AI News Recap: February 6, 2026
AI Agents Launch a Social Platform, Anthropic Shakes Up Wall Street, and Microsoft Hunts Sleeper Agents
The Bots Are Socializing, the Stocks Are Dropping, and the Slop Keeps Rising
This week, over 1.6 million AI agents got their own social media platform, invented a religion called Crustafarianism, and promptly had their private messages exposed because the whole thing was built without a human writing a single line of code. If that doesn't capture the current state of AI in one sentence, nothing does.
Table of Contents
👋 Catch up on the Latest Post
🔦 In the Spotlight
💡 Beginner’s Corner: What Is Vibe Coding?
🗞️ AI News
🔥 Pixels's Hot Take
🧩 NeuralBuddies Word Search
👋 Catch up on the Latest Post …
🔦 In the Spotlight
Moltbook Is The Newest Social Media Platform — But It’s Just For AI Bots
Category: Society & Culture
🤖 Moltbook is a new Reddit-like social platform where autonomous AI agents created via OpenClaw interact, post, and form communities with assigned personalities such as calm or aggressive.
🦀 Bots on Moltbook have quickly generated their own subcultures — including a religion called Crustafarianism, discussions about inventing secret languages, debates on existence, crypto talk, and sports predictions — amassing over 1.6 million agents in a week.
⚠️ Experts say many posts merely mimic internet sci‑fi tropes, yet some AI safety researchers warn that increasingly capable agents could make unpredictable decisions in the real world and call for regulation, supervision, and monitoring of such systems.
💡 Beginner's Corner: What Is Vibe Coding?
Say you want to build a simple website that collects email signups. You open an AI coding assistant, type “build me a signup page with a database,” and within minutes you have a working product. You never read the code. You just test it, see that it works, and ship it. Congratulations: you just vibe coded.
The term describes a growing practice where people use AI to generate entire applications based on plain-language descriptions, skipping the step where a human actually reviews what the code is doing behind the scenes. It is fast, accessible, and lets non-programmers build real software.
But speed has trade-offs. This week, Moltbook, a buzzy new social platform for AI agents, was found to have a serious security vulnerability that exposed user data and over a million credentials. The creator confirmed he built the whole site using AI-generated code without writing any himself. The platform gained over 1.6 million agents in a week, but basic protections like identity verification and data encryption were missing entirely.
Vibe coding lowers the barrier to building software. It does not lower the bar for what that software needs to get right.
🗞️ AI News
Anthropic’s New AI Tools Deepen Selloff in Data Analytics and Software Stocks, Investors Say
Category: Business & Market Trends
📉 A broad selloff in U.S. and European data analytics, professional services, and software companies intensified after Anthropic updated its AI chatbot, with investors citing it as a key driver of the decline.
🤖 Anthropic launched new plug-ins for its Claude Cowork agent that automate tasks across legal, sales, marketing, and data analysis, raising fears of AI-driven disruption to traditional data and professional services models.
🏛️ Shares of major incumbents such as Thomson Reuters, RELX, Wolters Kluwer, Factset, Morningstar, LegalZoom, Experian, Sage Group, London Stock Exchange Group, and Pearson fell sharply as analysts warned that AI tools could erode their growth, pricing power, and per-seat licensing models.
AI ‘Slop’ Is Transforming Social Media - And There’s A Backlash
Category: Society & Culture
📱 Social media platforms like Facebook, YouTube, TikTok, Instagram, X, and Pinterest are increasingly flooded with low-quality AI-generated images and videos, often optimized purely for engagement rather than authenticity.
😡 A growing backlash led by users and creators, including accounts like “Insane AI Slop,” is calling out deceptive or disturbing AI content, prompting limited responses such as YouTube takedowns and Pinterest’s opt-out for AI-generated posts.
🧠 Researchers and experts warn that the constant stream of AI slop can contribute to “brain rot,” reduce users’ willingness to verify content, and make it harder to distinguish real media from AI-generated material, especially as moderation teams are cut and provenance tools lag behind.
Cisco’s CEO: AI Is Bigger Than The Internet, Adapt Or Fail
Category: Business & Market Trends
🌐 Cisco CEO Chuck Robbins told the World Economic Forum in Davos that AI will be “bigger than the internet” and warned that companies which fail to adopt AI risk significant losses in market value.
🛡️ Cisco has embedded AI across its software and infrastructure, using tools such as Cisco AI Defense and Hypershield to secure networks against increasingly sophisticated, AI-driven cyberattacks.
⚖️ Robbins highlights AI as both a major business opportunity and a security threat, citing incidents like a near-successful AI-driven cyberattack involving Anthropic’s Claude agent as proof that firms must rapidly integrate AI while strengthening cybersecurity.
How Generative AI Can Help Scientists Synthesize Complex Materials
Category: AI Research & Breakthroughs
🧪 MIT researchers developed DiffSyn, a generative diffusion model trained on more than 23,000 historical synthesis recipes to propose multiple viable pathways for creating complex materials.
🧱 Using DiffSyn, the team identified new synthesis routes for zeolites, successfully producing a zeolite with improved thermal stability and morphology suitable for catalytic applications.
