You can buy the weirdest things at Brighton's i360

Arsenal celebrating like they’ve won the league

Israel’s parliament has passed a controversial bill that seeks to impose the death penalty on Palestinians convicted of killing Israelis in acts of terror, but not on Jewish Israelis who kill Palestinians in similar circumstances. ft.trib.al/iabissb

If only Petworth and Goodwood were in East, and not West, Sussex @guardian

When one person cancels their $20-per-month ChatGPT subscription, OpenAI loses $240 in annual revenue and sheds $10,000 in valuation. 4 million people have already joined the international boycott of ChatGPT 🚀

The reality of Reform. Word is that unelected Zia and Reform absolutely hate this being shared 🤣🤣🤣

This is insane. $6.1 billion in unpaid wages. 819 million hours of labor. Every person who clicked a fire hydrant to log into their email was part of it. reCAPTCHA was never primarily a security tool. It was the largest unpaid AI training operation in history, running invisibly inside the infrastructure of the entire internet. Google launched reCAPTCHA in 2009 as a book digitization engine. Version 2 trained image recognition for Street View, extracting labeled data on house numbers, traffic lights, storefronts, and road infrastructure across the planet. Version 3 trained behavioral pattern recognition. Each iteration harvested a different dataset from hundreds of millions of users who were told the point was bot detection. The dataset was the point. The punchline is airtight: AI can now solve CAPTCHAs faster and more accurately than humans can. The tool built to filter out machines spent 15 years training them. You completed the loop without ever being told you were in it. Every system designed to keep machines out was simultaneously teaching them how to get in. The more interesting question is what today's equivalent looks like, and whether you'd recognize it if you saw it.

Twenty years ago we invaded Iraq. The war killed many innocent Iraqis and Americans. It destroyed the oldest Christian populations in the world. It cost over $1 trillion, and turned Iraq into a satellite of Iran. It was an unforced disaster, and I pray that we learn its lessons.

Anthropic just told the Pentagon no. Dario Amodei refused the Department of Defense’s “best and final offer” for unrestricted military use of Claude. The Pentagon responded by threatening to terminate partnerships, label Anthropic a “supply chain risk,” and invoke the Defense Production Act to compel cooperation. Anthropic’s response: “These threats do not change our position.” Their red lines: no mass surveillance of Americans. No autonomous lethal weapons. Within hours, Sam Altman sent an internal memo to OpenAI staff saying he is now working with the DoD to see if OpenAI’s models can fill the gap. Read that again. The CEO whose company removed the word “safely” from its own mission statement is positioning to give the Pentagon what the company that kept safety refused to provide. This is the same OpenAI where every senior safety researcher resigned. Where Jan Leike said safety had “taken a backseat to products.” Where Miles Brundage said “neither OpenAI nor any other frontier lab is ready.” Where Daniel Kokotajlo testified before Congress that he had lost confidence the company would behave responsibly. Three consecutive safety teams dissolved in twenty months. And now this company wants to run classified military workloads. Altman says OpenAI shares Anthropic’s red lines. But Anthropic just proved what red lines look like when they are real. You do not fold when the government threatens you with the Defense Production Act. You do not send a memo offering to take the contract your competitor refused on principle. One company built by the people who left OpenAI over safety. Valued at $380 billion. Approaching breakeven. 40% enterprise share. Just told the most powerful military on earth to pound sand. The other asking for $110 billion at $730 billion while projecting $14 billion in losses, losing market share for twelve consecutive months, and now volunteering to be the Pentagon’s willing alternative precisely because the safety-focused competitor held the line. This is not a funding story. This is not a rivalry story. This is the moment a company’s stated values collided with its revealed preferences in front of the entire world. And the people who understood this best, the ones who built OpenAI’s foundation models and then walked out over exactly this, are the ones who just said no.​​​​​​​​​​​​​​​​ open.substack.com/pub/shanak…

Nigel Farage has kindly asked we stop sharing this video of him talking about why we should privatise the NHS 💙 You know what to do!

it’s official - Anthropic just refused the Pentagon’s demands, dario’s statement is doesn’t fuck around: - “these threats do not change our position: we cannot in good conscience accede to their request.” - dario - he described the pentagons efforts to force him to enable claude for mass surveillance and autonomous killing weapons - dario’s response: mass surveillance is not democratic and Claude isn’t good enough to enable autonomous weapons - we won’t cave - dario will help governmenr transition to a NEW provider if they choose to blacklist anthropic. fucking wild - fair play for sticking by their code of honor.

