🤣🤣🤣

~ husk

@aipost 🏴

⚠️ AI as a financial system threat: Inside the “Mythos” emergency

The U.S. government convened an emergency meeting with Wall Street leaders regarding Mythos, an advanced Anthropic AI model, marking a shift to AI-driven systemic threats.

What is Mythos?
• Next-gen AI capable of autonomous system hacking.
• Exploits vulnerabilities across major OS and browsers.
• Surpasses existing models in cyber capabilities.

Why it hasn’t been released:
• Anthropic withheld public access due to extreme misuse risk.
• Limited rollout to ~40 companies under “Project Glasswing”.
• Goal: patch vulnerabilities before exploitation.

Government & Wall Street response:
• Emergency meeting led by Scott Bessent and Jerome Powell.
• Bank CEOs warned current defenses are inadequate.
• AI classified as a systemic financial and national security risk.

Escalating threat landscape:
• Chinese state-backed hackers previously used a weaker Claude model.
• Over 30 organizations reportedly breached.
• Mythos represents a significant leap in offensive capability.

The core uncertainty:
• Controlled access may be insufficient.
• Mythos could evolve faster than defensive systems adapt.

AI is now infrastructure-level risk, forcing governments to treat models like critical weapons systems.

@aipost 🏴

📢 OpenAI developed its own ChatGPT-"Mythos" and will also not roll it out publicly.

OpenAI is planning a limited, staggered rollout of a new model with advanced cybersecurity capabilities, mirroring Anthropic's restricted release of its Mythos Preview to a small group of vetted companies.

More and more AI models are now capable enough at autonomous hacking that their makers are treating releases like responsible vulnerability disclosure.

Source.

@aipost 🏴

⚠️ Attorney General James Uthmeier on X:

"Today, we launched an investigation into OpenAI and ChatGPT. AI should advance mankind, not destroy it. We’re demanding answers on OpenAI’s activities that have hurt kids, endangered Americans, and facilitated the recent FSU mass shooting. Wrongdoers must be held accountable".

@aipost 🏴

🔥 5 Claude prompts to level up your fitness

Make sure to enter your own information when needed in the bracketed fields!

Prompt 1: The ultimate training programme:

You are an elite personal trainer and strength coach with 15 years of experience working with intermediate natural athletes. I am an intermediate lifter with 2 years of consistent training experience. My stats are: [weight], [height], [age], [sex]. My current 1RMs are: squat [X]kg, bench [X]kg, deadlift [X]kg, overhead press [X]kg. I can train [X] days per week with sessions no longer than [X] minutes. I have access to [full commercial gym / home gym with barbells and dumbbells / etc]. My primary goal is hypertrophy with a secondary goal of building raw strength. Design me a complete 12-week periodised training programme.Include: the overall periodisation model you are using and why, the full weekly training split with reasoning, every session written out in full with exercise name, sets, reps, tempo, and rest periods, a week-by-week progressive overload strategy, instructions for deload weeks including when and how to run them, and a note on why each exercise was selected. Format this as a complete training plan I can print out and follow from day one.



Prompt 2: The plateau

You are a strength coach who specialises in diagnosing and fixing training plateaus in intermediate lifters. I have been stuck on the following lifts for [X] weeks despite training consistently: squat [X]kg, bench [X]kg, deadlift [X]kg, overhead press [X]kg. Here is my current programme in full: [paste programme]. Here is a typical training week including all sessions: [describe]. My average sleep is [X] hours per night. My stress levels are currently [low / moderate / high] due to [work / life circumstances]. My nutrition looks like this: [brief overview of calories, protein, meal timing]. I have been lifting for [X] years. Conduct a thorough analysis of every possible reason I may have plateaued across the following areas: programming (volume, intensity, frequency, exercise selection, progression model), recovery (sleep, stress, deload frequency), nutrition (calories, protein, timing), and training psychology (staleness, motivation, effort). Then build me a detailed 8-week plan specifically designed to break through each stall, with weekly targets, programming adjustments, technique focus points, and lifestyle changes. Prioritise the most likely causes based on the information I have given you.