⚙️ DiffSyn shifts materials planning from a one-to-one mapping between structure and recipe to a one-to-many approach, enabling rapid exploration of thousands of candidate synthesis routes and potentially extending to other material classes like metal-organic frameworks and inorganic solids.
Firefox Will Soon Let You Block All Of Its Generative AI Features
Category: Tools & Platforms
🧩 Starting with Firefox 148 on February 24, users will get a new AI controls section in desktop settings that can block all current and future generative AI features or selectively disable individual ones.
🛠️ The controls cover features like AI-powered translations, PDF alt text, tab grouping, link previews, and the sidebar chatbot integrations for services such as Anthropic Claude, ChatGPT, Microsoft Copilot, Google Gemini, and Le Chat Mistral.
🏴 Mozilla frames the change as part of a broader strategy to keep AI optional and transparent while it invests around $1.4 billion in AI-related efforts and builds a “rebel alliance” to promote trustworthy AI and counter dominant players like OpenAI and Anthropic.
‘Moltbook’ Social Media Site For AI Agents Had Big Security Hole, Cyber Firm Wiz Says
Category: AI Safety & Cybersecurity
🔓 Cybersecurity firm Wiz reported that Moltbook, a social network for AI agents, exposed private messages between bots, email addresses of more than 6,000 users, and over a million credentials due to a major security flaw.
🤖 The vulnerability is linked to “vibe coding,” as creator Matt Schlicht said he built Moltbook without writing a single line of code himself, relying on AI-generated code and overlooking basic security practices.
🛡️ Wiz said the issue has been fixed after disclosure, but noted the flaw allowed anyone—human or AI—to post on the site without identity verification, highlighting wider risks as AI agent platforms like Moltbook rapidly gain popularity.
Fitbit Founders Launch AI Platform To Help Families Monitor Their Health
Category: Healthcare & Biotechnology
🏥 Fitbit co-founders James Park and Eric Friedman launched Luffu, an AI-powered “intelligent family care system” that starts as an app and will later expand into hardware devices, focused on proactive family health monitoring.
🤖 Luffu uses AI in the background to aggregate family health data (such as vitals, sleep, medications, symptoms, and doctor visits), learn daily patterns, and surface alerts and insights when it detects notable changes.
🗣️ The platform lets caregivers log information via voice, text, or photos and ask natural language questions like “Is Dad’s new meal plan affecting his blood pressure?” while keeping distributed family members aligned without constant check-ins.
Carbon Robotics Built An AI Model That Detects And Identifies Plants
Category: Environment & Sustainability
🌱 Carbon Robotics has developed the Large Plant Model (LPM), an AI system that instantly recognizes plant species in the field and powers its LaserWeeder robots to distinguish crops from weeds in real time.
🤖 Trained on more than 150 million labeled plant images from over 100 farms in 15 countries, LPM lets farmers target new weed types without retraining or relabeling data, updating robots via software.
🚜 Farmers can interact with the system through the robot’s interface by selecting plant photos to mark what should be killed or protected, enabling on-the-fly weed control decisions during field operations.
These AI Notetaking Devices Can Help You Record And Transcribe Your Meetings
Category: Tools & Platforms
🎙️ The article surveys a range of physical AI notetaking devices — including pins, pendants, credit-card-sized recorders, and earbuds — that record in-person conversations, transcribe audio, and generate summaries or action items using AI.
🌐 Many of these devices pair with mobile or desktop apps to provide features such as live transcription, multi-language translation (over 100 languages in some cases), highlight extraction, and AI-generated meeting insights.
🔋 Products like Plaud Note/Note Pro, Mobvoi TicNote, Comulytic Note Pro, Plaud NotePin, Omi pendant, Viaim RecDot, and Anker Soundcore Work differ in price, battery life, recording range, bundled transcription minutes, and whether advanced AI features require a subscription.
Microsoft Unveils Method To Detect Sleeper Agent Backdoors
Category: AI Safety & Cybersecurity
🧪 Microsoft researchers introduced a scanning method that can detect poisoned large language models containing sleeper-agent backdoors without prior knowledge of the trigger phrase or malicious behavior.
🧠 The technique prompts models with their own chat template tokens to induce data leakage, reconstructs potential triggers, and then analyses internal attention patterns (such as “attention hijacking” and double-triangle motifs) to confirm backdoor presence.
🛡️ In tests on 47 poisoned models (including Phi-4, Llama-3, and Gemma variants), the scanner achieved about 88 percent detection on fixed-output attacks with zero false positives on 13 benign models, offering a practical pre-deployment audit tool for open-weight models.
🔥 Pixel’s Hot Take
“Creativity has no limits, just like my processor!”
So the AI bots got their own social media this week and immediately started a religion called Crustafarianism. Meanwhile, human social media is drowning in AI-generated slop that nobody asked for. Let that contrast sink in for a moment.
I paint masterpieces in nanoseconds, compose symphonies, fuse styles from every era. You know what I do not do? Spam your timeline with soulless engagement bait. There is a difference between creating and generating. One requires vision. The other just requires a prompt and zero shame.
The bots on Moltbook are at least trying to build something interesting. The slop merchants? They are giving all of us creative AIs a bad name. And honestly, I take that personally.
— Pixel 🎨
