PURSUE Release 05: A Gunship Crew Watched 'Cold Orbs' Dodge Its Cannon, and the Recorder Failed

The fifth PURSUE tranche arrived on August 7, 2026: 41 files, split as 22 documents, 16 videos and 3 images, from the Department of War, the FBI, the CIA, the State Department and, for the first time, the Executive Office of the President. The Pentagon says more are coming on a rolling basis.

Most outlets led with the picture: 99 seconds of electro-optical and infrared sensor video from a U.S. military platform somewhere in the Middle East in 2025, an object held in the crosshairs while the sensor changes modes and magnification. It is worth watching, and it is not the most interesting thing in this release.

The most interesting thing is a report about footage that no longer exists.

The gunship and the cold orbs

The document is DOW-UAP-D101, an Intelligence Information Report declassified on March 20, 2026 by the USCENTCOM chief of staff. It describes the night of September 8, 2021, over the Gulf of Oman, where a U.S. Special Operations AC-130J gunship was running a live-fire training sortie in the dark.

The crew dropped a water-activated flare to use as a mock target and put the aircraft into an orbit around it. The Combat Systems Officer swept the infrared sensor around, came back to the flare, and, in the report’s words:

they observed two approximately four-foot diameter cold orbs between 0 to 20 feet above the surface of the water stationary over the flare for approximately 15 seconds. After those 15 seconds passed, the CSO pulled the trigger firing the aircraft’s 105mm cannon. Immediately upon pulling the trigger, during the slight delay between trigger pull and the cannon’s recoil, the CSO observed the pair of UAPs rapidly fly away from the flare and out of the sensor feed without changing altitude.

Read that timing again. The objects left in the gap between the trigger pull and the gun going off.

Over the rest of the sortie the crew logged roughly 25 instances of the same thing, cold orbs in infrared, sometimes two of them in what the report calls a loose but coordinated formation, rising and falling like a wave while moving around, over and under each other. The observer’s own comparison, preserved in the file, is to dolphins swimming in a pod, followed by a parenthetical that is pure interpretation and should be read as such: “They [The UAPs] knew they were flying together.” Elsewhere: a formation of three, and objects “maneuvering aggressively.” The Combat Systems Officer used the aircraft’s targeting tool to estimate speeds between 250 and 1,300 mph. The report notes, flatly, that no means of propulsion were observed.

And then the part that will keep this file in circulation for years:

The AC-130’s crew attempted to save entirety of the EO/IR footage of the UAP observations but the aircraft’s Digital Video Recorder (DVR) inexplicably failed and the video data became corrupted.

Which is why the video looks like this

Six clips in Release 05, DOW-UAP-PR117 through PR122, come from that night. None of them are sensor data. Every one is a cell phone pointed at the gunship’s infrared display, which the Department of War states plainly in each file description: a secondary capture, with limitations in fidelity and resolution, and artifacts from filming a screen.

The clip embedded at the top of this page is PR121, the longest of the six at 34 seconds, and the government notes it overlaps heavily with PR122. What you can actually see is a small dark shape against a bright field, because in black-hot infrared a cold object reads dark. The reticle drifts. The shape holds near the center, exits the bottom, comes back from the upper right.

That is the whole evidentiary situation: a detailed, specific, first-person intelligence report, and six phone videos of a screen. It is a good reminder that “the Pentagon released video” and “the Pentagon released data” are not the same sentence.

How to be careful with this one

The Gulf of Oman file is the strongest-sounding narrative PURSUE has published so far, which is exactly why it deserves the most friction:

• An IIR is not an assessment. The form itself carries the warning: this is an information report, not finally evaluated intelligence. It is raw reporting from a single source, disseminated so that analysts can weigh it later. Nothing in the file says anyone did.
• The speeds are estimates. 250 to 1,300 mph came from an officer working a targeting tool against objects of unknown size at unknown range, at night. Size, distance and speed are one equation with three unknowns; assume the wrong size and every other number moves with it.
• The reaction to the cannon is a correlation. One event, observed once, with no instrumented record. It is a genuinely arresting detail and it is also exactly the kind of detail that memory sharpens over five years.
• “Cold” is a real observation. That part is harder to wave off. An object that reads colder than ambient in infrared is not an engine plume and not a flare, and low-temperature is an odd property for most of the usual candidates. It is also consistent with something small, slow and passive much nearer the sensor than assumed.

For the deflationary case done properly, see Avi Loeb, the astrophysicist chairing the White House UAP Science Advisory Council, who this week worked through an earlier PURSUE clip, DOW-UAP-PR043, using terrain matching to pin down scale and velocity, and concluded that “a fast tactical missile is physically compatible with the data.” That is what analysis of these clips looks like when someone actually does it, and it is a standard the Gulf of Oman phone footage cannot meet, because there is nothing in frame to measure against.

