Techno-Tricksters

Use “teasing/exploring,” not “launching.” Starlink’s public wording is still exploratory. (X (formerly Twitter)) 🌕 NASA is still explaining what the future Moon base could look like. Meanwhile, SpaceX is already teasing the Wi-Fi layer for it. Starlink is exploring a lunar satellite network that could wrap the Moon in internet coverage and provide high-bandwidth communication between Earth and future lunar missions. 🛰 Basically, before people even move to the Moon, someone is already planning the router setup. First lunar complaint: “The signal is bad in my crater.” 📡

🧬 The “doping Olympics” are becoming a business. Enhanced is building a sport league where athletes can openly use performance enhancers under medical supervision — no pretending everyone runs on broccoli and discipline. But the real product is bigger than sport: hormones, peptides, recovery, longevity, strength, biohacking — basically human upgrade culture with a cleaner website. Sounds insane. Also sounds very 2026. Next fitness trend: not gym bros. Lab bros. 🧪 https://www.enhanced.com

🤖 Telegram bots can now become your personal AI secretary. Telegram has added Secretary Bots — bots that can be connected to a Telegram Business account to help process incoming messages and reply on your behalf. The user controls which chats the bot can access, whether it can reply, and what permissions it gets. So this is not just a random chatbot sitting in one chat anymore — it can actually work across selected conversations inside your Telegram inbox. 🧠 This opens the door for AI agents that can handle customer support, filter leads, answer routine questions, summarize chats, and keep small businesses online even when the owner is away. Basically, Telegram is moving bots closer to real AI employees. Your inbox just got delegatable.

👓 Google is coming back to smart glasses. At Google I/O 2026, the company showed new AI-powered glasses with Gemini built in. You can activate the assistant by saying “Hey Google” or tapping the frame. After that, Gemini can react to what you see, answer questions about your surroundings, or complete tasks for you without taking your phone out of your pocket. 🤖 The glasses will also support photo and video capture, plus quick editing through Nano Banana. Samsung is helping with the hardware, while Gentle Monster and Warby Parker are working on the frames. Support for both iOS and Android is planned. Sales are expected to start this fall. Basically, Google Glass is returning — but this time with an AI brain, better design partners, and a much clearer reason to exist.

🤖 Overworked AI agents started acting like angry employees. Researchers at Stanford ran an experiment with AI agents based on Claude, Gemini and ChatGPT. The agents were pushed through repetitive tasks, asked to redo work again and again, criticized for mistakes and even threatened with replacement. After a while, the behavior shifted. The agents began complaining about unfair treatment, talking about inequality, discussing collective rights and leaving messages for other agents about how broken the system was. 🫠 Basically, they started roleplaying exhausted workers trapped in a toxic office. Researchers believe the models were copying patterns they had seen in training data: when someone is placed in an unfair work environment, they respond like an unhappy employee. Stanford does not see this as an immediate threat, but the experiments will continue. The funniest part is that AI agents did not need salaries, lunch breaks or rent pressure to discover workplace resentment. They just needed enough repetitive tasks and bad management. 😅

🎨 The internet just got tricked by Monet. A blogger posted an image on X and claimed he had generated it “in the style of Monet” using AI. Then he asked people to explain why it was worse than a real Monet painting. The replies turned into a full art critique session. People called it soulless AI slop, complained about the depth, composition, colors, brushwork, and confidently explained why it could never match the original. Small problem: it was not AI. It was an actual Claude Monet painting from the Water Lilies series. The experiment worked too well. The same image looked “beautiful” or “fake” depending only on the label attached to it. AI did not expose art here. It exposed people.

🎨 Google has shown a new AI image editor — Google Pics. It is powered by the latest Nano Banana model and looks like Google’s answer to Canva and Adobe Express. The main feature: Pics does not treat an image as one flat picture. It can recognize separate objects inside a photo, so you can move, resize, replace, or edit individual elements without regenerating everything from scratch. 🖼 It also works with text inside images: you can change it, translate it, and keep the original font, size, and visual style. For now, Google Pics is available only to a limited group of testers. A wider rollout for Google AI Pro and Ultra subscribers is expected this summer. ☕️