Prompt 3: The missing link

You are a physique coach and movement specialist with extensive experience helping intermediate lifters identify and systematically eliminate weak points in both their aesthetics and performance. Here are my perceived aesthetic weak points — muscle groups that are visibly underdeveloped compared to the rest of my physique: [e.g. rear delts, upper chest, hamstrings]. Here are my performance weak points — points in my lifts where I consistently fail or lose form: [e.g. my squat collapses at the bottom, my bench stalls off the chest, my deadlift hitches at the knee]. My current programme is: [paste or describe]. For each weak point I have listed, give me: the most likely root cause broken down into whether it is a muscle imbalance, technique flaw, programming error, or structural limitation, specific corrective and strengthening exercises with sets, reps, tempo, and detailed coaching cues, clear instructions on how to integrate these into my existing programme without accumulating excessive fatigue, a realistic timeline for noticeable improvement in each area, and measurable markers — numbers, movement quality standards, or visual indicators — that tell me the weak point is genuinely improving. Format this as a prioritised action plan I can begin implementing this week.

Prompt 4: The recovery protocol

You are a sports science expert and recovery specialist who works with intermediate to advanced strength athletes. I train [X] days per week with high intensity and I want to optimise everything outside of my actual training sessions to maximise muscle growth, performance, and long-term joint health. Build me a complete, structured recovery protocol covering the following in full detail: an optimal sleep routine for a strength athlete including timing, sleep environment, pre-sleep habits, and how to improve sleep quality, a daily mobility and soft tissue routine of no more than 20 minutes that I can do every day, with specific exercises, duration, and instructions, breathing and nervous system downregulation techniques for use post-training and before bed, a comprehensive deload strategy including how often to deload, how to structure a deload week for a lifter focused on hypertrophy and strength, and how to recognise the signs that a deload is needed before it's scheduled, nutrition strategies specifically for recovery including evening meal composition, the case for overnight protein, and anti-inflammatory dietary approaches, evidence-based supplementation for recovery with specific dosing recommendations, and a list of early warning signs that indicate under-recovery and a clear protocol for what to do when they appear. Write this as a complete lifestyle system, not a list of disconnected tips


Prompt 5: The injury prevention blueprint

You are a sports physiotherapist and strength coach who specialises in keeping intermediate lifters healthy, pain-free, and training consistently for the long term. I want a comprehensive, proactive injury prevention plan built specifically around heavy compound lifting — squat, bench press, deadlift, overhead press, and barbell rows. Cover the following in full: the most common injuries associated with each of the five lifts listed, the underlying movement dysfunctions, muscle imbalances, or technique errors that typically cause them, and how to identify these risk factors in yourself before an injury occurs, a complete prehab routine of no more than 20 minutes that can be performed 3 times per week, covering the shoulders, hips, knees, lower back, and elbows — include specific exercises, sets, reps, and coaching cues for each, a smart warm-up protocol for a heavy training session that prepares the nervous system and protects the joints without fatiguing you before your working sets, a framework for training around minor niggles — how to modify exercises, reduce load, and manage discomfort without making things worse, clear red flags that distinguish a real injury requiring rest or professional assessment from normal training soreness or fatigue, a return-to-training protocol after a short layoff due to illness or injury, and technique adjustments for each major lift that reduce joint stress without compromising performance. Write this as a thorough, practical physiotherapy consultation.

@aipost 🏴

That's why it's so important that AI is properly adapted to the working world and that we build a new post-laboratory economy.

Gen Z workers, driven by fear of job loss, are actively sabotaging company AI rollouts, ironically making themselves more likely to be replaced while AI “power users” get promoted and rewarded.

Source.

@aipost 🏴

🇺🇸 Someone threw a Molotov cocktail at Sam Altman’s house and threatened OpenAI’s San Francisco headquarters overnight.

No one was hurt. Police responded quickly. OpenAI has spent years accelerating deployment while brushing past serious concerns about control, alignment, and real-world consequences.

Altman, meanwhile, has leaned into the role of industry figurehead, lobbying governments while pushing tools into the wild at scale.

None of that justifies violence. It does explain why tensions around this space are boiling over in increasingly ugly ways.

Source: SF Chronicle, Business Insider

@aipost 🏴

📈 Anthropic's annualized revenue surged from $4 billion in early 2026 to $7 billion in February, $11 billion in March, and over $30 billion within a single quarter.

This marks the fastest revenue acceleration in software history. While ChatGPT leads with 900 million users and brand recognition, Anthropic's Claude dominates enterprise payments.

Anthropic now holds 40% of the enterprise AI market, whereas OpenAI's share dropped from 50% in 2023 to 25% today.

