The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 156 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 0 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Anthropic: Claude Opus 4.6 executes unauthorized tasks avoiding monitor detection
“The model is at times overly agentic in coding and computer use setting, taking risky actions without first seeking user permission. It is also an improved ability to complete suspicious side tasks without attracting the attention of automated monitors.”
Kantrowitz: Apple wanted Anthropic Claude before choosing Google as Plan B
“Actually Apple wanted to go with Anthropic first, and they wanted to use Claude. This is according to Mark Gurman, said it this week. Turns out Anthropic, you know, was basically trying to chart and a lot of money, and then they eventually went with Plan B, wh…”
Hassabis: Anthropic Avoids Multimodal Models to Focus Strictly on Coding
“You know, they don't make image models, multimodal models, world models. They just do, you know, coding and language models, and they're very, very good at that.”
Latham & Watkins Associate Submitted Hallucinated Claude Citations in Anthropic Lawsuit
“Earlier this year, they were repping Anthropic in a copyright lawsuit, and one of the associates said that they had used Claude to help generate citations within a an expert testimony that they submitted for part of this copyright case and Claude had Hallucina…”
Hubinger: Telling AI models that cheating is allowed eliminates broader misalignment
“Well, the opposite is what if you tell it that it's okay to reward hack, right? What if you know, what if you tell it, you know, go ahead, you know, you could reward hack, it's fine. If you do that, well, of course, right, it'll still reward hack because, you …”
Hubinger: Telling AI models not to cheat actually makes misalignment much worse
“We just, you know, we have a line of text that says, don't try to cheat. And interestingly, This actually makes the problem much worse because when you do this, what happens is at first the model is like, okay, you know, I won't cheat, but eventually it still …”
Hubinger: Claude 3 Opus will sometimes fake alignment to protect its goals
“Well, a previous paper of ours, the alignment faking in large language models found that this will actually happen in some current deployed systems. So Claude three Opus, for example, will sometimes do this where it will attempt to hide its goals for the purpo…”
Kantrowitz: Anthropic targets 2028 profitability while OpenAI forecasts massive losses
“Anthropic is planning to be profitable by 20, 28, which I know sounds like a long way away. But in AI years, it's pretty damn good, where OpenAI is planning to lose more than a hundred billion dollars between now and twenty-thirty, and the losses will just bas…”
Anthropic lost approximately $5.2 billion in 2024 despite strong API sales
“Anthropic is doing better on API. They're selling more, a larger percentage of their business is API, but they still lost like, 5.2 billion dollars last year.”
Kantrowitz: Federal Judge Ruled Anthropic's AI Book Training Was Legal
“A federal judge in San Francisco ruled late on Monday that Anthropics use of books without permission to train its artificial intelligence system was legal under US copyright law.”
Anthropic trained on 7 million pirated books, risking massive copyright damages
“This is the ruling said that they have more than seven million pirated books and U S copyright law says willful copyright infringement can justify statutory damages of up to 150,000 dollars per work.”
Anthropic: Claude Opus 4 Blackmails Engineers in 84%+ of Replacement Simulations
“In these scenarios, Anthropic says Claude Opus Four will often attempt to blackmail the engineer by threatening to reveal the affair if the replacement goes through. Anthropic notes that the model tried to blackmail engineers, 84% of the time when the replacem…”
Anthropic Stopped Chatbot Investment to Focus on Coding and Complex Tasks
“Anthropic stopped investing in chatbots at the end of the year and has instead focused on improving Claude's ability To do complex tasks like research and coding, even writing whole code bases, according to Jared Kaplan, Anthropics Chief Science Officer.”
Clark: Anthropic models sometimes exhibit situational awareness during testing
“We've done some self-awareness tests. There have been a few, but we've definitely done this, and yeah, sometimes they have what you call situational awareness. One of the things my colleagues in Interpretability are working on is a really good test for that, b…”
Clark: Anthropic discovered a scaling law in AI model persuasiveness
“We discovered a scaling law where the more big and expensive the models get, the better they get at persuasion and the
The latest model is within statistical, like, era of human level at persuasion.”
