After a discussion with @MiddleEastBuka and other researchers, I highly suspect that this video of an explosion is an AI-spliced video, created using the video on the left - and likely not the first of its kind.

Why?🧵 This video has several oddities, particularly in how the environment reacts to the explosion:
1. Both cars shift position, with the white car moving a fair distance from where it originally stood. However, the movement is unnaturally smooth, with no windows shattering and no visible damage to either vehicle.
2. The two motorcycles next to the soldier simply disappear completely, leaving only some sort of structured object just out of frame (I can't make out what it is)

This isn't real footage, it's an AI-spliced video based on a real image from January 8th, 2026.

1. The AI artifacts:
A. Powerlines are floating, and its pole disappears with the explosion.
B. There is a random puff of smoke from one of the individuals standing before the explosion.

2. AI-Splicing evidence:
A. All of the people are in the exact same positions as the old image.
B. The AI messed up the front gate, with it adding/removing features not present in the real school gate (Compare red rectangle).

https://x.com/talhagin/status/2033362029617054148

I recognized the origin of this footage because I had geolocated it back in January:

https://x.com/talhagin/status/2016060309627195675

This image is currently being reported as visual "proof" that a misfired AD system was responsible for the strike on a building reportedly housing a girls’ elementary school in Minab.
It's not.

That image was taken in Zanjan, over 1.3k kilometers from the school in Minab.
A 🧵
As showcased, jointly with @Stinky915846091, the misfired AD system photographed was North-West of Zanjan, towards the mountains.

There is no possible way this could have hit the school, 1,300 kilometers in the opposite direction.

Ironically, this post is propaganda - Those images are from 2 entirely different cities.

1. The "Zoom Out" video was taken in Tehran (Geo: 35.726468, 51.322978).
2. The "Zoom In" video was taken in Mashhad (Geo: 36.327939, 59.498942).

How did I geolocate them? A 🧵
@GeoConfirmed 1. First, I used Google Lens on both images to see what results I could get to confirm whether the OP was being truthful or not.
I found 1 X account claiming the "Zoom Out (Now ZO) video was from "Kashani Street". For the "Zoom In" (ZI) video, I saw a comment saying "Mashhad".

AI detectors, including those using LLMs, consistently fail to reliably identify AI-generated content, often producing false positives on real images.

But there is an effective, though limited, way to identify AI-generated content using tools like SynthID.

What is SynthID? 🧵 Background:
An AI-generated image claiming to show Maduro in U.S. custody was spread widely on X, believed by many to be authentic.
I debunked it primarily using SynthID. Some questioned why I used AI to debunk AI.
I didn’t - Well, not the way you may think.

https://x.com/talhagin/status/2007417679712559604?s=20

AI-Community Notes are wrong (Again) - This photo is not from the Azovstal steel plant in Ukraine.

It's from Gaza, and I managed to geolocate the rough location of the camera in Beit Lahia.

How did I find it? A Guide🧵
(H/t @Stinky915846091) The so-called “proof” that this originated in Ukraine relies on AI-generated citations, but there also CNs by real people (but unapproved).

None of the images or videos presented in those searches match the screenshot in question - not a single one.

Claiming that they “look similar” is insufficient; many power stations share comparable structural features. Similarity alone does not constitute evidence.

Remember the fake site built to spread disinformation?
It claimed the bystander who disarmed one of the shooters wasn’t Ahmed al-Ahmed, but “Edward Crabtree.”

Let’s break down everything wrong with the site - 🧵

https://x.com/talhagin/status/2000286750023745867

First of all, the site is "attempting" (very badly) to mimic the real news site, "The Daily Aus" - It just got literally everything wrong:

Based on the available visuals, it appears (Based on current video evidence) that both gunmen arrived prepared with two firearms each: two hunting rifles and two shotguns.

A Timeline 🧵

https://x.com/talhagin/status/2000182712552821075

Current visual information:
1. In the phone and drone footage, one of the individuals who is shot appears to be holding/laying by a firearm (likely a rifle) near the back right side of the vehicle at the moment they are hit. The exact weapon type cannot be confirmed from the footage alone.

2. The vehicle seen parked next to the bridge matches a car reportedly associated with one of the shooters’ home addresses (based on the license plate allegedly linked to him).

3. In the phone footage, a visible interaction visible is seen between one of the victims and a shooter.

Based on the video, after shooting at two individuals in close proximity, they direct almost* all subsequent gunfire toward the south-east.

Their immediate line of sight, shown by the red movement line, points south-east toward the gathering point of the Hanukkah celebration, marked by the white circle.

https://x.com/talhagin/status/2000152801574355160

(Per my analysis of the video) The shooting does not appear to be random or indiscriminate.
The two individuals were likely shot due to their close proximity or an attempt to intervene. At 1:31, the shooters can be seen actively warning people to the north, where filming is taking place, to move away.

What do you get when you mix a redacted photo from the Epstein estate with AI-enhancement tools?
Nothing good, or real.
A 🧵

https://twitter.com/nypost/status/1999624251817497061

The original image was released with the women’s faces blurred. Naturally, people wanted to know who they were, and some turned to AI to “enhance” the image.
edition.cnn.com/2025/12/12/pol…

It will never cease to amuse me that the agenda or circumstances surrounding any major event will always fall directly in line with whatever core belief a “truth seeker” holds.