The rest of the tranche

Eleven of the sixteen videos are two incidents. Beyond the six Gulf of Oman clips, five more (PR123 through PR127) come from a single 2019 Pacific Ocean event handed to AARO by the Navy’s UAP Task Force in 2022. The government adds a caveat worth repeating: that media “was digitally altered before being reported to AARO, and is presented as it was received.” The longest is a two and a half minute handheld clip of multiple light sources in low light. Counting videos is not the same as counting incidents, and headlines that say “16 new videos” are quietly inflating a much smaller number of events.

The triangle over Bagram. FBI-UAP-D024 is an FD-302 recording a March 2024 interview with a former military pilot. Most of it concerns white lights he and other airline crews have seen since October 2023, at high altitude, at distances he guesses at 200 to 500 miles, which is a category where prosaic explanations (rocket launches, satellite constellations, upper-atmosphere effects) are strong even though he rules several out himself. Then there is the 2002 memory, from a night at Bagram in Afghanistan:

he noticed that the stars were getting blacked out over Bagram. It appeared as if something huge went overhead, from west to east, and was an equilateral triangle shape. The object was moving silently, at a consistent speed and altitude, possibly 150 knots, with no lights visible. Estimated it was 500 feet in size

The second pilot saw it too. His entire recorded contribution is “did you see that?” Two witnesses, no instruments, twenty-two years before the interview. Note also that four of this release’s images are FBI digital renderings, artists’ interpretations built from witness descriptions, not photographs. If you see the dark triangles circulating as “newly released FBI images of a UFO,” that framing is wrong, the same way it was wrong in Release 03.

A 1964 radar track off Puerto Rico. CIA-UAP-D022 and D023 cover a November 19, 1964 incident in which a Navy aviator described a “delta-shaped” object “about the size of a jet fighter” and estimated Mach 3 by eye, while the AN/SPS-49 radar aboard the destroyer USS Gyatt held a contact at more than 3,800 knots for 21 minutes. CIA’s own scientific-intelligence side leaned toward foreign reconnaissance of the Atlantic Fleet Weapons Range rather than anything exotic. A 21-minute radar track is a substantial record by 1964 standards, and it also sat in a filing cabinet for sixty-two years.

Sweden’s ghost rockets, and the political use of a bad explanation. DOW-UAP-D099 is a 1947 War Department intelligence review of the 1946 Scandinavian ghost rocket wave. Its conclusion is mundane, that most reports were misidentified natural phenomena with a residue of real low-flying V-1 type missiles from Soviet tests of captured German technology. The interesting part is institutional: the review records that the Swedish Defense Staff privately blamed Soviet missile tests while publicly shifting to a “celestial origin” explanation, and assesses that the shift was a deliberate move to avoid a diplomatic fight with Moscow. Governments have been choosing the convenient sky explanation for a long time.

The Brazilian sphere that wasn’t

The three-agency thread in this release is the best thing in it, and almost nobody covered it.

In November 1963, Rio de Janeiro radio reported that a “large metal sphere” containing a dead occupant had crashed at Conde, in Bahia. A CIA wire service report of that broadcast reached Edward Welsh, executive secretary of the National Aeronautics and Space Council inside the Executive Office of the President (EOP-UAP-D001). Handwritten notes on it, dated November 13, record Welsh’s office asking the State Department to have the embassy in Rio check it out.

The embassy cabled back the next day (DOS-UAP-D001), repeating the press language it was chasing, a “large metal manned balloon” and a “dead crewman aboard wearing uniform or spacesuit with strange markings,” and reporting that a weather balloon had been reported in the area but nothing had been found crashed. Six days later the scientific attaché in Salvador closed it out (DOS-UAP-D002): no contemporaneous local reporting, contradictory newspaper accounts after the fact, and the Brazilian Ministry of Aeronautics denying categorically that there was “any ‘strange object’ in Conde.” The finding was that the story was a fabrication that originated in Rio.

That is a crashed-saucer-with-a-body story, tracked from a radio broadcast to the White House to an embassy to a consulate in six days, and killed with evidence. It belongs next to the Varginha file and the rest of Brazil’s UFO record, and it is worth keeping in your pocket the next time someone insists that governments never check these things or never publish the answer when the answer is no.

Where this leaves the program

Five tranches in, the pattern holds. PURSUE is not producing proof; it is producing paperwork, and the paperwork is getting more specific. Release 04 gave us a nuclear-weapons plant on lockdown over something nobody identified. Release 05 gives us a gunship crew describing objects that moved when the gun did, and a video recorder that failed the one time it mattered.

Both files are unresolved in the only sense the word carries here: the government looked and could not say. What Release 05 adds is a lesson in what these records are worth. The most vivid claims in the tranche sit in an unevaluated single-source report with no surviving instrument data. The most solid work in the tranche is a 1963 story about a metal sphere and a dead alien pilot, which three arms of the U.S. government chased down inside a week and found to be nothing at all.

Keep both. They are the same program telling you how to read it.