Orbs Launching Orbs: We Read the Whole AARO File. One Agent Says There Is Video.

On October 13, IBTimes UK ran a headline saying the Pentagon had released 155,000 pages of UFO documents. It spread fast. It is also wrong.

The 155,000 figure belongs to The New York Times. In an October 2 story, Danny Hakim’s team said it had assembled roughly 155,000 unique pages of records from seven countries. That is a newspaper’s research pile, not a government release. The Pentagon’s actual program, the Presidential Unsealing and Reporting System for UAP Encounters, or PURSUE, has published six tranches since May 8. We pulled every document bundle from war.gov/UFO onto our own machine and counted 270 PDF files.

Both stories lean on the same case, though, and that case is worth the attention. It is the one AARO calls “Orbs Launching Orbs.” The government has published a signed analysis memo, five first-hand witness statements, a map and a set of AI illustrations about it. Most coverage has quoted a few lines from them. We read all of them, page by page.

Update, October 23, 2026. PURSUE Release 07 adds a second western U.S. case at a sensitive national security site: a silent black triangle reported by federal special agents in June 2026 (DOW-UAP-D176 and FBI-UAP-D043). AARO says it closely matches the “morphological features and kinematic profiles” of other reports from the region. Release 07 still contains no video, audio or photograph from this case or the 2023 orbs case; the FBI file’s images are reconstructions drawn on a photo of the terrain. Our read of the release.

What the file is

The case first appeared on May 8, in PURSUE Release 01, as a four-slide summary titled “Western US Event.” The full paperwork followed on June 12 in Release 03:

• DOW-UAP-D077, a four-page Memorandum for Record from the All-domain Anomaly Resolution Office, signed by AARO Director Jon T. Kosloski on 5 June 2026.
• D079 through D083, five free-form narratives from Witnesses 1 to 5, each dated 2 June 2026. Six agents saw things. Five statements were released.
• D078, a “notional” map of four incidents. Its own caption says all images “were artificially generated using https//genai.mil/” and that the locations “do not accurately represent the relative position of the observers or reported phenomena.”

The setting, assembled from the statements: a valley ringed by mountain ranges, a “sensitive national security site” in the western United States, an airfield, a gate, a base. The agents were not out stargazing. Witness 2 says the job was “to survey an area for individuals launching drones.” That detail matters later.

What the agents say they saw

AARO’s summary is careful and short. Over two days in October 2023, six federal law enforcement special agents, working in teams of two, “reported observing ‘orbs launching other orbs,’” at about dusk. “A luminous orange ‘mother orb’ appeared to produce smaller red ‘orbs,’ one after another, multiple times over a period of several hours.” The orange orb showed for one or two seconds, released two to four red ones, and vanished. The red lights moved in what looked like coordinated formations. The agents called it all “silent.”

The statements are stranger, and more human.

Witness 2, the longest account at eight pages, describes the first night. A light low over the northern horizon, at first “not dissimilar to that of a luminescent planet,” grew brighter and “expelled three or four red lights from it. The red lights accelerated instantly and maneuvered with perfect, smooth coordination into a horizontal formation.” The process repeated about five times in half an hour. The last group of red lights drifted to the southern horizon, and from about 7:30 to 11 pm “hovered above the southern horizon. The red lights sometime dipped below the horizon, out of sight but consistently rose back up.”

Witness 2 reaches for comparisons, and flags them as comparisons. The event “looked like a mix of mechanical deployment and biological division.” The opening and closing of the bright light was “not dissimilar to the opening and closing of a portal,” though “I am not saying or suggesting that was what we observed.” The red row on the second night was “reminiscent of the Phoenix lights.”

Witness 3 gives the cleanest image of the launch itself: the red orbs “were like grapes being expelled from a basketball.” The expulsion “appeared to be mechanical,” and there were “no exhaust trails coming from either the orange or the red orbs.” On a second outing Witness 3 watched four red lights stop “above an airfield and hovered in what appeared to be a square over each corner of the airfield.”

Our AI illustration of the dusk launches as Witnesses 2 and 3 describe them. Made from the statements, not from any photograph. Not evidence.

Witness 4 saw the same pattern, in different words: “multiple singular very bright orange lights that appeared to shoot out smaller orange lights as if being hatched.” Then, in the next sentence: “The smaller lights seemed to travel downwards like a falling flare.”

Keep that sentence in mind.

The “large, fiery orb” Witness 1 saw against a hillside: “a circle of light that looked like a swirly pattern of bright lava,” with what “almost appeared to have a line or spindle below.” AARO later put it about 1,050 metres away and 12 to 18 metres across. AI illustration, not evidence.

The car that was not a car

The orbs are the headline. The ground encounters are the part of the file nobody seems to have read.

