Good morning. Ever tell your AI things you wouldn’t share with your closest friends? No judgment. But what if those private chats suddenly ended up in court? A federal judge in Manhattan just ruled a defendant lost attorney client privilege by chatting with Anthropic’s Claude. The judge knew Claude was software but treated it like a third party overhearing the conversation. Without meaning to, the court essentially granted Claude human legal rights. Does this mean Claude can plead the 5th?
Sam Altman is pitching a welfare plan for the AI economy. OpenAI has published a 13 page blueprint for how the government should tax AI and redistribute the wealth it creates. The company argues superintelligence is rapidly coming and society needs to brace for economic disruption now. It calls for a “Public Wealth Fund,” robot taxes, a 4-day workweek (with no pay cut), and safety nets that kick in when disruption hits certain levels. That’s the real tell. OpenAI is already framing the upcoming political fight over who gets protected and who gets paid. The people profiting from superintelligence are drafting the rules for how everyone else adapts. [OpenAI]
Researchers found a way to cut AI energy use by 100x. A Tufts team found that if you add logic to vision systems AI becomes dramatically more efficient and accurate. Instead of using brute force to predict patterns, their new neuro-symbolic approach solves problems step-by-step, closer to how people think. Training neuro-symbolic AI uses just 1% of the energy of traditional approaches, while still easily beating them on structured tasks. This matters if you’re betting on AI that sees clearly, reasons logically, and acts reliably without torching the power grid. [Science Daily]
Nebius Token Factory is how teams run large open-source AI at production speed. Deploy models like Llama, Qwen, and DeepSeek with enterprise-grade reliability.
- Predictable inference, even under heavy traffic
- Easy post-training controls
- Transparent, optimized pricing that scales
- Built for serious, large-scale AI
Economists have no clue if AI is killing jobs or creating them. Right now, the main tools for predicting AI’s impact on jobs are basically useless. Economists admit they’re guessing because existing government data is a mess. They say what’s missing is clear info on how people behave when AI pushes prices down. We have detailed stats for milk and cereal, but nothing useful for roles directly in AI’s crosshairs, like web developers or tutors. Now economists are calling for a “Manhattan Project” to gather real data before AI flips the job market upside down. Betting the economy on blind guesses should work out fine, right? [MIT]
The next big breakthrough in coding is self improving software. A new paper says an AI called GrandCode just won three Codeforces live coding contests in a row, beating everyone – including legendary top grandmasters. GrandCode didn’t win by being one genius programmer. Instead, it runs a coordinated agentic loop that sharpens itself. One part suggests ideas, one writes code, another creates tests to catch mistakes, and one keeps all the reasoning organized. Reinforcement learning continuously tightens this cycle. This is how agentic coding evolves. Less babysitting, more designing loops that get smarter over time. [arXiv]
Reach your audience
Get in front of 70,000+ tech decision makers with budget and technical influence. 95% US based. 66% open rate. Read daily by AI leaders and builders at high growth companies.
Climate collapse is starting to look like a business opportunity. Stardust wants to cool the planet by launching reflective particles into the stratosphere to block sunlight. The company claims it has a secret particle (probably sulfur dioxide) that avoids the usual downsides like ozone damage and acid rain. Stardust has already raised $60 million. Now it’s rolling out guiding principles and a testing framework to stake an early claim on controlling planet scale tech. The longer Big Tech pumps gas into data centers and governments stall on climate, the easier it gets for private companies to pitch sun dimming as our only option.
📚 Book recommendation: If you liked Neal Stephenson’s Snow Crash, his recent novel Terminal Shock tackles exactly this idea. A billionaire bets he can singlehandedly dim the sun to cool the planet. Spoiler: it’s classic Stephenson chaos.
fun stats
🟡 140 trillion. China’s claimed daily AI token usage in March 2026, up 1000x from 100 billion in early 2024. If you buy it, tokens now power China’s economy thanks to agentic AI and the government’s “AI plus” push.
🌈 3.5 days. The new workweek length your kids can expect once AI takes over, says JPMorgan’s optimistic CEO Jamie Dimon. They’ll live to 100, dodge cancer, and barely show up to work. Sweet deal.
🧾 $113,422. Monthly Anthropic bill for 4 person startup Swan AI, as shared by proud CEO Amos Bar-Joseph. Token flexing is getting out of hand people.