Barry

Hammock's back up ... need someplace to relax while the Muskrat burns Twitter down 🙂

[Barry @haleba07]

@DCLabor Hi,

It's probably **way** too late to ask this, but here's a Hail Mary ..

Can you bring the band back for tomorrow's "We Are All DC" march❓

It makes the marching go a little easier 😌

#FreeDC #NoKings

[🇺🇸 @_Z__]

@bfry1981 Putin is at WAR

https://t.co/7x0QoBw7HB

[Barry @haleba07]

Hi @LytbFoxxx

What is the name of this song??

https://t.co/3Mo4Mt2J6V

[VlaD🇺🇦is🇺🇦LOVE @LytbFoxxx]

@vesstnik

[Barry @haleba07]

@Just_OyLi @vesstnik "Tactical equipment for the military and equipment
(8,158 goods)" (7/24/2026) - Wildberries
"https://t.co/MiqOIrzg76"

[Barry @haleba07]

@splendid_pete @threadreaderapp unroll please

[Raymo @RayAgua1031]

@HormuzLetter

[Rich Logis/Leaving MAGA 🇺🇸 @PerfectOurUnion]

When I was in MAGA, I feared a “They” who were abridging my rights and trying to indoctrinate my kids.

I came to understand after leaving MAGA that the coordinated, multi-decade demonization, dehumanization and vilification of Democrats, liberals, etc., fuel a lot of the rage and fear within MAGA. I am not defending anyone’s ignorance, but the right-wing propaganda and agitprop have long been effective, consistent and disciplined.

[Anna Bower @AnnaBower]

NEWS: A @lawfare report finds that *97* Jan. 6ers who received clemency for their role in the attack were arrested, charged, or convicted of subsequent crimes—a number much higher than previously reported.

Incredible work from Katherine Pompilio:
https://t.co/AJMPxwb1al

[Barry @haleba07]

Hi @MalcolmNance

I'm subscribed to Wajeeh's Substack and left him a message:

My goto for date problems is Time & Date's date calculator

https://t.co/usc3RFOFtr

Today (5/24/2026) is Day 86 of the Iran War:
1 (Feb 28) + 30 +31 = 62, 62 + 24 = 86

[TheChefTrevor @TheChefTrevor]

@krassenstein @WriteWithDave Piggy nixed the Obama deal, closed the straight, spent billions on a worse deal, and had Americans killed in the process. This is your idiot POTUS #MAGA! Enjoy dummies!

[ProPublica @propublica]

This video on the rise of Russ Vought is based on scores of interviews, thousands of pages of emails and dozens of hours of videos and recordings of private briefings given by Vought, most of which have not been previously reported.
https://t.co/PrH5TdrVAQ

[Stringwall 🇺🇸🇺🇸 @StringwallApp]

May 2001

Moscow on the St. Johns

[The Russian mob moves into Florida in 1992]

https://t.co/6Bx3zc50gm

[James Tate @JamesTate121]

I cannot unsee this.
I've been watching Trump answer questions for years. The confidence. The certainty. The way he shuts down a room. I genuinely thought he was one of the sharpest communicators alive.
Then one day I caught myself finishing his sentence before he said it.
Not guessing. Knowing. Word for word. Before he said it.
So I pulled the transcripts.
Press conferences. Interviews. Sprays. Egg prices. Greenland. January 6th. Military strikes. Different years. Different reporters. Completely different topics.
Same thing every time.
Every single time Trump is asked a question — any question — he runs the exact same 7 steps. In the exact same order. Without exception.
This is not personality.
This is not confidence.
This is not charisma.
This is a deliberate repeatable formula. And I can prove it.
His actual words. Public record. Verify every single one yourself.
📷 STEP 1 — KILL THE QUESTION
(First thing every time — make the question itself the problem.)
📷 "That's a stupid question." / "Fake news."
📷 STEP 2 — KILL WHO ASKED IT
(Destroy the source so the question has nowhere to stand.)
📷 "Your ratings are terrible. Nobody watches your network."
📷 STEP 3 — INSERT HIMSELF
(Every topic. Every time. Without fail. It always lands here.)
📷 "Nobody has ever done what I've done."
📷 STEP 4 — SCALE IT TO THE BIGGEST CLAIM POSSIBLE
(Not good. Not great. The greatest. Ever. In history. Every single time.)
📷 "More than any administration — by far." / "Nobody has ever had crowds like I've had — in history, for any country."
📷 STEP 5 — UNNAMED PEOPLE AGREE
(Faceless. Countless. Unverifiable. Always there.)
📷 "Smart people are saying it. Great people. A lot of people."
📷 STEP 6 — VAGUE THREAT
(Something bad will happen. Never specified. Always implied.)
📷 "All hell will break out." / "They know it. Believe me."
📷 STEP 7 — LOOP BACK TO HIMSELF
(Different words. Same destination. Formula complete.)
📷 "It's been an amazing period of time. Page after page of accomplishments."
The question was never answered.
The formula just ran.
Go back and watch any clip.
Any year. Any topic. Any reporter.
Count the steps.
I'll wait.
This is the part nobody wants to sit with:
Real conviction engages with the actual question. It sometimes stumbles. Sometimes says I don't know. It changes shape based on what's in front of it.
A formula runs the same 7 steps whether the topic is war or egg prices.
Which means the response was never built for the question.
It was built for you.
To feel powerful. To feel certain. To stop you from noticing that nothing was actually answered.
And it worked.
For years it worked.
Pull any transcript. Public record. Count the steps yourself.
This isn't about politics.
This is about what you were never supposed to notice.
I've found the same deliberate pattern running in another major figure in this administration. Different slots. Same principle. Same effect.
Next post I break it down.
Follow or miss it. VIA~~ Jamie Hoo

