What am I missing? Looking for more on-the-record examples of @sama lying: named source, sworn testimony, or paper trail. (Not looking for vibes).
There seems to be a pattern at OpenAI. cc @sama 2026-09-04 - OpenAI found its agents' German wiki message board on June 21, didn't disclose it until outside researchers published 2026-07-10 - Apple sues OpenAI for trade secret theft 2026-05 - Murati testifies Sam falsely told her legal had cleared skipping safety review for GPT-4 Turbo 2026-05 - McCauley testifies Sam told the board three ChatGPT versions had passed the safety board when one had 2024-05-28 - Toner: Sam didn't tell the board he owned the OpenAI Startup Fund while claiming no financial interest 2024-05-28 – Toner: employees were afraid to go against Sam because they'd seen him retaliate for past criticism 2024-05-22 - Sam says he didn't know about the vested-equity clawback; Vox publishes documents he signed 2024-05-20 - Johansson: OpenAI shipped a voice imitating hers after she declined twice 2023-12-08 - NYT: Sam told directors other directors wanted Toner removed; they hadn't said so 2023-11-17 - OpenAI's board fires Sam: "not consistently candid"
7
4
1,586
chair retweeted
Sometimes I read on flights, and sometimes…
33
7
185
12,141
I think we just let Anthropic do one more generation of models to Claude 6.5 family before pausing AI progress. Then we can call it a day, good game, well played, Timmy’s mom brought snacks.
Phew, Anthropic, you win this week, well played 🤌 Congrats to the whole team, and special thanks to @bcherny, I had left Claude Code for a while there, but I'm back, and damn is it better than ever. For anyone who hasn't tried it in a while, I dunno, you might want to check it out, esp. now that you have the trifecta of Opus 5.5, Sonnet 5.5, and Haiku 5.5. With these, plus Cloud Computers, it's kinda insane.
24
Among other things that weren't fully priced in by people who thought industry Super PACs could buy Congress -- AI timelines are shorter than electoral ones, so post-November, the next election will be well after superintelligence (even with some slowdown).
2
5
50
2,462
chair retweeted
haiku 5.5 is clocking in at 122 tps on openrouter btw nearly 2x faster than luna
anthropic may have just put the final nail in the coffin for luna haiku 5.5 looks ridiculously good crazy to see a claude model getting so cheap competition in this space is amazing
14
3
261
4,272
"As for Chamath, forcing him to humiliate himself on stage is neither clever nor novel; this is called the All-In podcast, and occurs weekly."

ALT F Murray Abraham Book GIF

Scott gemmed out
3
37
908
52,192
Cautious optimists/accelerationists unite!
Turns out I am a Radical Centrist 🔥
20
chair retweeted
It's so strange to have access to Mythos Anthropics marketing really won me over. I feel both respect and fear toward this model like I would to a powerful human
We’re expanding our Cyber Verification Program to give security professionals broader access to our most capable models. Through this program, verified security professionals can access Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5 with safeguards designed for defensive work. We’re also opening up new tiers to allow for authorized offensive work, like penetration testing and red-teaming. anthropic.com/news/cyber-ver…
18
14
859
94,414
chair retweeted
Okay this is fun. Audited 10 open source projects I contribute to (30k to 60k stars). 5 came back clean. The other 5 are getting private security reports from me today: an SSRF guard that fails open, a DNS rebinding hole, an allowlist bypass, an auth middleware gap, and a prefix-match bug. I heart Mythos.
What just happened?! My prior career as an agentic security researcher has paid off. Hi Mythos!
8
1
31
3,337
chair retweeted
Auto mode is free. We do not charge for it as a matter of principle.
5
4
56
4,732
chair retweeted
it's cute how the robots really are just absolute nerds. the life of a mind is the life of the mind.
1
1
45
1,816
chair retweeted
Your Mac is slowing you down. Moving to Linux has exponentially improved performance for my agents, in particular on the file system side.
192
106
2,680
1,180,663
Trump administration refuses international election monitors for first time in over 2 decades. The decision not to invite observers from the OSCE puts the U.S. alongside Russia and Belarus, while the Justice Department prepares to deploy its own poll monitors. ms.now/news/trump-midterms-o…
356
4,673
11,648
1,851,027
Between this and his “midterm surprise,” the time to prepare for him to try to steal an election is now.
Trump shouts out Russ Vought and then says, ominously, "we have some big surprises coming up in a few weeks, right Russ? He's gonna be one of the important people in the history of our country."
2,129
4,400
19,477
465,776
chair retweeted
Big, big, big, big props for quasiriemann and no Siegel zeroes (with many other beauties in there), I’m kicking myself for talking all over about it being within reach but not actually having pushed. There are some sad stories related to their users getting scooped / conflicts of interest (kakeya maximal comes to mind), and obviously we had some of the problems (without commenting on precise rumours). We should put that aside for today though. It’s obviously the most significant moment in mathematical history.
46
153
2,643
201,518
mistaking the David and the Goliath this badly is a sign of horrible epistemics at least, if not worse
If you’re willing to push doomer narratives about AI you can make hundreds of thousands if not millions of dollars. And keep in mind this is just what *one* EA org is doling out. It’s an endless money fountain and there’s no equivalent to it on the other side
1
1
26
1,899
chair retweeted
Just saw Artificial. Phenomenal. Savvy of them to frame it as a Sama biopic, but it’s really about Ilya. AI heads will enjoy the zillion little cameos of lesser-known characters, from Sam Teller to Jakub Pachocki. These people did their RESEARCH. It’s a horror movie, a tragedy, and when necessary, a comedy. I wish it were coming out sooner.
4
17
113
4,792
Replying to @firesidealpha
The problem w @profgalloway at this point is he is addicted to performative clickbait for attention. In depth Analysis is typically not part of his current schtick.
1
8
238
chair retweeted
Claude's day off 🏝️. Animated by Opus 5.5.
36
70
1,316
48,806
chair retweeted
Mom, can we have Steven Pinker AI Doom debate? No we have Steven Pinker AI Doom debate at home. Steven Pinker AI Doom debate at home: youtu.be/-tIq6kbrF-4
19
12
361
109,053
Replying to @firesidealpha
This sounds like pretty much every company. Remember, this guy is only 27 and only worked at OpenAI and Anthropic. He sounds more like an anthropologist observing a distant, isolated people's who have never encountered other cultures "People are very willing to sacrifice themselves for the good of the mission. They’ll take on work that’s less interesting, less big, bold and exciting, and less flashy. They’ll do it at a hit to their own interests for the good of the company."
3
1
50
4,347