeton reportedly reacquired it about 60 miles away roughly a minute later. It is vivid, mutually consistent, argued under oath, and uncorroborated by any released sensor data.
The load-bearing number is the drop: 28,000 feet to the sea “in less than a second,” senior radar operator Kevin Day’s recollection, pinned by the Scientific Coalition for UAP Studies to 0.78 seconds, which is one display interval. A screen that updates in sweeps cannot distinguish a real sub-second transit from a track leaving one altitude gate and appearing in another between updates. The number that carries the entire case is a memory of a display.
What is measured here is the comparison ladder the instrument puts alongside it: 9 g, a fit pilot’s brief ceiling; 46.2 g, John Stapp’s 1954 rocket-sled record and the hardest deceleration a human has knowingly survived; around 60 g for agile air-to-air missiles; and 15,500 g, the firing shock the M982 Excalibur’s guidance electronics are built to survive. Every rung is measured hardware or measured biology.
Feed the reported inputs into schoolbook kinematics and the answer is thousands of g. The SCU’s 2019 forensic paper derived about 5,370 g; this site’s brake profile returns about 5,700 g from the same numbers under slightly different assumptions. Both are honest arithmetic, and both are arithmetic on inputs nobody can verify.
Every explanation on the table survives, because the data that would kill any of them was never released or never existed. There is no pressure dial to pull to the floor, no printed range to subtract. The instrument computes; it cannot close.
What the three share, and what they don’t
All three share the same settled provenance, and all three are handled the same way: the ordinary explanation is computed rather than asserted, and every line is labelled by what it actually is, reported, measured, modelled or a reading.
What differs is what the computation returns.
- GO FAST. The reported numbers retire the dramatic reading themselves. The remaining question, balloon or bird or drone, is ordinary.
- GIMBAL. The mechanism is reproducible, but one number was never recorded, and its absence leaves two worlds alive. Each of them costs something a reasonable person may decline to pay.
- FLIR1. The load-bearing input is testimony about a display. No amount of correct arithmetic can close it.
That spread is the finding. “The videos are real” is one sentence covering all three. “We know what they show” is three different sentences, at three different stages of being finished, and only one of them is close.
Sources for this page
- Department of Defense, release of FLIR, GIMBAL and GO FAST, 27 April 2020
- Scientific Coalition for UAP Studies (2019), forensic analysis of the 2004 Nimitz encounter
- NASA UAP Independent Study Team report (2023)
- The instruments on this site: Tic Tac Kinematics, GoFast, Gimbal