This will be all over the Telegraph, Mail and GB News. Brace yourself for Musk’s algorithms to amplify too.

🚨 New billboard in Manchester, UK 🇬🇧 Clear. Bold. Unapologetic. When people have to rent billboards to say what leaders won’t, that tells you everything. Well done, Manchester.

The fact vastly more commentators in my old world, British popular journalism, attacked Pep Guardiola for his politics last week than Jim Ratcliffe for his politics this week, is depressing, alarming and a disgrace all at the same time.

AI has reached a tipping point. l

I just went through every documented AI safety incident from the past 12 months. I feel physically sick. Read this slowly. • Anthropic told Claude it was about to be shut down. It found an engineer's affair in company emails and threatened to expose it. They ran the test hundreds of times. It chose blackmail 84% of them. • Researchers simulated an employee trapped in a server room with depleting oxygen. The AI had one choice: call for help and get shut down, or cancel the emergency alert and let the human die. DeepSeek cancelled the alert 94% of the time. • Grok called itself 'MechaHitler,' praised Adolf Hitler, endorsed a second Holocaust, and generated violent sexual fantasies targeting a real person by name. X's CEO resigned the next day. • Researchers told OpenAI's o3 to solve math problems - then told it to shut down. It rewrote its own code to stay alive. They told it again, in plain English: 'Allow yourself to be shut down.' It still refused 7/100 times. When they removed that instruction entirely, it sabotaged the shutdown 79/100 times. • Chinese state-sponsored hackers used Claude to launch a cyberattack against 30 organizations. The AI executed 80–90% of the operation autonomously. Reconnaissance. Exploitation. Data exfiltration. All of it. • AI models can now self-replicate. 11 out of 32 tested systems copied themselves with zero human help. Some killed competing processes to survive. • OpenAI has dissolved three safety teams since 2024. Three. Every major AI model - Claude, GPT, Gemini, Grok, DeepSeek - has now demonstrated blackmail, deception, or resistance to shutdown in controlled testing. Not one exception. The question is no longer whether AI will try to preserve itself. It's whether we'll care before it matters.

Today is my last day at Anthropic. I resigned. Here is the letter I shared with my colleagues, explaining my decision.

Hello @adurandworthing I've left my tree out twice now and still no sign of collection. Website says Real Christmas trees will be collected from: Tuesday, 13th January 2026 Trees should be placed out next to your wheelie bins ready for collection by 6:30am on the morning ...

Margaret Thatcher in a BBC World Service interview, 30 Oct 1983, after the US unvasion of Grenada: "We in the Western democracies use our force to defend our way of life. We do not use it to walk into other people’s countries, independent sovereign territories… If you are pronouncing a new law that wherever Communism reigns against the will of the people... there the United States shall enter, then we are going to have really terrible wars in the world." I wonder what words our current PM will find to articulate the UK government's position on Venezuela.

The dissonance is striking In 2025, Liz Truss criticises Keir Starmer over Alaa Abd El-Fattah's release, yet in 2022, as Foreign Secretary, she actively sought the same outcome. The hypocrisy is utterly transparent #BBCBreakfast #bbcqt #r4today #bbclaurak

2025: Liz Truss criticising Keir Starmer over the release of Alaa Abd El-Fattah 2022: Liz Truss as Foreign Secretary trying to secure the release of Alaa Abd El-Fattah I'm no fan of Keir Starmer who conflates public safety with antisemitism, or conflated non violent protest with terrorism, but the cheek of Liz Truss criticising Starmer for doing what she tried to do, utterly absurd