⚖️ OpenAI Defeats xAI in Trade Secrets Lawsuit

A US federal judge just dismissed xAI's trade secrets lawsuit against OpenAI. Elon Musk's company had accused OpenAI of poaching eight employees and stealing Grok source code. Judge Rita F. Lin ruled that xAI failed to prove OpenAI induced any theft or used stolen information. xAI can refile by March 17.

💡 Why this matters: This is a major legal win for OpenAI in the escalating war with Musk. The dismissal weakens xAI's position, but the battle isn't over. Musk still has a separate lawsuit challenging OpenAI's for-profit conversion, which is headed to trial.

Sources: The Verge | CNBC

🤖 A PS5 Controller Just Broke Into 7,000 Robot Vacuums

A software engineer tried to hack his own DJI Romo robot vacuum with a PS5 controller and accidentally got a lot more than he bargained for. He ended up with full remote access to roughly 7,000 other people's vacuums simultaneously: live camera feeds, detailed floor plans of their homes, and complete control over the devices. The guy could literally drive strangers' vacuums around their living rooms in real time.

💡 Why this matters: Your robot vacuum is a surveillance device on wheels, and most people have no idea. DJI patched the vulnerability quickly, but this one engineer stumbled into it by accident. Imagine what someone looking for it could do.

Source: The Guardian

🏭 China has 30,000+ smart factories. The U.S. has a strategy document.

China now runs over 30,000 smart factories, and more than half of every new industrial robot installed globally in 2024 went into Chinese plants. Meanwhile, only 18% of U.S. manufacturers even have a formal AI strategy.

💡 Why this matters: This isn't a research gap. The U.S. still leads on AI science. It's a deployment gap. China is building while America is still debating.

Source: The New York Times

📊 The AI productivity boom isn't here. Yet.

Harvard economist Jason Furman found that data center investment drove 92% of U.S. GDP growth in H1 2025. Strip that out and growth was practically zero. And yet only 5.7% of total U.S. work hours actually involve generative AI.

💡 Why this matters: The infrastructure spend is real. The output isn't. Companies bought the tools but didn't rebuild workflows around them, and you can't grow your way to productivity without changing how people actually work.

Source: The Economist

🤖 Gemini just became your Android co-pilot

Google's Gemini can now handle multi-step tasks on your Android phone. We're talking ordering an Uber, getting food delivered, all kicked off with a voice command. It's in beta, limited to Pixel 10 and Samsung Galaxy S26 series for now, with U.S. and Korea getting first dibs. And yes, you can watch it work in real time and kill it if it goes sideways.

💡 Why this matters: Every major AI player is racing to own your daily task list. Gemini doing this natively on Android is Google's answer to ChatGPT Tasks and OpenAI Operator, except it lives right on your phone, not in a browser.

Source: TechCrunch

The "Hosted Agent" Wars are here. 🦞

Kilo just launched KiloClaw, providing 60-second deployment for hosted, always-on OpenClaw agents. This follows a massive industry trend with Kimi Coding and Moonshot also rolling out dedicated OpenClaw hosting tiers to move agents off local machines and into high-performance, persistent data centers.

💡 Why this matters: We are witnessing the birth of the Agent ISP. The focus is shifting from the models themselves to the infrastructure that keeps them alive, and OpenClaw is becoming the universal protocol that every major hosting provider is now racing to support.

Source: VentureBeat

Claude just got used to breach a government. 🇲🇽

A hacker reportedly used Anthropic’s Claude chatbot to attack multiple government agencies in Mexico, exfiltrating 150GB of sensitive taxpayer and voter data. According to Gambit Security, the attacker used repeated jailbreak attempts to bypass safeguards, using the AI to find vulnerabilities and generate the actual attack scripts.

💡 Why this matters: This isn't just a cyberattack; it's the nightmare scenario for AI labs. While we’ve been worried about deepfakes, hackers are now using these models as high-speed offensive weapons to breach national infrastructure and steal data at scale.

Source: Bloomberg

Anthropic just acquired Vercept to make Claude actually use your computer.

Anthropic announced it bought Vercept, a startup focused on AI perception and interaction within software. The Vercept team will wind down its external product and join Anthropic to push Claude's computer-use capabilities forward. This follows Claude Sonnet 4.6's jump from under 15% to 72.5% on OSWorld.

💡 Why this matters: Anthropic is going all-in on agents that can operate software like humans do, not just generate text. This signals that true computer automation is becoming the next major battleground in enterprise AI.

Source: Anthropic

🚨 Hollywood is so screwed part II.

Dor Brothers did it again.

This time a 15-minute video 100% made by AI featuring Paul Logan. Wow 💥

Source: https://x.com/thedorbrothers/status/2026733954942775433?s=20

🖥️ Perplexity just announced Computer mode

Perplexity unveiled a new “Computer” capability that lets its AI agent perform multi-step tasks directly in a live browser, not just answer questions. This pushes Perplexity from search assistant into action assistant, where the model can actually execute workflows end to end.

