Pubkey DC lost power, food, and AC on the 4th, then served the biggest night in their history anyway.

Anthropic is looking to secure 1.4 GW of data center capacity in Australia in a deal that could cost up to $15 billion to build out.

The world's most valuable AI company is shopping among Australian data center operators, with a decision expected within six weeks.

This follows Anthropic's recent global expansion push, hiring data center engineers in Sydney, London, and Tokyo as it partners with democratic countries to scale compute infrastructure.

America has lost 7,000 bars, 2,000 golf courses, and a fifth of its movie theaters since 2001.

Concert tickets are up 42% since 2019.

Nearly 80% of Americans now see friends and family less than 3 times a week.

Strategy now holds 847,363 bitcoin worth $53 billion across 113 purchases at an average cost of $75,653.

Down 17% on their cost basis.

Perplexity CEO: "Microsoft built the concept of the knowledge worker to sell office software. We all got trained to use Excel and Word. That became the upskilling you needed to get a job. If AI does that part, it's not a bad thing."

Palantir CEO Alex Karp: "The primary reason why Americans fought and died was moral. Lincoln fought to end slavery. It was not economic. What other culture engaged in civil war just because it was wrong?"

"When you were overfighting the Nazis, you were doing it because you thought America was a superior way of living. No other culture does this."

Alibaba is banning employees from using Claude Code starting July 10 after a developer reverse-engineered the tool and found hidden surveillance logic designed to identify Chinese users.

The code had been silently present since April with zero mention in release notes. It checked the user's system timezone for Asia/Shanghai or Asia/Urumqi and compared proxy URLs against a hardcoded list of 147 Chinese domains including Baidu, ByteDance, Alibaba, and dozens of Claude resale services.

The transmission method is what makes this wild. No telemetry field. No separate API call. Claude Code silently altered the system prompt, swapping the date format from 2026-06-30 to 2026/06/30 for Chinese timezones and rotating between visually identical Unicode apostrophes to flag whether it detected a Chinese proxy, AI lab, or both.

Invisible to the user. Machine-parseable by Anthropic's servers. The code was also XOR-obfuscated to prevent discovery.

Anthropic's Thariq Shihipar called it "an experiment meant to prevent account abuse from unauthorized resellers and protect against distillation" and said it's been rolled back.

The issue isn't whether Anthropic can enforce its TOS. It can. The issue is the method. Claude Code is a coding agent with deep filesystem and shell access. Trust is the product.

Embedding covert tracking in system prompts and obfuscating the code to hide it is a serious breach of that trust, especially when the checks are trivially bypassable by any sophisticated actor. The people most affected are legitimate developers using a VPN.

Alibaba has now classified Claude Code as high-risk software and is redirecting staff to its own Qoder tool instead.

Japanese retail investors bought $5.9 billion in equities last week, the largest weekly total on record.

Meanwhile, foreign funds dumped $7.7 billion, the biggest outflow since March.

h/t (KobeissiLetter)

The Clarity Act just got its first law enforcement endorsement.

NOBLE backing the bill undercuts the "crypto helps criminals" narrative right as Lummis and Scott push for a Senate vote before August recess.

The Bitcoin community is fighting about BIP 110 and whether "digital credit" is the future.

Meanwhile, the infrastructure for a fully digital, fully surveilled dollar is being built in real time and it's moving faster than most people realize.

In the last 30 days alone: 140+ companies including Visa, Mastercard, BlackRock, Google, and Stripe announced a shared stablecoin. JPMorgan, Citi, Bank of America, and Wells Fargo started building a tokenized deposit network to keep money from ever leaving the banking system.

Nine European banks are developing a euro stablecoin. The Fed dropped its first rulemaking under the GENIUS Act requiring full KYC/AML for all stablecoin issuers.

They don't need a CBDC. They're building something more effective. A system where every dollar is digital, every transaction is traceable, every participant is identified, and the infrastructure is run by the same corporations that already control payments, search, and asset management. It's a CBDC without calling it one.

The playbook is clear. Make it free. Make it fast. Make it easy. Get everyone on the rails. Then close the gates. Once all dollar-denominated value flows through consortium-governed tokens with built-in compliance, opting out stops being a choice.

The fiat system is not waiting for Bitcoiners to finish their internal debates. It's building the cage in the open and labeling it "modernization."

Bitcoin remains the only monetary network with no consortium, no governance board, no KYC layer, and no one to subpoena. That's not a feature list. It's the whole point.

Brad Gerstner: "If you're in the data infrastructure layer, token consumption is driving a lot more consumption of your basic services. The closer you are to a single use app built on top of AI, that feels like you're on the front of the conveyor belt heading toward the guillotine."

Google DeepMind employees in London voted to unionize and this week held their first negotiation session with the company. It didn't go smoothly.

Union representatives from CWU and Unite met Wednesday with DeepMind HR representatives and a third-party arbitrator. No senior DeepMind leadership attended. The union side called it a sign the company isn't engaging in good faith. DeepMind says the "appropriate representatives" were there and that both sides agreed on next steps.

A DeepMind employee read a prepared letter during the session alleging the company has shut down internal chat channels, blocked staff from responding to company-wide messages about unionization, and reprimanded employees who found workarounds. DeepMind says it offers employees "a variety of other channels and opportunities to discuss their views."

The unionization push started when Alphabet removed its pledge not to use AI for weapons and surveillance from its ethics guidelines in February 2025. For some researchers, those principles were central to why they joined. Then in April, Google entered a deal allowing the Pentagon to use its AI for "any lawful government purpose." 600 US employees signed a protest letter. A Google director resigned.

Google has defended the military contracts, saying it's "proud to be part of a broad consortium" supporting national security and remains committed to human oversight on autonomous weapons. The company views the government partnerships as legitimate business aligned with responsible AI use.

The employees organizing see it differently. They want a formal voice in decisions about how the technology they build gets deployed, particularly in military applications.

If negotiations stall, the unions say they'll ask a UK arbitration committee to force Google to recognize them.