Witness 2’s second incident is the one IBTimes excerpted. Two apparent headlights sat still in the valley five to seven miles away. They flashed “toward us; the flashing appeared to be consistent with a white pulse.” A second pair appeared and flashed the same way. The agents drove at them with their own headlights off, on night vision, at one point near 70 miles per hour. The headlights “morphed into one bright white/orange orb-like light with a halo-like glow.” At about 50 yards, “the object then accelerated instantly to the south.” Later it reappeared as “a group of orange and purple lights in the shape of a box (not dissimilar to a semi-truck in shape and opaqueness)” and “covered approximately four miles of rugged terrain within approximately one minute.” That works out to around 240 miles per hour, if the distance estimate holds.

The agent’s verdict: “The objects initially appeared to mimic vehicles (my best example is the flying car from the Harry Potter book series).” The object “appeared to ‘toy’ with us.”

Witness 2’s chase, lights off and on night vision, as the “headlights morphed into one bright white/orange orb-like light with a halo-like glow.” AI illustration, not evidence.

The “box of lights” that crossed “approximately four miles of rugged terrain within approximately one minute.” AI illustration, not evidence.

Witness 3 describes the same trick from another team. A vehicle below them was “flashing its headlights at a slow, methodical tempo” while red lights hung over the airfield. “It seemed evident that it was involved with whatever was happening.” They raced down the hill, lights off, and “could get no closer than what seemed to be about 100 yards.” Then the vehicle “seemingly floated off the road to the south, across rough terrain. There were no roads and later we searched for tire tracks but didn’t find anything.”

Then the pre-dawn hours. Witness 1 and Witness 5 were partners, and they tell the same story separately. Around 4:30 am, on a road near a lake on the way to the base, they closed on what they took for a car with a broken taillight: one light red and round, one white and square. As an oncoming car approached, the pair of lights “changed direction and moved off the road to the right.” Witness 1: “It did not bounce or move with the terrain, but seemed to hover. It did not kick up any dust either.” Witness 5 watched it through night vision until the lights went off and “a very thin line remain[ed], maybe 4 ft wide.” And then, from Witness 1, the line that has no place in a flare report: “It reminded me of a Toyota Prius for some reason.”

The “dark kite”: one red round light and one white square light, two to three feet off the ground, that slid off the road “with zero resistance.” AI illustration, not evidence.

An hour or so later, back at the spot with thermal binoculars, they saw “dust and thick spider web looking material floating in-front of us.” Then “what looked like the same object floating in a more vertical orientation.” Witness 1: “as it floated past I could see a star through the object, as well as the crest of the hill and some vegetation. It was almost like looking at something through water.” The Release 01 slide adds a detail from a spotlight: “at one point my beam went from shining far into the distance to stopping about 50 yards away on nothing in particular.”

The “transparent kite” Witnesses 1 and 5 describe, with a star showing through it and “thick spider web looking material” in the air. A third colleague with them did not see it. AI illustration, not evidence.

When they came back in daylight, Witness 5 noted “odd scrapes in the ground around a certain plant,” fresh and “symmetrical and round,” on a night with no wind.

Two cautions belong right here. This is one team of two, not two independent witnesses. And a third colleague who came out with them that morning, the slides say, “did not see the object.”

What AARO actually concluded

The memo is better than most of the coverage of it. It runs through seven hypotheses and grades each one.

Hypothesis	AARO’s grade
Military aircraft exhaust	Ruled out. Aircraft were “too high.”
Drones	Unlikely. Hours of loitering exceeds battery limits.
Military aircraft dispensing flares	Partially plausible. About 60 percent.
A U.S. program (Blue Force)	Plausible, inconclusive.
Foreign intelligence (Red Force)	Highly unlikely.
Weather, atmosphere, stars, satellites	Unlikely.
Unrecognized technology	Pending. Up to 40 percent.

The flare finding is the heart of it. Flight logs confirm U.S. military aircraft in the area “were equipped with, and actively deployed, infrared countermeasure flares as a component of a standard exercise.” The position, direction and timing of the reports “strongly align with historical radar and ADS-B data.” The agents, AARO notes, “are professionally familiar with the visual characteristics of military flares” and said this was not that. AARO’s reply is quiet and important: the flares those aircraft carried “differ from those of standard illumination flares.” In other words, the agents may have been comparing the lights to the wrong kind of flare.

The witnesses’ own words point the same way. Witness 4: “like a falling flare.” Witness 1 and Witness 5 each report that when an orange orb lit up and released red ones, there was a white flash on the horizon at or around the same moment. Witness 1: “At the same time the orange light lit up, there appeared to be a flash from over the horizon to the left.” That is what you might expect if something over the horizon was dispensing and burning flares.

Then the residue. For the remaining 40 percent, “radar and ADS-B data indicate that no known aircraft were active within the observers’ estimated line-of-sight.” And the red light that stayed put above a ridgeline for several hours “is physically incompatible with the burn-time and descent rate of any known military flare.”

Two things are worth weighing against that. First, the hours-long red light appears to come from Witness 2’s account of lights that “sometime dipped below the horizon, out of sight but consistently rose back up.” Distant lights near the horizon that drop and return are not obviously a single stationary object. Second, military aircraft on exercise do not always broadcast ADS-B, which is why AARO pairs it with radar. AARO had both and still found nothing in line of sight.

