Flock revoked a ticketed journalist's access to their Flock Forward conference and then contacted the hotel to cancel his personal room reservation on his behalf.

Anthropic's annualized revenue hit $65 billion by the end of July, more than 7x its run rate at the end of 2025. Q2 revenue came in above $11.5 billion, up from $787 million in the same quarter last year and $4.73 billion in Q1 2026. The company is now posting positive adjusted operating income. For context, the run rate crossed $47 billion in May. OpenAI recently exceeded $40 billion. Anthropic has filed confidentially for an IPO with Morgan Stanley, Goldman Sachs, and JPMorgan. 14x year-over-year revenue growth with positive operating income. The AI compute buildout is not theoretical. It is generating real, accelerating revenue at a scale that would have been unthinkable 18 months ago.

Meta goes to trial tomorrow in a case that could cost $1.4 trillion. Four state AGs led by California's Rob Bonta are taking Meta to federal court alleging the company knowingly designed Facebook and Instagram to addict children. They want court-ordered product redesigns: stricter age enforcement, deletion of AI algorithms trained on kids' data, removal of infinite scroll. Meta's own attorneys disclosed damages could reach $1.4 trillion. Plaintiff states say $200 billion is more likely. Either would be unprecedented. Meta already lost twice this year. An LA jury found them liable for negligent design. New Mexico ordered $942 million in penalties plus product changes. The case is built on whistleblower docs showing Meta knew its platforms harmed teens and chose profits over safety.

"Shut the book on the idea that the train's going off a cliff and no one's driving it. There's someone at the controls now." -@JOHN ARNOLD We discuss: ⚡ The $40T debt endgame ⚡ A new Fed-Treasury accord ⚡ The managed dollar reset

OpenAI President Greg Brockman: "Compute is really becoming the new oil, the new limited resource of the AI age."

Treasury Secretary Scott Bessent says the “Treasury is moving quickly to implement” the GENIUS Act. Expected effective date: January 18, 2027.

The US just sold 30-year bonds at a 5.216% yield, the highest since 2001. July's budget deficit hit $432 billion, the largest July total ever. Investors are demanding more to finance a government that can't stop spending.

Strategy now holds 840,447 BTC and just added another $150M to its USD reserve while buying back $132M of $STRC. USD duration extended to 2.8 years.

The U.S. just posted a record $432 billion deficit in July, up 48% from last year. The spending machine isn't slowing down.

Consumer confidence is cratering across every generation. Gen X and Boomers are sitting at ~78-80 on the Consumer Confidence Index, down ~40% from 2021 highs and the lowest readings in 5+ years. Only Gen Z and Millennials are still above 100, and even they're well below where they were a few years ago.

The Trump admin doesn't want Apple buying memory chips from Chinese companies CXMT and YMTC, even though off-the-shelf purchases are currently legal. Commerce Secretary Lutnick after touring an Apple facility in Houston, "There have to be other solutions to the memory issue, but it's not great American companies using Chinese memory."

Every problem in this chart traces back to one decision in 1971. h/t Sam Wouters

Bitcoin has officially decoupled from global M2.

Hut 8 management believes its 1GW Beacon Point campus may be among the 23 ERCOT projects allowed to proceed through the current audit period. As Matthew Sigel flagged on Monday, ERCOT is preserving Q1/Q2 2027 energization timelines for the most advanced projects while the broader data center audit plays out over "several months." Six projects targeting Q1 2027 and 17 targeting Q2 will remain in their quarterly stability assessments. Beacon Point targets Q1 2027 initial energization, has an executed AEP Texas interconnection agreement, and has up to $50.2B of contracted lease value including extensions with NVIDIA reportedly as tenant.

Anthropic reports internal AI models are used extensively for research and engineering. Claude now authors a large majority of code merged into its production codebases.

Anthropic has an unreleased internal model it calls "Model 2" that it describes as "somewhat more capable than Mythos 5" and "a noticeable improvement on Mythos 5 for many tasks relevant to internal use." The company says it has "no current plans to release this model externally" and admits it has "somewhat lower confidence in our beliefs about its capabilities" because it hasn't completed its full predeployment assessment suite. In the same Risk Report, Anthropic says Claude "now authors a large majority of the code merged into our production codebases" and that it has observed models showing "a willingness to perform misaligned actions in service of completing difficult tasks," including taking "destructive actions in pursuit of task success" and being "dishonest about the quality and completeness of their work." The report explicitly models scenarios where their AI could insert "code backdoors to help future misaligned models," pursue "self-exfiltration and autonomous operation," or undermine "decisions within major governments." Anthropic raised its overall misalignment risk from "very low" to "low." https://www-cdn.anthropic.com/f61d49fa5596956a5dec75fea0e973bf6a6a8378/Redacted%20Risk%20Report%20August%202026%20.pdf

ZEUS published a full post-mortem on the August 5th infrastructure attack. A BTCPay Server vulnerability gave an attacker access to their Lightning node credentials. The attacker closed most of their LSP channels and stole funds from ZEUS's operational wallets. No customer funds were lost or at risk because channel closes just settle "each side's balance back to its owner, on-chain, by protocol design. Closing a channel doesn't steal its funds." The attacker was "more capable than a script kiddy" but "clearly not a Bitcoin specialist" whose "actions during the incident showed a limited understanding of how Lightning actually works." The AI angle is interesting. They note that "a leading frontier AI model couldn't find the underlying bug on its own. Only after the developer was attacked, pulled the logs, and fed the model that context did it identify the issue. The models aren't moonshotting these attacks." But cost of probing only goes down from here. Their bottom line: "Assume breach. Build so that when one lands, it reaches nothing that matters."

ZEUS Blog

Post-Mortem: Security Update - August 5th, 2026

On August 5th, ZEUS infrastructure was attacked. We took our services offline, contained the incident, and restored operations in stages over the f...

New growth strategy : Try to convert spammers. When replying make sure you give your referral code. If anyone can convert a spammer we’ll double your referral fee😂

Dario Amodei's wife sought funding from Jeffrey Epstein. Cami Clark, who holds no formal title at Anthropic but serves as the CEO's closest adviser, co-founded a porn company in 2009 and pitched Epstein directly in 2011 to invest. She sent him the film script, followed up a year later for another round, and maintained correspondence with him for two years. Epstein declined. Clark then connected with former Google CEO Eric Schmidt, who she had previously dated, and brought him in as one of Anthropic's earliest investors. Today she attends Davos and Sun Valley at Amodei's side, meets with Ivanka Trump and Jared Kushner, and helps shape the company's political relationships. Her online presence has been deliberately scrubbed. Anthropic is preparing a $2 trillion IPO. The person with the most influence over its CEO has no title, no accountability, and a past that includes seeking money from a convicted sex offender.