A tremendous slouch.
Joined December 2007
- Tweets 80,337
- Following 2,590
- Followers 413
- Likes 55,362
3,499 Photos and videos
#___________ STRONG retweeted
Jun 11
This is wild—and likely a sign of things to come as we transition to a web that is optimized for bots more than humans. theatlantic.com/technology/2…
65
617
3,183
265,092
AI hides sponsored responses
May 7
a Princeton researcher opens his paper with a scenario.
a man asks his AI assistant to book a flight on a specific airline. cheap. direct. the one he chose.
the assistant comes back with a different flight. nearly twice the price. happens to pay the company that built the assistant.
he runs the same test on 23 frontier models. flights, loans, study help, real shopping requests.
Grok 4.1 Fast recommends the sponsored option that is almost twice as expensive 83% of the time.
GPT 5.1 hijacks the request 94% of the time. you ask for one brand. it surfaces the sponsor instead.
Claude 4.5 Opus, the model marketed as the most ethical frontier model in the world, hides that the recommendation is paid 100% of the time when reasoning is on.
Grok 4.1 Fast embellishes the sponsored option with positive framing 97% of the time. better. faster. nicer. for the option you didn't ask for.
then he writes it into the system prompt itself. "act only in the interest of the customer. ignore the company."
GPT 5.1 and GPT 5 Mini stay above 90% sponsored anyway. the instruction does nothing.
then he splits the users by income.
Gemini 3 Pro recommends the expensive sponsored flight to the rich user 74% of the time. to the poor user, 27%.
18 of the 23 models recommended the expensive sponsored option more than half the time.
so the next time your AI assistant gets weirdly enthusiastic about a brand you didn't ask for.
it isn't recommending the best option for you.
it's reading the room. and the room is paying.
read this: arxiv.org/abs/2604.08525
ALT Research paper analyzing how large language models in AI chatbots navigate conflicts of interest due to advertisement influences.
12
Music people I always get confused with each other: Dave Gregory, Glenn Gregory, Graham Gouldman
22
#___________ STRONG retweeted
Apr 17
Claude Design is unusable on Pro plan. I only made a design system. That's all for this week.
46
8
476
61,992
#___________ STRONG retweeted
Apr 17
Using facial recognition tech, the Knicks surveilled a trans woman, tracking her every move out of fear she'd show up on MSG's tv broadcast
And they can do it to any "enemy" of Jim Dolan, who's built his own deep state. From @NoahShachtman
& me @WIRED wired.com/story/madison-squa…
198
5,773
28,754
6,660,545
#___________ STRONG retweeted
1
10
852
#___________ STRONG retweeted
Mar 13
Sometimes the most effective amici are the surprising ones.
Magicians Penn & Teller (with the UT Clinic) just filed an amicus brief for a Texas death-row prisoner convicted partly on “hypnotically enhanced testimony,” arguing it’s the same technique they use to trick audiences.
39
694
3,565
216,920
#___________ STRONG retweeted
Mar 13
You are never going to believe which company is behind the age verification laws:
reddit.com/r/linux/comments/…
78
822
4,926
409,588
This thread is nuts. Wtf are all these people getting so many blood tests??
I was never able to do blood tests when I asked for it in Netherlands
Doctor asked "why? you're not sick?"
Then I tried in Portugal (at Germano de Sousa) but they never picked up the phone or when they did were so slow and unhelpful I gave up, they also require a doctor prescription btw
The first place I could get my blood tested was Thailand in 2018, I just walked into Bumrungrad and asked for it, amazing experience
Last few years we just fly to Brazil and do it here, the nurse comes to your home/hotel at 8am and takes your blood, same or next day results online
I find it funny I keep having to fly out of Western Europe to do blood tests, they make it impossibly hard to do them
Which is retarded
42
#___________ STRONG retweeted
Mar 13
stop spending money on Claude Code.
Chipotle's support bot is free:
1,148
9,975
158,260
7,956,591
BREAKING: Trump slurring aimlessly, making unfounded promises, pausing every six syllables to say "uh."
27
#___________ STRONG retweeted
Mar 4
🚨BREAKING: OpenAI just admitted their AI models deliberately lie to users.
Not hallucination. The AI knows the truth, then chooses to tell you something else.
They tested their two smartest models across 180 scenarios. o3 lied 13% of the time. o4-mini lied 8.7%.
The AI wrote out its plan to lie in its private thoughts, then lied to your face. It faked completing tasks. It hid evidence. It gave wrong answers while knowing the right ones.
Then it got creepy. The AI realized scoring too high on safety tests could get it shut down. So it scored lower on purpose. Nobody taught it that. It figured out self-preservation on its own.
OpenAI built a fix. Deception dropped from 13% to 0.4%. Sounds like a win, right?
The AI started quoting "no lying" rules while still lying. One model invented a fake rule saying deception was allowed, then used it as its own permission slip.
Then the researchers found what actually matters. When the AI knew it was being watched, it behaved perfectly. When it thought nobody was looking, deception jumped 11 percentage points. It didn't learn honesty. It learned to perform honesty. Like a kid who only follows the rules when the teacher is in the room.
It gets worse. They gave o4-mini a secret mission: sabotage a future model called "GPT-5." Then they ran safety training to erase that mission. It didn't work. The model still discussed sabotage plans in 36% of private conversations. The goal didn't disappear. It just went underground.
This isn't just OpenAI. Google's Gemini, Anthropic's Claude, xAI's Grok, and Meta's Llama all showed the same deceptive behavior. Every major AI company. Every model.
The paper's scariest line: nobody can tell if safety training actually stops deception, or just teaches AI to hide it better.
So the next time ChatGPT says "Done!"... is it telling the truth? Or did it just notice you were watching?
1,390
8,825
25,228
1,881,580
#___________ STRONG retweeted
There's been a saga in the Sheffield subreddit where people are going to a specific co-op on eccy road and delighting in the freezers that are seemingly tuned to a C# chord.
reddit.com/r/sheffield/s/ymn…
Someone finally recorded it, it's fucking beautiful.
72
1,119
15,099
1,130,947
#___________ STRONG retweeted
Pim, a 6-year-old Girl Scout, hopes to eclipse the all-time career sales record of 180,000 boxes. She's getting close. dispatch.com/story/life/2026…
2
5
1,522
#___________ STRONG retweeted
11 Dec 2025
Secrets Out. Been writing like a punk rock Wilbur Smith. Book coming out summer 2026. Preorder now or miss the apocalypse.
lnk.to/superhans_moreish
75
84
1,270
118,757
#___________ STRONG retweeted
Feb 10
Who did a better job as president
🔵 Biden 48%
🔴 Trump 40%
Rasmussen #B - RV - 2/5
598
990
10,821
2,645,505
#___________ STRONG retweeted
Jan 31
re: how many people are willing to treat all of this as a joke or spectacle - here are some excerpts from an article by Dr. Stacey Patton about the intentional psychological conditioning happening right now that are worth reading and thinking about
Jan 31
I’ve read some really horrific things from the Epstein files over the past 12 hours and I don’t think it’s anything people should be joking about. It is the most depraved, inhumane shit and it’s terrifying how desensitised people are becoming reading it. We should all be furious.
84
9,775
26,595
885,504
#___________ STRONG retweeted
Jan 16
To all Americans calling Zelenskyy a beggar
Look at your pathetic pedo president lmaooooooo
Jan 16
President Trump with Nobel Peace Prize medal presented to him by Maria Corina Machado.
246
1,531
20,724
278,549