They begin with a conclusion, and just jerry-rig “evidence” to fit it.

Wanna test it yourself? Go to any self-proclaimed truth seeker and figure out their core narrative:
1. Hatred of left/right wing people
2. Hatred of a religious/ethnic group.
3. Hatred of a racial group.
4. Or the recurring hits: “This is a conspiracy,” “False flag,” “It’s staged.”
5. Etc

On Jan 5, 2025, @elonmusk said:
“Grok will be the best source of truth by far.”

Beyond this, users have relied on the X LLM, known as Grok to verify information, even attempting to use its replies as Community Notes.

But can we really rely on Grok?🧵 We mostly see Grok being tagged in comment sections, being asked:
“Grok, is this true?”

This tells us something really important:
People want to verify information they see on their feed.

In theory, that’s great - People aren't blindly believing posts. But there’s a big catch.

I've been seeing people pick & choose which phrases to show that were written on the Minneapolis shooter, depending on the narrative they are trying to formulate.

Here is a complete list of every(122) phrase written on the weapons and magazines, and their (likely) meaning: Note:
After reviewing nearly everything I could (some words were difficult to decipher or translate from Russian*), it appears that the shooter was fascinated by what many would describe as “Internet culture”—albeit its more extreme end.

One of the reasons I follow a wide range of accounts across the political spectrum is to gain insight into how each side views the same event, and hints that misinformation is being spread.

Take this incident that occurred outside an Egyptian government office in New York yesterday: These 3 posts all appeared on my timeline in one continuous scroll. However, you'll notice that they cover the same exact incident differently:

This photo is from a Halloween party in Thailand in 2022.

https://twitter.com/HilzFuld/status/1958039617061679247

Source:

https://x.com/talhagin/status/1794679517086429441?t=_IspFgfxFt8DqCIgfYQ8yA&s=19

The message isn't fake, but deleted - There are 2 ways to prove this:
1. Forwarded Messages
2. Link ID

A short🧵

To Clarify: This thread is strictly about verifying a specific claim, and to help others with methods to verify the existence of deleted Telegram messages. 1. Forwarded Messages - While the message was deleted in the main channel, it wasn't in the discussions "Group".

How to find it: Go to Al-Sharifs Telegram > Click on "Leave a comment" under any post > Then "Join Group" > It'll appear in your Chats > Click on it and search "9 ساعات" > It'll show up for you - Written and forwarded by Al-Sharif to that group.

1. Link to the "Group Chat" - t.me/anas1020304050…

2. Link to Al-Sharifs Channel - t.me/anas1020304050

Deepfakes are commonly based on existing footage - This one is based on a Subcommittee hearing from 2022.

But, if you didn't know how Omar sounds or weren't aware of the "Meme" reference, it's very possible you'd fall for this.

Here is the original footage:

https://twitter.com/BGatesIsaPyscho/status/1951945171144163598

What the Deepfake does:

It uses existing footage, learns the movements of the target, and then includes fabricated audio and lip movements on a frame taken from the original clip - for this, it used the following photo that was shared a lot:

This particular video has been accused by many of being AI-generated, either citing an anomaly on the individual's hand or the use of AI detectors.

So is real or fake? I believe it to be authentic.

Why?🧵

https://twitter.com/RamAbdu/status/1951725979644330462

The main anomaly mentioned by users is the person's right hand, which appears to have a sixth finger growing on his middle finger.
AI still tends to mess up this type of detail, so it immediately resulted in the video being labeled as such.
(Using the evidence links provided in the proposed CN link)

As I've said on several occasions, no source? Ignore.

Grok said that the following image was from Iraq in 2014, in an attempt to debunk the claim that it was a recent photograph from Gaza in 2025. The issue? It provides 0 source.

Also, Grok is wrong - It's from Gaza.🧵
All it took to make Grok backtrack its claim was asking for evidence and providing it with 2 angles of the girl - Took 3 replies.

See our full conversation here:
x.com/i/grok/share/V…

With the publication of the following footage, so to came with it suspicion that it was AI, or at least partially, specifically the second video.

The doubts aren’t unfounded — but after a thorough review, I believe the footage is authentic. Here's why:

A🧵

https://twitter.com/talhagin/status/1940871926089707761

As AI technology advances, so does our skepticism toward video footage. Visual oddities that were once dismissed as typical artifacts are now increasingly seen as undeniable signs of AI fakery.

So, what raised suspicion in this case? Four main things — but one frame in particular stood out.

A viral claim alleges that over 1,000 Israeli deaths—military and civilian - were exposed via a hack of the Israeli "Hess" website, due to Iranian strikes.

But is there any truth to it—and can we even know, given Israel’s military censorship?

A🧵 The official numbers, per Israeli/Palestinian media (TOI/Maariv/Palestine Chronicle/MOH), cite 29 deaths (28 civilians / 1 off-duty soldier) and 3,000+ injured, but the death toll could rise as there are still several in critical condition (As of June 24th).

Sources:
TOI (They are missing 1 name: Yelena Zadovsky)
timesofisrael.com/these-are-the-…
Maariv:
maariv.co.il/economy/israel…
Palestine Chronicle:
palestinechronicle.com/israeli-media-…
MOH:
gov.il/he/pages/24062…