AARO’s Blue Force paragraph is the most revealing in the memo. It says the reported characteristics “align with those of certain U.S. military technologies,” but “historical records are inconclusive regarding whether those technologies were present,” and “no single Blue Force capability fully accounts for all the phenomena’s reported characteristics.” Read plainly: some of this looks like our own hardware, and the paperwork cannot prove it either way. At a sensitive site with an airfield, during an exercise, while agents were hunting drone operators, that is not a small caveat.

Finally the line that will get quoted most. “AARO’s preliminary exclusion-based hypothesis is that … unrecognized technology may account for up to 40 percent of the phenomena.” The same sentence goes on: “this provisional assessment is based solely upon narrative testimony, as well as the elimination of competing hypotheses, and is unsubstantiated by technical data or physical evidence.” Unrecognized means unrecognized. It does not mean non-human, and AARO does not say it does.

The video line

Section 3a of the memo states: “The reporting agents did not collect video footage, photographic imagery, or other technical data during the incident.”

Witness 2’s statement, published in the same release, describes a third encounter. An orb-like light sat on a rural highway, then glided down the road for 10 to 15 miles while the agents drove 80 miles per hour and could not catch it. A semi-truck came around a bend at them as the light vanished. And then, on page 4: “In the video footage, I believe we say something about thinking it was going to hit us.” And a few lines later: “The video footage captures some of it, but more valuable is the audio of our reactions.”

Witness 5, meanwhile, writes that about three miles from their vehicle, “we set up a camera in hopes to catch photos if anything returned to sight.” The statement never says what, if anything, the camera caught. The same team also used thermal binoculars that morning.

There is a charitable reading. AARO’s memo is about “the incident,” the dusk orb launches, and Witness 2’s video may come from a different night of the same deployment. If so, the memo is technically right. But the government has released a statement in which one of its own federal agents says video and audio exist from that deployment, and has not released them or said why. That is the single most concrete thing in the file, and no outlet we have seen has mentioned it.

The pictures are not pictures

Every image attached to this case was generated by AI. The map says so. Witness 2 says the sketches were made from “prompts and descriptions I provided” and that “these are being generated two and a half years after the events.” The narratives themselves are dated June 2026, written for AARO long after the fact. Memory does not improve with time, and AI image generators fill gaps with whatever looks plausible. Treat the images as illustrations of what the agents remember, and nothing more.

The only moving images of the case are no different. PURSUE Release 03 includes two clips tagged to it, FBI-UAP-PR005 and FBI-UAP-PR006, titled “Digital Recreation, Narrative Statement 3-1” and “3-2.” Their official description calls each one “an artistic interpretation” based on “a first-hand description provided by a federal law enforcement special agent,” prepared by the FBI “at the request of the Department of War in 2026.” The numbering appears to match Witness 3’s two sightings. They are animations, not footage. We listed the contents of all six PURSUE video bundles, 150 clips in all, and found no recorded video of this case.

FBI-UAP-PR005: FBI digital recreation of Narrative Statement 3-1, Western U.S. Event, 2023. An animation made in 2026, not footage.

FBI-UAP-PR006: FBI digital recreation of Narrative Statement 3-2, Western U.S. Event, 2023. An animation made in 2026, not footage.

The same goes for the illustrations in this article. They are ours, made with AI from the agents’ written words, and each caption says which statement it draws on. None of them is a photograph of anything.

What would settle it

The case is unusually testable, because the government already holds most of what would decide it.

1. Release the video and audio Witness 2 describes. If it exists, it is the only recorded data from the whole deployment.
2. Release whatever Witness 5’s camera captured, even if the answer is nothing.
3. Name the flare type, or at least describe it. AARO’s case for 60 percent rests on the claim that these flares look unlike illumination flares. Show how.
4. Close the Blue Force question. AARO says the features match “certain U.S. military technologies.” Either those systems were there or they were not. That is a records question, not a physics question.
5. Release the sixth agent’s statement, or say why it was withheld.

Until then, the honest summary is the government’s own: a well-documented, multi-witness report, most of it plausibly explained by flares, a residue that nobody can explain from the data, and no technical evidence for anything exotic. Unresolved means unknown. It is not a synonym for alien, and it is not a synonym for debunked.

About our crawl

For this piece we downloaded all six PURSUE document bundles from war.gov/UFO, 270 PDFs in all, and ran every file through a reading and scoring pass on our own hardware, then checked the quotes above against the original pages by hand. The Western U.S. Event file came out near the top. A few others that deserve their own write-ups: a 1968 CIA report of a “huge metallic disc-shaped object” found in a crater near Pokhara, Nepal; a 1971 Australian defence intelligence paper alleging the CIA used Project Blue Book to publicly “debunk” UFOs; and a 1948 Air Force Top Secret file quoting Swedish intelligence on objects “which cannot be credited to any presently known culture on earth.” We will get to them.