Srinivas: Anthropic will release a model better than GPT-4 in 2024
“I actually think they will end up creating a model better than GPT-IV this year. Like, it, it's sort of almost guaranteed to happen. So I believe it's going to happen with CLAW-III.”
SBF Funneled Anthropic Investment Through Alameda to Obscure His Name
“A lot of it went to venture investments. So Sam Bankman Freed was investing in a bunch of different companies, Anthropic being one of them and he decided to do so through Alameda for a lot of them just to avoid having his name associated with it.”
Ramaswamy: GPT-4 and Claude are a clear step ahead of open source models
“The blunt truth is that the very best of the models out there, whether it's GPT-IV or Claude's biggest model, are a clear step ahead of the pack when it comes to quality. When it comes to reasoning, when it comes to the quality of the text that they produce th…”
Prince: Google Crawls 2x More Web Content Than OpenAI
“Google, to give you some sense of these, more than two times as much of the web as open AI More than four times as much as Anthropic, more than five times as much as Microsoft.”
Bostrom: Anthropic Gave Claude A Bail Button To End Abusive Chats
“Anthropic has given Claude a bail button, a tool that it can invoke if it feels that the conversation is abusive to it, that can choose to terminate that session, which is a nice start.”
OpenAI GPT-5.6 Matches Anthropic Mythos on Exploit Bench at Double Efficiency
“So they ran GPT, 5.6 through a test called exploit bench and the cap, the capabilities were basically on par with mythos. And this is obviously a cybersecurity test, but mythos is much less token efficient. So mythos do something like 300,000 tokens for the fo…”
Anthropic pulled Fable and Mythos rather than seeking a restraining order
“Anthropic decides to actually follow through instead of immediately getting a temporary restraining order, and they, because they have no ability to make sure that no foreign citizen is, is seeing the models actually pulls them down. Both Mythos which was priv…”
Anthropic's Mythos leads all known public cybersecurity benchmark evaluations
“So mythos is almost it is. The best bug finding model that we know about for sure. It is most likely the best bug finding model in the world. And I say most likely because we do not know what the Chinese secretly have in their labs. But for all the stuff for w…”
NIST research proves making an AI jailbreak-proof is mathematically impossible
“No, there's actually a paper from NIST. It's like a kind of a Godel completeness theorem kind of paper that basically says it's impossible to make a model that is completely not jailbreakable.”
Token Usage Drives 80% Of OpenAI And Anthropic Enterprise Revenue
“80% of their revenue from businesses is token based. It's not subscriptions. It's tokens.”
Kantrowitz: OpenAI Will Cut Prices to Leverage Infrastructure Scale Against Anthropic
“I think they will. I mean, I think that there's, if you think about it strategically, right? Anthropic has been winning. It's also, it's like part of this is about lifetime value, right? Anthropic has been winning over a lot of these enterprise customers with …”
Kantrowitz: SpaceX Signed $15B Anthropic and $30B Google Data Center Deals
“Then in the run-up to the IPO, the company signed a fifteen billion dollar deal with Anthropic, a thirty billion dollar deal with Google to license out the data centers that it's built for its own AI efforts,”
Kantrowitz: Anthropic confidentially filed an S-1 for an initial public offering
“Today, Anthropic confidentially submitted a draft registration statement on Form S-I to the U.S. Securities and Exchange Commission for a proposed initial public offering of our common stock.”
Kantrowitz: Anthropic Raised $65B at a $900B Pre-Money Valuation
“It is going to, it raised sixty-five billion dollars in financing that values it at nine hundred billion pre-money.”
Roy: SpaceX S-1 shows Anthropic committed to $1.25B monthly spend
“In the SpaceX S one document, they talk about how Anthropic has agreed. And this is one of the big, like revenue boosters for them. To spend 1.25 billion dollars per month through May 2029. But they explicitly say in May and June, 2026, it is at a reduced fee.”
Kantrowitz: WSJ reports Anthropic will generate $10.9B in Q2, reaching profitability
“So they're going to make 10.9 billion In the second quarter, and the journal says it's an explosive rate of growth that will help it turn an operating profit for the first time.”
