🦀 The “Trash Panda” of AI: Meet Moltbot (formerly Clawdbot, now OpenClaw), the open-source AI assistant that went viral for doing your chores locally. It’s got 100,000+ GitHub stars and a personality described as “trash panda energy,” but security experts are horrified that people are giving it their Amazon logins.

🇪🇺 Europe’s Unicorn Stampede: While the US obsesses over LLMs, Europe just minted five new unicorns in a single month. Highlights include Harmattan AI (French defense tech) and Preply (holding it down from Kyiv).

🛡️ The Default Password Crisis: Russian hackers breached Poland’s power grid not using quantum zero-days, but default usernames and passwords. In 2026. The wiper malware they deployed was described as “digital arson.”

🤖 Anthropic’s “Cowork” Update: Claude is getting plug-ins. Anthropic’s new “Cowork” feature lets enterprises dictate exactly how the AI handles workflows, pushing us further from “chatbot” to “colleague.”

🚗 Waymo’s $16B Raise: The robotaxi wars are heating up with a massive funding round valuing Waymo at $110B. Uber is also doubling down, betting big on AV startup Waabi.

📜 Google’s Scaling Law Reality Check: Google dropped a bombshell paper: More agents does not equal better AI. In fact, multi-agent systems can wreck performance on sequential tasks (like planning) by up to 70%.

The News: Anthropic introduced “agentic plug-ins” for Cowork, allowing Claude to pull from specific tools and datasets to execute workflows. Meanwhile, Google’s Auto Browse agent attempted to book symphony tickets and ended up booking seats in separate rows because it followed instructions too literally.

The Takeaway: We are in the awkward toddler phase of AI agents. They can book the ticket, but they lack the “common sense” to know you want to sit next to your date. The potential is massive for enterprise automation (Anthropic’s play), but the consumer UX is still hilarious—and sometimes frustrating.

The News: A report surfaced detailing how Russian government hackers breached Polish energy facilities. The entry point? Default credentials on wind and solar farms. The attack deployed wiper malware designed to physically destroy systems.

The Takeaway: This is a stark reminder that as we digitize critical infrastructure (and fill it with AI-enabled sensors), the basics of cybersecurity (MFA, changing passwords) are being ignored. You can have the smartest grid in Europe, but if the password is “admin123,” it’s just a waiting room for a blackout.

Google ATLAS: New scaling laws for multilingual models show that training on diverse languages (like those in Europe) actually helps performance, challenging the “English-first” dogma.

Internal Monologue: Research shows AI that “talks to itself” (mumbling with short-term memory) learns faster and smarter, hinting at more human-like adaptability.

SpaceX’s Starshot: Elon Musk isn’t just launching satellites; he’s seeking approval for 1 million solar-powered data centers in orbit. They call it a step toward a “Kardashev II-level civilization.” We call it extreme cloud computing.

Physical Intelligence: Stripe veteran Lachy Groom is raising buzz (and cash) for “robot brains,” aiming to bridge the gap between digital AI and physical robotics.

The Toy Breach: An AI chat toy company, Bondu, left 50,000 logs of children’s conversations exposed to anyone with a Gmail account. It raises terrifying questions about privacy in the era of “always-listening” toys.