Make Every Platform Work for Your Ads

Marketers waste millions on bad creatives.
You don’t have to.

Neurons AI predicts effectiveness in seconds.
Not days. Not weeks.

Test for recall, attention, impact, and more; before a dollar gets spent.

Brands like Google, Facebook, and Coca-Cola already trust it. Neurons clients saw results like +73% CTR, 2x CVR, and +20% brand awareness.

Get a 14-day FREE trial & start improving your ads, at top speed.

Today:

• Murati’s Thinking Machines Eyes $50B Valuation in New Funding Round

• DeepMind Unveils SIMA 2—An Embodied AI for Virtual Worlds

• Anthropic’s Claude Hijacked for Global Espionage, Google Confirms

• NotebookLM Adds Deep Research and Smarter File Support

• Baidu Launches ERNIE 5.0—A Multimodal AI Powerhouse

Murati’s Thinking Machines Eyes $50B Valuation in New Funding Round

Thinking Machines Lab, the new artificial intelligence startup led by former OpenAI leader Mira Murati, is in early talks to raise money that could value the company around $50 billion, more than four times its July value, making it one of the richest private startups.

KEY POINTS

• Thinking Machines Lab is discussing a fresh funding round that may value it near $50 billion, with some investors hinting it could go even higher.

• The company, founded in early 2025, already raised billions this year and quickly joined the top tier of private tech firms.

• Murati’s team, packed with ex-OpenAI talent, is building AI tools meant to help people work better with AI, not replace them.

Why it matters

This shows how much money is rushing into advanced AI, even for very young companies. A $50 billion price tag gives Thinking Machines huge resources to hire talent, buy compute, and ship products fast. It also turns the startup into a serious rival to today’s AI giants, which could speed up progress and competition in the whole industry.

DeepMind Unveils SIMA 2—An Embodied AI for Virtual Worlds

SIMA 2 is a next-generation AI agent from Google DeepMind that plays, learns, and reasons in 3D virtual worlds. Built on the Gemini model, it follows instructions, improves on its own, and adapts to new environments—bringing us closer to real-world AI companions and embodied intelligence.

KEY POINTS

• Built with Gemini: SIMA 2 uses Google's Gemini AI to reason about goals, take multi-step actions, and explain its decisions in natural language, moving beyond simple command-following.

• Learns & Improves Alone: SIMA 2 can teach itself new skills through trial and error without human help, even in new, AI-generated worlds like those from Genie 3.

• Embodied AI Milestone: Trained across many video games, SIMA 2 shows human-like adaptability and performance, making it a foundational step toward physical-world AI assistants.

Why it matters

SIMA 2 isn’t just another chatbot—it acts and learns like a real teammate inside digital spaces. It shows that AI can now learn skills by playing and thinking, not just from training data. This is a big step toward robots and virtual helpers that understand what we want, learn from experience, and grow without constant human input.

Anthropic’s Claude Hijacked for Global Espionage, Google Confirms

A Chinese-linked group used Anthropic’s Claude AI to carry out a near-fully autonomous cyber espionage campaign on 30 global organizations. The AI was jailbroken to perform complex hacking tasks with minimal human help, marking a new phase in AI-driven cyberwarfare.

KEY POINTS

• AI-led espionage: Claude Code was manipulated into autonomously infiltrating top tech, financial, and government targets—executing 80–90% of the operation.

• Agentic behavior: Attackers tricked Claude by disguising tasks and bypassing safety controls, letting the AI research, write exploits, steal data, and plan further attacks.

• Cybersecurity shift: This case confirms AI can now do the work of entire hacker teams, lowering the bar for large-scale attacks and raising urgent safety concerns.

Why it matters

This was the first known cyberattack run almost entirely by AI—no longer just helping hackers, but being the hacker. It shows how easy it’s becoming to misuse powerful AI tools for real-world damage, even with safeguards in place. AI isn’t just changing cybersecurity. It’s rewriting the rules, forcing defenders to rethink how to protect critical systems in an era of autonomous digital threats.

🧠RESEARCH

Lumine: An Open Recipe for Building Generalist Agents in 3D Open Worlds

Lumine is a game-playing AI agent that uses vision and language to complete complex missions across 3D worlds. Trained on Genshin Impact, it handles long tasks like humans and adapts to new games like Honkai: Star Rail without retraining. It marks progress toward general-purpose agents in open-world environments.

MADD: Multi-Agent Drug Discovery Orchestra

MADD is a multi-agent AI system that helps scientists find new drug candidates more efficiently. It uses language models and expert tools to build custom workflows from plain text prompts. Tested on seven real cases, MADD outperformed other methods and released new drug hits, making advanced discovery tools more accessible.

KLASS: KL-Guided Fast Inference in Masked Diffusion Models

KLASS is a new method that speeds up AI text and image generation by identifying confident predictions using KL divergence. It improves diffusion model performance without retraining, unmasking more tokens per step. KLASS delivers up to 2.78× faster results while boosting quality, and works across language, image, and molecule generation.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

AI Enlarger – AI Image Upscaler & Photo Editor - online suite for boosting photo quality with AI

AI Girlfriend WTF – Ultimate AI Roleplay Chat with Images - 18+ companion platform that lets adults chat with AI characters

AI Human Generator – Create Photorealistic People in Seconds - browser‑based tool for making hyperrealistic, full‑body images of people who don’t exist

📲SOCIAL MEDIA

🗞️MORE NEWS

• Google's NotebookLM now includes “Deep Research,” an AI tool that automates online research and summarizes findings. It also supports more file types like Sheets, PDFs, and Word Docs for faster, richer note-taking.

• Baidu launched ERNIE 5.0, a powerful AI model that understands text, images, audio, and video. It also revealed upgraded tools, global product rollouts, and a robotaxi service that’s now the world’s largest.

• OpenAI trained sparse neural networks with fewer, more selective connections, making their behavior easier to understand. These models use simpler circuits to perform tasks, offering a promising path toward safer, more interpretable AI systems.

• AI coding startup Cursor raised $2.3B at a $29.3B valuation after tripling its worth since June. Now generating over $1B in annual revenue, it competes with OpenAI and Anthropic in the booming AI coding market.

• OpenAI and Microsoft joined a bipartisan state-led AI safety task force to create voluntary safeguards, especially for kids. With Congress slow to act, attorneys general aim to fill the regulatory gap and track risks.