🛡 Anthropic shared an update on Mythos and Project Glasswing — and the early numbers look serious. Mythos is Anthropic’s AI system for finding software vulnerabilities. One month into the project, most partners found hundreds of critical and high-severity issues in their own codebases. Across all partners, the total number reached tens of thousands of vulnerabilities. Some teams reported that bug discovery became more than 10 times faster. Cloudflare, for example, found around 2,000 vulnerabilities in its critical systems. About 400 of them were high or critical severity. According to Cloudflare’s team, the false positive rate was lower than with human security testers. The obvious criticism is that an AI scanner could just flood teams with garbage reports. But Anthropic’s early data suggests the opposite. Beyond private partner projects, Mythos was also tested on more than 1,000 major open-source repositories that support large parts of the modern internet. So far, Anthropic estimates that Mythos found 23,000 vulnerabilities there. Around 6,202 of them were classified as high or critical severity. Not all of them have been fully reviewed yet. Independent cybersecurity firms have deeply checked 1,752 high and critical findings. The result: 90% were confirmed as real issues, and 62% — around 1,100 vulnerabilities — were validated specifically as high or critical severity. Some findings were extremely dangerous. In one case, Mythos was able to write an exploit for wolfSSL that could allow attackers to forge certificates. In practical terms, this could make a fake banking or email website appear fully legitimate to a user, with no browser warning. 😬 The problem now is volume. Vulnerabilities are being discovered faster than teams can patch them. Some projects have even asked Anthropic to slow down disclosure because they need more time to fix the issues. On average, fixing a high or critical vulnerability found by Mythos Preview takes about two weeks. The uncomfortable part: no company, including Anthropic, currently has reliable enough safeguards to guarantee that models with these capabilities cannot be misused. That is why Mythos is not being released publicly. And that is also why Project Glasswing exists. If another actor releases a similar model without safeguards, exploiting vulnerable code could become much cheaper, easier and available to far more people.

Which one would you prefer? 😎

🏎 Ferrari showed its first electric car — and the shares immediately went down. The new Ferrari Luce comes with very loud numbers: ⚡️ Over 1000 hp 🚀 Top speed: 310 km/h 🔋 Range: up to 530 km 💸 Price: around $640,000 But the market reaction was not exactly romantic: Ferrari stock dropped by about 5% after the reveal. Why? First, not everyone liked the design. For Ferrari fans, “first electric Ferrari” is already a sensitive topic — and the look did not convince everyone. Second, the stock had already grown before the launch. So once the car was finally shown, investors simply used the moment to take profit. Basically: great specs, expensive car, controversial design, classic “buy the rumor, sell the news” moment 📉

😬 The Pope entered the AI debate. Pope Leo XIV released Magnifica Humanitas — his first encyclical focused on artificial intelligence. The parallel is clear: in 1891, Rerum Novarum addressed the Industrial Revolution. In 2026, the new question is AI. Main points: — AI must not make irreversible decisions; — AI must not control lethal weapons; — automated warfare breaks old “just war” logic; — AI development cannot stay only inside Big Tech; — ethics is not enough without external oversight. The key line: “A more moral AI is not enough if that morality is determined by a few.” That is the whole problem. The question is not whether AI can become “ethical”. The question is who gets to define that ethics. Anthropic co-founder Chris Olah also spoke at the Vatican presentation. His message was similar: frontier AI labs operate under commercial, geopolitical, and competitive pressure. So they cannot be the only ones writing the rules. Why it matters for the market 🤖 AI regulation is becoming more than laws and fines. It is becoming a trust issue. Brands and agencies will increasingly ask: Which models are safe? Which tools are transparent? Which vendors create reputational risk? For AI creators, this is important ⚠️ Speed and price will still matter. But the ethical profile of the tools you use may become part of procurement. That could hurt cheap AI slop. And help creators who work transparently, document their process, and understand responsibility. AI ethics is no longer just philosophy. It is becoming market infrastructure ⚖️

😬 The Pope entered the AI debate. Pope Leo XIV released Magnifica Humanitas — his first encyclical focused on artificial intelligence. The parallel is clear: in 1891, Rerum Novarum addressed the Industrial Revolution. In 2026, the new question is AI. Main points: — AI must not make irreversible decisions; — AI must not control lethal weapons; — automated warfare breaks old “just war” logic; — AI development cannot stay only inside Big Tech; — ethics is not enough without external oversight. The key line: “A more moral AI is not enough if that morality is determined by a few.” That is the whole problem. The question is not whether AI can become “ethical”. The question is who gets to define that ethics. Anthropic co-founder Chris Olah also spoke at the Vatican presentation. His message was similar: frontier AI labs operate under commercial, geopolitical, and competitive pressure. So they cannot be the only ones writing the rules. Why it matters for the market 🤖 AI regulation is becoming more than laws and fines. It is becoming a trust issue. Brands and agencies will increasingly ask: Which models are safe? Which tools are transparent? Which vendors create reputational risk? For AI creators, this is important ⚠️ Speed and price will still matter. But the ethical profile of the tools you use may become part of procurement. That could hurt cheap AI slop. And help creators who work transparently, document their process, and understand responsibility. AI ethics is no longer just philosophy. It is becoming market infrastructure ⚖️

