Welcome back to another entry of my weekly AI Diaries. Every time I sit down to write these updates, I feel like I’m trying to catch water with my bare hands—the pace of innovation in artificial intelligence is just that relentless.
We used to look at AI through incredibly rose-tinted glasses. We talked endlessly about how it would cure diseases, revolutionize clean energy, and act as a universally positive force for humanity. We even saw massive tech conglomerates sign manifestos swearing they would never weaponize these models. Well, after looking at the news crossing my desk this week, it is safe to say those promises have officially been forgotten.
Let me walk you through what happened over the last seven days, because we just crossed several major thresholds that are going to fundamentally change how we interact with technology.
The “Dark Manifesto” and the Militarization of AI
The biggest shift I noticed this week wasn’t a new piece of software; it was a shift in ideology. The boundaries of how AI can be used are being aggressively redrawn, and practically speaking, everything is now on the table.
The clearest indicator of this pivot came from Palantir. They just published a 22-point document titled the “Technological Republic” manifesto. Reading through it gave me chills—social media is already calling it the “dark manifesto,” and for good reason. It essentially normalizes the use of AI as a weapon. Palantir, whose influence in the US defense sector is rapidly expanding, used this document to publicly argue that utilizing AI for lethal military purposes is not only acceptable but necessary. It’s an open declaration of what we all kind of suspected was happening behind closed doors.
And they aren’t acting alone. Two massive developments dropped right alongside Palantir’s manifesto:
- Anthropic’s Reversal: Anthropic previously stated their most advanced model, Mythos, could be highly dangerous in the wrong hands. Now? It turns out they are exploring opening federal agencies’ access to it.
- Google and the Pentagon: Reports broke that Google is currently in talks with the Pentagon to create a framework that would allow Gemini to operate in highly classified, secure environments.
The AI arms race is no longer a sci-fi concept; it is our current reality.
Google Chrome Gets a Gemini Upgrade: Welcome to the “Skills” Era
Moving away from the heavy stuff, Google dropped a genuinely fantastic update for those of us trying to speed up our daily workflows. They are deeply integrating Gemini into the Chrome browser with a new feature they call “Skills.”
Here is how it works: Instead of typing out the same complex prompts over and over again, you can now save them as a “Skill.” Let’s say you have a specific way you like AI to summarize long articles or format data tables. You save that prompt, and the next time you need it, you just type a “/” in the Gemini interface or click a plus button, and your custom Skill runs instantly on your active tab.
What I really love about this is the ecosystem integration. If you are signed into your Google account on desktop, your saved Skills sync seamlessly across your devices. It can even pull data from multiple open tabs simultaneously. It’s a massive time-saver.
Microsoft’s Panic Button: “Copilot Code Red”
It seems like Microsoft is feeling the heat. Despite pushing Copilot into basically every product they own, they aren’t seeing the user adoption or performance results they wanted. To prevent falling behind in the AI race, they’ve initiated a massive internal pivot officially dubbed “Copilot code red.”
The goal is to drastically overhaul the user experience, but the strategy is what fascinates me. Just days after this “code red” leaked, we found out Microsoft is testing ways to turn Copilot into a 24/7 autonomous digital assistant.
By integrating AI agents similar to the OpenClaw architecture into the Copilot ecosystem, Microsoft wants this tool to stop waiting for your prompts. They want it running in the background, independently executing tasks and managing your digital life around the clock. If they can pull this off without it becoming an intrusive mess, it could change everything about how we use Windows.
Robots Are Finally Getting “Smart”: Gemini Robotics-ER 1.6
If you read my articles regularly, you know I have a soft spot for robotics. This week, Google DeepMind blew my mind with their new Gemini Robotics-ER 1.6 model.
Until now, most robots just blindly followed pre-programmed scripts. DeepMind’s new model gives machines “embodied reasoning.” This means the robot doesn’t just see its environment; it actually understands the context. It can analyze visual data, plan a multi-step task on the fly, and self-evaluate whether it completed the job correctly.
Boston Dynamics wasted absolutely no time putting this into action. They equipped their famous robot dog, Spot, with the ER 1.6 model, and the demo video is wild. Spot is shown looking at a handwritten to-do list and just… doing the chores. I watched a robotic dog organize scattered shoes, pick up boxes, and actively sort laundry into a basket. In one scene, it even grabbed a leash to take a real dog for a walk based entirely on natural language commands. We are watching the transition from mindless machines to reasoning physical entities in real-time.
The Posthumous Return of Val Kilmer
Finally, we have to talk about the ethical gray area Hollywood just sprinted into. The legendary actor Val Kilmer, who sadly passed away last year after his long battle with throat cancer, is “starring” in a new film called As Deep as the Grave.
Before his death, Kilmer had agreed to play the role of an American priest in the film, but his health prevented him from ever shooting a single scene. Now, with the explicit permission of his daughter, the studio has used AI to completely recreate Kilmer’s likeness and voice to play out the entire role.
Even with the family’s blessing, this has caused a massive uproar in the entertainment industry. While the technology is undeniably impressive, it leaves us asking deep questions about legacy, art, and the commercialization of digital ghosts.
Rapid Fire: New AI Tools & Industry Bites
A lot of smaller, but equally important, updates flew under the radar this week. Here is my quick breakdown of the rest of the news:
New Tool Drops:
- Claude Opus 4.7: Anthropic released what they are calling the most powerful model currently available to the general public.
- Claude Design: Powered by Opus 4.7, this new suite combines prototyping, slide deck creation, and visual generation. It is specifically built for people with zero design background.
- Happy Oyster & Lyra 2: Both Alibaba (Happy Oyster) and Nvidia (Lyra 2) launched massive tools that let users generate fully interactive, navigable 3D worlds using AI.
- Gemini 3.1 TTS: Google dropped an update to their Text-to-Speech model, and frankly, it’s one of the most natural-sounding vocal generation models I’ve heard to date.
Industry Shorts:
- China’s Master Plan: China is rolling out a massive national initiative to integrate AI directly into their education system, from elementary schools all the way to lifelong adult learning. They are aggressively building a digital-first workforce.
- Meta & Broadcom: The two giants signed a long-term deal (stretching to 2029) to co-develop custom AI silicon chips.
- Opera Browser Connector: Opera One and GX browsers just got an update that allows ChatGPT and Claude to natively read open tabs and interact directly with the browser interface.
- Solving the Unsolvable: A Chinese AI model independently solved a complex math problem proposed by American mathematician Dan Anderson in 2014—completely without human intervention.
- Nvidia Ising: Nvidia revealed the first AI model capable of breaking quantum barriers. “Ising” accelerates calibration and error correction in quantum computers, boosting performance by 2.5x and accuracy by 3x.
- Subconscious Bias: A fascinating new study published in Nature proved that when AIs generate training data for other models, they pass down hidden biases and prejudices on a “subconscious” level.
- Native Mac Gemini: Google officially released a standalone, native Gemini app for macOS.
- Vercel Hack: The cloud deployment platform Vercel confirmed a recent cyberattack. The scary part? The breach originated entirely from the hijacking of a third-party AI tool.
Over to you: I want to hear your take on the Val Kilmer situation. If a family gives their blessing, do you think it is acceptable for studios to resurrect deceased actors using AI for brand new movies, or should a performer’s filmography end when they pass away? Drop your thoughts in the comments, because I’m genuinely torn on this one!
You Might Also Like;