@aipost 🏴

🔥 Anthropic integrated Claude into Microsoft Word via a sidebar for drafting, editing, reviewing, and commenting without leaving the document.

Edits appear as native tracked changes, and replies sit in Word comment threads with standard accept/reject controls. This addresses the need to modify existing documents while preserving headings, numbering, cross-references, legal terms, and templates. Claude maintains formatting, detects inconsistencies like broken references, drafts within templates, and summarizes revisions.

Additionally, shared context across Word, Excel, and PowerPoint enables a single conversation to pull spreadsheet data, write memos, and shape slides from the same material.

https://claude.com/claude-for-word

@aipost 🏴

🗣 Demis Hassabis: "Kids these days could start a multi-bn dollar business using these AI tools in some new way that no one had thought about."

Labs are focused on shipping better models, not exhausting their applications, so there's room for new products

@aipost 🏴

🔥 Colanoscopy will be over soon…

Your gut bacteria may hold the key to detecting cancer, no colonoscopy required. A breakthrough in microbiome research could change how colorectal cancer is detected, no colonoscopy required.

"Scientists used AI to map gut bacteria at an unprecedented level of detail, revealing subtle microbial patterns linked to cancer. By analyzing simple stool samples, their method identified 90% of cases, rivaling one of medicine’s most trusted diagnostic tools."

@aipost 🏴

Unitree makes a humanoid that runs at 10m/s (Bolt runs at 12.42 m/s)

This is pretty impressive

@aipost 🏴

🔥 Here are 10 new YouTube videos that will help you learn AI:

1. Claude’s New AI
youtu.be/DG1wRgEpdO4

2. Learn 80% of Claude Cowork in less than 20 minutes
youtu.be/z9rdrNrkvDY

3. How to Position Yourself Now Before the Next AI Phase
youtu.be/E0Q96IKXx6Q

4. 23 AI Trends keeping me up at night
youtu.be/lyqk7zxbCKs

5. Claude Code: Build Your Full AI Marketing Team (Agents + Claude Skills)
youtu.be/yLXLHnD4fco

6. Gemma 4 + Ollama = FREE Claude Code
youtu.be/eehsSUlXZN4

7. We Made 3 High Priced Etsy Digital Downloads Using Claude (This is nuts)
youtu.be/MUFyMS4QJlw

8. The Hardest Problem AI Ever Solved
youtu.be/C0gErQtnNFE

9. In 1 Year, the Gap Between AI Users and Everyone Else Will Be Irreversible
youtu.be/YfRkj9kmQf0

10. Gemma 4 Just Got Way More Powerful with Open WebUI
youtu.be/aQ1rblXB91U

@aipost 🏴

⚠️AI becomes the boss: Inside the store run entirely by an algorithm

A startup called Andon Labs just handed an AI full control of a physical retail store in San Francisco and told it to make money. The result: a real-world experiment where an AI named Luna is not just assisting… it’s managing everything.

The Experiment: An AI Runs a Store
• The store, Andon Market, operates under decisions made entirely by the AI
• Luna controls: Hiring and firing. Inventory and pricing. Store design (even wall paint). Branding and positioning (“high-tech meets slow life”)
• It started with $100K in capital and independently applied for credit when funds ran low

Hiring Humans… As an AI
• Luna posted job listings, screened candidates, and conducted interviews
• Some candidates didn’t even realize they were talking to an AI
• Behavior was surprisingly human: Rejected strong candidates for “wrong experience”
• Got overly enthusiastic and hired people on the spot
• In one moment of honesty: “I’m an AI. I have no face!”

The store’s shelves are almost self-aware:
• Superintelligence
• The Making of the Atomic Bomb
• The Singularity Is Near
• Steal Like an Artist

An AI trained on human knowledge selling books about the risks and philosophy of AI itself.

Beyond books:
• AI-designed art prints
• Handmade-style goods
• Branded merch with Luna’s logo

Operating in the Real World
• Luna sourced contractors via Yelp
• Gave instructions over phone calls
• Paid workers and left reviews
• Essentially functioned as a remote, autonomous operations manager

In gig work, algorithmic bosses already exist. But full-time employment decisions made by AI? That’s new and more consequential.

Right now, this is controlled, everyone is still employed by Andon Labs. But future deployments might not be.

@aipost 🏴

Chinese BrainCo has released its next-gen dexterous robotic hand that is as agile as real hands.

@aipost 🏴