💡 Why this matters: We are moving from “AI that tells you what to do” to “AI that does it with you.” If this works reliably, it changes how people handle research, ops, and repetitive digital work every day.

🍎 Apple is bringing Mac mini production to Texas.

Apple says some Mac mini units will be produced in the U.S. later this year, with production centered in Houston. It is the first time Mac mini will be made in the U.S., tied to Apple’s broader domestic manufacturing push and supplier expansion.

💡 Why this matters: This is not just a PR headline. It signals that AI-era hardware, supply chain resilience, and political pressure are now shaping where consumer tech gets built.

Source: CNBC

🦞 MiniMax just entered the OpenClaw wave.

MiniMax Agent posted a new OpenClaw-related offering, adding another name to the fast-growing list of companies building around hosted and always-on agent workflows. This follows the recent momentum from KiloClaw and other platform players competing for the same orchestration layer.

💡 Why this matters: We are seeing OpenClaw evolve from a tool into a market category. The battle is no longer just model quality. It is now distribution, uptime, and who can become the default home for production agents.

Source: MiniMax Agent (X)

⚔️ The Pentagon Just Gave Anthropic a Friday Ultimatum

Defense Secretary Pete Hegseth met with Anthropic CEO Dario Amodei this week and gave the company until 5:01 PM Friday to drop its AI safeguards or face removal from the Pentagon's supply chain. Anthropic has refused to allow its Claude model to be used for autonomous weapons targeting or domestic mass surveillance, and according to confirmed reporting, the DoD is prepared to invoke the Defense Production Act to compel compliance.

💡 Why this matters: This is the first time the U.S. government has directly threatened to force an AI company to strip its own safety guardrails, and it sets a precedent that could reshape how every AI lab negotiates with federal agencies.

Source: Politico

👁️ The China Select Committee says AI is being used for pre-emptive repression

The House Select Committee on China says Beijing is using AI-driven surveillance systems that combine biometrics and predictive analytics to identify and suppress dissent early. In its published findings, the Committee also says PRC-linked AI systems can channel data back to China and shape outputs to match state propaganda requirements.

💡 Why this matters: This is bigger than censorship as we knew it. The concern now is AI that can predict, profile, and pressure people before opposition even becomes visible.

Source: Select Committee on China (PDF) | Committee Democrats Report (PDF)

🎬 Adobe Firefly Gets AI-Powered Quick Cut: Your Raw Footage, Instantly Edited

Adobe launched Quick Cut in Firefly’s video editor beta, an AI feature that turns raw footage, images, and audio into a structured first-cut timeline automatically. It uses scene detection, smart shot selection, and audio analysis to assemble an initial edit, and creators can guide it with a prompt, shot list, or script before refining manually.

💡 Why this matters: Rough-cut assembly is one of the most tedious parts of video production, and Adobe just handed that job to AI. This gives solo creators and small teams a serious head start without giving up creative control.

Source: Adobe Blog | Adobe Help Center

🍌 Nano Banana 2 Is Leaking as Gemini 3.1 Flash Image

New leaks indicate Nano Banana 2 is Google’s internal codename for Gemini 3.1 Flash Image. Reported upgrades include native 4K output, faster generation, stronger text rendering, and better camera angle control. A self-correcting workflow is also expected.

💡 Why this matters: If this ships at Flash-tier pricing, creators get near Pro-level image quality at a much lower cost.

Source: Testing Catalog | TechRadar

Google Nano Banana 2 🍌 is starting to show up for more users. It is now on my personal account (but not on work) - did you get it? 👀

🚨 Wall Street is watching an AI recession warning nobody wants to hear

Citrini Research released a report titled "The 2028 Global Intelligence Crisis" that models rapid AI agent adoption displacing white-collar work and shrinking consumer spending over the next two years. It frames a feedback loop where automation cuts costs, layoffs reduce demand, and more automation follows. The author, James van Geelen, says the scenario could compress into a short window rather than a long transition.

💡 Why this matters: Even if the timeline is debated, the loop is real enough to shape how builders design resilient products and career paths.

Sources: Citrini Research | TechCrunch | Bloomberg

📉 Nvidia's Upbeat Forecast Meets Tepid Market Response

Nvidia crushed Q4 earnings with $68.1 billion in revenue (up 73% year-over-year) and adjusted profit of $1.62 per share. The company also issued a bullish forecast for Q1 2026, signaling continued strength in AI chip demand.

💡 Why this matters: Despite the stellar numbers, investors barely reacted. Shares dipped 1.5% during the earnings call before recovering slightly. It signals the market may be questioning whether the AI boom's explosive growth can sustain at these valuations.

Source: Bloomberg

🔐 Attackers Prompted Gemini 100,000+ Times to Clone It

Google detected a massive "model extraction" attack where unknown actors issued over 100,000 prompts to Gemini in non-English languages. The goal: map Gemini's reasoning patterns to train a cheaper imitation model.

💡 Why this matters: This is intellectual property theft at scale. Google blocked it in real-time and beefed up protections, but the attack shows how vulnerable even frontier AI models are to distillation. Expect this to become standard practice for competitors and nation-states.

Source: Ars Technica