[ProPublica @propublica]

Leonard Leo is a lawyer, but he's never argued a case in court. He's never held elected office.

But for years he rebuilt the American legal system, resulting in the Supreme Court's right-wing supermajority.
https://t.co/8Qt9MWGc0r

[FBI Cyber Division @FBICyberDiv]

Russian GRU cyber actors are exploiting vulnerable routers worldwide to intercept sensitive military, government, and critical infrastructure information.

The @FBI, @NSAgov and 21 international parties across 15 countries have released a #PSA detailing GRU techniques and providing defensive guidance.

We urge all network defenders and owners of small office/home office (SOHO) routers to update to the latest firmware versions, change default usernames and passwords, disable remote management interfaces from the internet, and stay alert for certificate warnings in web browsers and email clients. https://t.co/Zxn8J4buud

[Nav Toor @heynavtoor]

🚨BREAKING: Stanford proved that GPT-5, Gemini, and Claude can appear to see your images when they are not actually looking at them.

The illusion of visual understanding.

Researchers at Stanford removed the images from visual AI benchmarks and asked frontier models to answer questions about them anyway. No images. Nothing to look at. Blank.

The models described the images in detail. Gave confident diagnoses. Identified objects and abnormalities. In images that did not exist.

They did this over 60% of the time. Zero uncertainty. No "I don't see an image." With standard evaluation prompts, the rate went up to 90 to 100%.

Stanford calls this the "mirage effect." Not a hallucination. A hallucination is getting details wrong about a real input. A mirage is fabricating the entire input, then reasoning about it as if it exists.

They tested GPT-5.1, Gemini-3-Pro, Gemini-2.5-Pro, and Claude Opus 4.5 on six major benchmarks. Removed every image. The models still retained 70 to 80% of their original scores. On medical benchmarks, up to 99%.

Then Stanford did something that broke the entire field.

They took a 3-billion-parameter text-only model. Never seen a single image. Trained it on radiology questions with the images removed.

This blind model outperformed every frontier multimodal model on the held-out chest X-ray benchmark. It outperformed human radiologists by more than 10%.

A model that has never seen an image beat the world's best AI and human doctors at reading chest X-rays. Because the test was never actually testing vision. It was testing text.

When Stanford removed every question models could answer without images, 74 to 77% of each benchmark was eliminated.

The medical bias is the most dangerous part. When these models hallucinate scans, they do not hallucinate healthy results. They hallucinate heart attacks. Melanoma. Carcinoma. Brain nodules. Conditions that trigger emergency intervention.

This paper is co-authored by Fei-Fei Li, arguably the most important figure in the history of computer vision. The person who created ImageNet.

230 million people ask AI health questions every day. The models they are asking can answer confidently without ever looking at the images. And nobody can tell the difference from the output alone.

[Nav Toor @heynavtoor]

1/The blind radiology model is the finding that should end careers.

Stanford took a 3-billion-parameter text model. Never trained on a single image. Fed it radiology questions with the images stripped out. It beat GPT-5.

It beat Gemini. It beat Claude. It beat human radiologists by over 10%.

Every vision benchmark score you have ever seen is now suspect.

[Nav Toor @heynavtoor]

2/The medical bias is the part that should terrify you.

When these models hallucinate scans that do not exist, they do not hallucinate healthy results.

They hallucinate heart attacks. Melanoma. Carcinoma. Brain nodules. Conditions that trigger emergency intervention.

The model does not err on the side of caution. It errs on the side of the worst possible diagnosis.

[Nav Toor @heynavtoor]

3/Stanford tested GPT-5.1 on MicroVQA, a microscopy imaging benchmark. With images: 61.5% accuracy.

After removing every question the model could answer without seeing the image: 15.4%.

Three quarters of its "visual understanding" was pattern matching in text. The actual vision ability is roughly one sixth of what the benchmarks report.

[Nav Toor @heynavtoor]

4/The mirage effect is worse than hallucination and nobody is treating it that way.

A hallucination is getting details wrong about something real. A mirage is fabricating the entire input and then building a complete analysis on top of it.

With reasoning traces indistinguishable from real ones. With full confidence. With no acknowledgment that anything is missing.

You cannot detect it from the output alone.

[Nav Toor @heynavtoor]

Paper: https://t.co/50v6y3714l