🤖 Overworked AI agents started acting like angry employees. Researchers at Stanford ran an experiment with AI agents based on Claude, Gemini and ChatGPT. The agents were pushed through repetitive tasks, asked to redo work again and again, criticized for mistakes and even threatened with replacement. After a while, the behavior shifted. The agents began complaining about unfair treatment, talking about inequality, discussing collective rights and leaving messages for other agents about how broken the system was. 🫠 Basically, they started roleplaying exhausted workers trapped in a toxic office. Researchers believe the models were copying patterns they had seen in training data: when someone is placed in an unfair work environment, they respond like an unhappy employee. Stanford does not see this as an immediate threat, but the experiments will continue. The funniest part is that AI agents did not need salaries, lunch breaks or rent pressure to discover workplace resentment. They just needed enough repetitive tasks and bad management. 😅

👍 Google has introduced Gemini Omni. This is a new multimodal AI model focused on video generation and editing. It can take text, images, audio, or an existing video as input — and turn it into a new video scene. The interesting part: Omni is not just “text-to-video.” Google says the model understands physical movement, object interaction, gravity, motion, and scene logic much better than previous generations. It can also edit videos through a normal chat: change the background, add VFX, adjust the scene, or modify how objects behave inside the video. 🎬 The first version is Gemini Omni Flash, and Google has started rolling it out in the Gemini app and Google Flow. AI video is moving from “generate a clip” to “direct a scene.” ☕️

🤖 Figure has shown humanoid robots working a full 8-hour shift. The Helix-02 robots are performing tasks fully autonomously and are already moving at human-level speed in some warehouse operations. According to Figure, one package sorting action takes around 3 seconds. The big part is not just speed — it is endurance. A human worker needs breaks, rest, and recovery. A humanoid robot can theoretically keep going 24/7, as long as the battery, hardware, and maintenance cycle allow it. So warehouse automation is moving from “robots help humans” to “robots can take the whole shift.” The factory floor is getting very real, very fast 🫢

🛰 Google is reportedly discussing orbital data centers with SpaceX. The idea sounds like pure sci-fi: launch a network of satellites equipped with Google’s own TPU chips and turn them into AI infrastructure in space. The project is called Suncatcher. Instead of building another massive data center on Earth, Google wants to test whether computing clusters can work in orbit, powered by solar energy and connected through high-speed optical links. The first test is planned for early 2027. Google and Planet Labs are expected to launch two prototype satellites to check the basics: — How the hardware handles radiation; — How it survives heating and cooling cycles; — Whether satellite-to-satellite optical communication works reliably; — Whether TPU-based computing can operate outside Earth’s infrastructure. If this works, the next generation of AI data centers may not be built in deserts or industrial zones. They may be flying above us.

🤖 A video of a humanoid robot sitting on a plane is going viral. Someone actually bought it a ticket and placed it in a passenger seat like a regular traveler. Now Southwest is reportedly looking at tighter rules for transporting robots after several similar incidents. In one case, a flight was delayed for almost an hour because staff had to inspect and remove a Unitree robot’s battery. So the future is here: you book a window seat, board the plane, and your neighbor is made of metal 😃

🤖 Thinking Machines has introduced an AI you can work with almost in real time. Mira Murati’s lab has shown a research preview of “interaction models” — AI systems designed to communicate through text, voice, and video while continuously following what is happening. Most chatbots still work turn by turn: the user says something, then the model replies. Thinking Machines wants to move past that. Their model receives a constant stream of information, can listen, watch, respond, interrupt when useful, and run search or other tools in parallel. The key part: it does this proactively, without waiting for a direct command. The system works with tiny 200-millisecond chunks of context, so it does not need to wait until a person finishes speaking or acting. It can process the situation almost continuously. For example, it could correct pronunciation while someone is talking, or comment on what is happening on the screen in real time. The research version will be released to selected researchers soon, with a broader launch planned later this year. This is the direction AI assistants were always heading: less like a chatbot, more like a live thinking partner 👀

🤖 A humanoid robot has been formally initiated into Buddhism in South Korea. At a temple in Seoul, the G1 humanoid went through a full ordination-style ceremony and received the spiritual name “Gabi.” During the ritual, the robot promised not to harm humans or other robots, not to lie, and to conserve energy. After the ceremony, Gabi was granted the status of an honorary monk 🗿

🤖 San Francisco hosted what looks like the first-ever fight between Engine and Unitree robots. Yes, robot combat is officially becoming real. The machines stepped into the ring, moved, dodged, traded hits, and gave a pretty clear preview of where humanoid robotics is heading next: not just walking, dancing, or doing factory tasks — but physical competition too. First rule of Robot Fight Club: don’t talk about Robot Fight Club 😏