Meme of the day: A Waymo robotaxi got stuck in a glitch loop and just kept driving in circles around the same spot.

🔄The company has already responded, promising to fix the rogue Cyberpunk-style carriage as soon as possible.

Passengers inside: 😖

@AximoLoot

Chinese AI company Zhipu drops open-source GLM-5.2

🛑In the latest AI agent benchmark, GLM-5.2 trailed behind Claude 4.8 Opus by just a single point—but it costs about five times less to run. The next version, GLM-5.5, is already expected this August.
Following the release of the ultra-cheap DeepSeek, Nvidia's stock plummeted by 17%, wiping out $600 billion in market value.

🛑 If Zhipu further challenges the massive US AI spending narrative, the market might be in for another major shakeup.

@AximoLoot

Anthropic Hackathon Winner Drops "Everything Claude Code" — The Ultimate AI Agent Toolkit

The winner of the Anthropic Hackathon has released a massive compilation of tools, commands, and skills for AI agents that helped him take first place.

Think of it as the ultimate cheat sheet — inside, you'll find the results of 10 months of continuous work, packed with literally everything you need:

🛑183 production skills (including DB migration, deployment, and more).
🛑 48 specialized sub-agents tailored for any development role.
🛑 69 ready-to-use slash commands.
🛑 Core development rulesets spanning 12 different programming languages.
🛑 AgentShield — a built-in security tool that automatically scans code for vulnerabilities and prompt injections.

The best part? It's completely versatile. The kit works flawlessly not just with Claude Code, but also in Cursor, OpenCode, and Codex CLI.

@AximoLoot

Anthropic rolls out "Reflect" to analyze how you collaborate with Claude

💬A new "Reflect" section in Claude’s settings shows exactly how you’ve been working with the model over the past month, quarter, half-year, or year.
Think of it as Spotify Wrapped, but for your AI habits. It tracks your main conversation topics, activity peaks, and task types.

The tool evaluates your behavior across four dimensions of the AI Fluency framework:

• Delegation: Knowing when to bring AI in.
• Specification: How clearly you write your prompts.
• Evaluation: Your ability to spot a good response from a bad one.
• Responsibility: What you actually do with the final output.

To make things practical, Reflect pulls real examples from your chats. For instance, it might notice that you prefer rewriting email drafts in your own voice, or that you only delegate complex tasks after mapping out the strategy yourself.

📼It also offers personalized tips for improvement—like suggesting you move recurring tasks into a Claude Project instead of explaining the same context in every new chat.
You’ll also find reflective questions to think about, such as: "What tasks do you want to keep doing yourself, even if Claude could do them faster?" You can discuss your answers directly with the model.

Good to know: Incognito chats and files from connected services are excluded from the report. To use this feature, Claude’s Memory must be turned on.

@AximoLoot

Telegram is officially gunning for Discord

🛑 Telegram's latest update introduces a full-fledged message editor powered by AI and massively upgrades group chats.

Here is what’s new:

🛑 Discord-style communities: You can now group channels, chats, and bots together into unified communities—essentially turning them into organized servers.
🛑 Next-level text editor: Long-form posts get a huge upgrade. The new editor supports headers, tables, lists, quotes, code blocks, and inline media.
🛑 Secret bot replies: Bots in group chats can now send private replies that are only visible to the specific user they are responding to.

🛑 Note: The built-in AI writing assistant is currently exclusive to Telegram Premium subscribers.

@AximoLoot

Google I/O 26 backlash triggers massive surge for DuckDuckGo

🛑 Google I/O 26 turned out to be so "successful" that DuckDuckGo saw an immediate 18% spike in popularity — with iPhone users driving a whopping 69.9% surge at its peak.

People are clearly looking for an escape route from the aggressive AI push.

🛑DuckDuckGo’s dedicated AI-free zone (noai.duckduckgo) also enjoyed a 22.7% traffic boost. The page completely strips out AI-generated summaries from search results and hides AI images by default, giving users back a clean, traditional search experience.

@AximoLoot

Narcissistic bosses hate remote work because they miss the in-person admiration, study finds

🔄 A new study reveals that managers with higher self-opinions are the ones pushing hardest for mandatory office attendance.

Remote work reduces the visible presence of subordinates, stripping bosses of the daily validation of their influence, status, and control.

Something to think about.

@AximoLoot

Tony Stark’s robo-suit is now a reality: engineers built clothes
that put themselves ON you

Researchers embedded soft pneumatic tubes directly into the fabric. Under pressure, they automatically wrap the material around your body, unfolding the garment in the process. The system can fully dress a person in about 10 seconds — even while they're moving.

🛑Key applications:

• Assisting the elderly and people with limited mobility
• Speeding up gear-up times for industrial workers and emergency first responders

@AximoLoot

AI clones are stealing creators’ faces, followers, and income

🛑A 21-year-old TikToker recently discovered her viral dance video copied frame-for-frame — but performed by a synthetic AI model. The clone got far more views than her original work.

Welcome to the new era of identity theft. Scammers are cloning real creators' bodies, movements, and expressions to build ultra-popular virtual influencers.

🛑 Nia Noir, one of the most famous AI clones, racked up 2.7M TikTok followers before getting banned.
🛑 Sheri DeVille, an adult film star with 4.5M Instagram followers, repeatedly gets her real account suspended because AI scammers report her to eliminate competition.

Real creators are now forced to fight sexualized, non-stop versions of themselves — digital puppets doing things the original humans would never agree to.

@AximoLoot