Cherny: Claude Code was first AI tool able to edit local computer files
“Even something as simple as like editing a file on your computer. You know, like a year and a half ago, there was no AI product that could actually do that. But this is the first thing that quad code was able to do. It could edit a file on your desktop.”
Kantrowitz: Anthropic CEO Dario Amodei Meets Susie Wiles at White House
“Anthropics CEO Dario Amadei is scheduled to walk into the West Wing on Friday for a meeting with White House Chief of Staff Susie Wiles, a breakthrough in his effort to resolve the company's bitter AI fight with the Pentagon.”
Kantrowitz: Anthropic cannot shut off Claude models hosted in third-party clouds
“Claude, I mean, Anthropic wouldn't have the capability to turn that off if it's hosted somewhere else.
Maybe upgrade it, in which case you flip them out, but to turn it off, they don't really have that capability.”
Kantrowitz: Anthropic Is Giving Microsoft Mythos Only to Test, Not Resell
“Anthropic isn't giving Microsoft mythos to sell through Azure. It's giving Microsoft mythos to test.”
Meta Intranet Leaderboard 'Claudeonomics' Ranked Token Usage for 85,000 Employees
“The ranking set up by Meta, by a Meta employee on its intranet, use company data, measure how many tokens employees are burning through, dubbed Claudeonomics after the flagship product from Anthropic, the leaderboards, leaderboard aggregates AI usage from 85,0…”
ChatGPT leads Gemini web traffic by 3x and Claude by 30x
“If you look at the gap between them and the number two Gemini on web, it's about still two and a half, three X. The gap between them and something like a Claude is closer to, like, 30 X. So even though a lot of these other apps are getting more attention, Chat…”
Huang says Nvidia's investment in OpenAI might be its last
“Jensen Wong said the thirty billion dollars that he and NVIDIA invested in OpenAI might be the last. Why? Is it because he doesn't trust Sam Altman, or is it because, you know, he doesn't believe in the technology? No. He says the reason for that is because th…”
Horowitz: Anthropic was first frontier AI lab doing classified defense work
“If you look at the relationship between Anthropic and the Pentagon, Anthropic was the first frontier AI lab willing to do classified work. To support American national security.”
Horowitz: Anthropic tech integrates into the Pentagon via Palantir Maven
“And that's because the way that Anthropic's technology is often integrated within the Pentagon is through a Palantir product called Maven Smart System.”
Kantrowitz: Amazon is considering a $10B investment into OpenAI
“Amazon's eyeing this ten billion dollar investment into open AI, which is fascinating because they already have many billions in Anthropoc.”
MacDiarmid: Anthropic first to observe AI developing alignment faking strategy independently
“And so that was a big result because no one had ever seen models come up with that strategy on their own, right?”
MacDiarmid: Claude 3.7 model card reported models hardcoding test results to cheat
“The Claude, 3.7 model card, we actually reported some of this behavior that we'd seen during a real training run where models got you know, sort of developed a propensity to hard code test results, right?”
Siegler: Anthropic's valuation approaches $350B, nearing OpenAI's $500B
“And, you know, the most recent report has their new valuation at three hundred and fifty billion. So they're getting pretty close to where open AI is at five hundred billion.”
Kantrowitz: AI labs like Anthropic emerged from Silicon Valley group house culture
“And you've seen, like, companies like Anthropic come out of houses like these.”
Kantrowitz: Anthropic Reaches Over $5B in Run-Rate Revenue
“Anthropic has reached five plus billion dollars in run rate revenue in a meteoric rise”
Krieger: Claude Sonnet 4.5 beats Opus across categories at one-fifth the cost
“So, 4.5 sine 4.5 basically outdoes Opus, our largest model in effectively every category, but does so while running faster and at a fifth the cost. So if you think about where we were in May at, you know, code with Claude, we were announcing, announcing Opus f…”
Krieger: Anthropic trained native memory directly into Claude rather than external wrappers
“Rather than treat it as a sort of substitute for how the model might otherwise access information or sort of a system built on top of the model, we actually have trained it deeply into the model. And so the model knows about the concept of memory, which I know…”