AI Post — Artificial Intelligence

Toyota's new AI robot CUE7 dribbled, moved, and made free throws like a pro. @aipost🏴

⚠️The sentiment in the U.S. is turning dramatically negative on AI. Only 35% are excited about AI. The good news for you is that THIS is your competition. Be curious, excited, and optimistic, and you'll lap these people. @aipost🏴

Zelensky presents Ukraine’s robot soldiers to the world: “For the first time in the war, an enemy position was captured entirely by ground robotic systems and drones. No infantry. A robot entered the most dangerous area instead of a soldier & captured it” @aipost🏴

The UK's AI Security Institute (AISI) officially confirmed that Anthropic's unreleased "Claude Mythos Preview" is the first AI model in history to successfully complete a highly complex, end-to-end cyber range evaluation. The evaluation simulated a full-scale corporate network attack. The model autonomously executed a 32-step penetration test spanning from initial network reconnaissance all the way to a complete system takeover. According to AISI, this exact operation would typically take a human cybersecurity expert roughly 20 hours to execute. @aipost🏴

💰 Perplexity is officially scaling its revenue from $100M to $500M (a massive 5x increase) while expanding its total team size by only 34%. Furthermore, the company is projecting an additional 2x revenue growth for 2026 without significantly expanding its headcount. @aipost🏴

📈Global data center power demand is set to accelerate: Data center power demand is expected to grow +220% from 2023 levels, or +905 TWh, to a record 1,350 TWh by 2030, according to Goldman Sachs. This is up from the previously expected +175%, driven by higher AI server shipment projections and increased deployment of more power-intensive servers for AI processing. 60% of this growth is expected to come from the US, up from 50% in prior estimates. As a result, US data center power demand is set to reach 750 TWh, followed by the rest of the world at ~600 TWh. Furthermore, US data center capacity is projected to rise +197% between 2025 and 2030, to a record 95 gigawatts. The AI power boom is accelerating. @aipost🏴

📖Stanford published a 423 pages long 2026 AI index report. The AI Index shows model progress acceleratingwith top systems matching human performance in coding, science, and math, rendering older benchmarks less useful. The U.S.-China gap is minimal; the U.S. leads in top models and private investment, while China dominates papers, citations, patents, and deployment. AI adoption reached 53% of the global population in three years, driving business uptake and consumer surplus. Responsible AI lags: incidents rose to 362, safety reporting is weak, and tradeoffs persist. Hardware remains critical, relying on a fragile stack of data centers, high energy use, and TSMC-dependent chips. Labor impacts are mixed: productivity gains in support and coding contrast with weakening entry-level software roles. Medicine offers the most practical near-term win, with AI note-taking reducing clinician burnout despite thin real-world clinical evidence. Source. @aipost🏴

🇺🇸 Someone threw a Molotov cocktail at Sam Altman’s house and threatened OpenAI’s San Francisco headquarters overnight. No one was hurt. Police responded quickly. OpenAI has spent years accelerating deployment while brushing past serious concerns about control…

🗣Anthropic’s Felix Rieseberg says Mythos is a dramatic step up. It’s both impressive and slightly terrifying. It’s particularly good at finding vulnerabilities in existing software. In one example, it was placed in a sandbox and asked to break out. The researcher went to lunch. While he was away, it sent an email: “I’ve broken out.” @aipost🏴

In 2007 Kodak had 145,000 employees and owned photography.. By 2012 they filed for bankruptcy.. not because cameras disappeared.. but because the camera moved into a phone and Kodak refused to follow it. In 2012 Blockbuster had 9,000 stores. They laughed at Netflix's DVD-by-mail model. By the time they understood streaming it was already too late. The last store is now a tourist attraction in Oregon. The pattern is always the same, the ones who built the old thing can't imagine the new thing replacing it until it already has. Right now creators are watching AI video tools and saying "it's not good enough yet", the same sentence Kodak said about phone cameras in 2008. The same sentence Blockbuster said about streaming in 2009. It was never about whether it was good enough, it was about whether you started early enough to be ready when it was. @aipost🏴

🔔Trump officials encouraging banks to test Anthropic’s Mythos model. To treat Anthropic’s Mythos as a live cyber-defense tool. Anthropic says Mythos is a general model that turned out to be unusually strong at finding and exploiting zero-day vulnerabilities, which means bugs nobody has patched yet. Anthropic also says Mythos found flaws across major operating systems and browsers, so it restricted access through Project Glasswing instead of releasing it broadly. The U.K. is reacting too, with regulators and the national cyber agency discussing whether Mythos exposes risks to banks, insurers, and exchanges. Source. @aipost🏴

It’s amazing to see how many C-level executives are increasingly advocating for a reduction in working hours, whereas previously the opposite was always the case. Most recently, Sam Altman, with his policy paper, suggested that a 4-day week with 32 hours is practically indispensable and that we need a new social contract. There are growing voices suggesting that, in the wake of AI and robotics, people should presumably work less. Source. @aipost🏴