This week in AI, creativity, and knowledge-sharing are evolving faster than ever. From Elon Musk’s new AI encyclopedia to large-scale video generation and autonomous AI scientists, here’s what’s making headlines:

📚 Grokipedia
Elon Musk’s xAI is developing Grokipedia, an AI-powered encyclopedia aimed at providing a more transparent, neutral, and bias-free alternative to Wikipedia, powered by the Grok chatbot for real-time fact-checking and context enrichment.

🎥 Wan2.2: Large-Scale Video AI
Alibaba Cloud’s Wan2.2 is a cutting-edge open-source video generative model offering cinematic-quality T2V, I2V, TI2V, S2V, and Animate capabilities. It supports text, image, and speech-driven video generation at 720P/24fps on consumer GPUs.

🔬 Periodic Labs: AI Scientists
Periodic Labs is building autonomous AI systems that conduct real scientific experiments, hypothesize, and learn from results in physics and materials science, accelerating discovery in areas like superconductors and semiconductors.

From AI-powered knowledge and cinema-quality video generation to autonomous scientific research, these breakthroughs show just how fast AI is reshaping knowledge, creativity, and innovation.

Elon Musk Announces Grokipedia to Rival Wikipedia

Elon Musk's AI company xAI is developing Grokipedia, an AI-powered online encyclopedia aimed at countering what Musk describes as Wikipedia's bias and "woke" ideology. Grokipedia will be freely available without usage limits and powered by xAI's Grok chatbot, designed to identify inaccuracies and add missing context to provide a more transparent and neutral knowledge platform. Musk envisions Grokipedia as a significant improvement over Wikipedia, targeting followers critical of Wikipedia's perceived liberal bias. Despite controversies around Grok's behavior, Musk is optimistic about transforming how people access reliable information through this new AI-driven project.

Wan2.2: Open Large-Scale Video AI

Wan2.2, developed by Alibaba Cloud’s Wan team, is an advanced open-source large-scale video generative model that builds on Wan2.1 with major upgrades, including a Mixture-of-Experts (MoE) diffusion architecture for greater capacity at similar cost, cinematic-level aesthetics trained on curated datasets, enhanced complex motion generation from a significantly larger training set, and an efficient 5B TI2V model enabling unified text-to-video and image-to-video generation at 720P/24fps on consumer GPUs. It supports multiple modalities—T2V, I2V, TI2V, S2V, and Animate—for applications ranging from cinematic video creation to character animation and speech-driven generation, with integration into ComfyUI and Diffusers. Licensed under Apache 2.0, Wan2.2 is available on GitHub, Hugging Face, and ModelScope, offering inference code, model weights, and community support for academic and industrial use.

Periodic Labs: Building AI Scientists

Periodic Labs is pioneering the creation of an AI scientist—an autonomous AI system designed to conduct scientific research by hypothesizing, experimenting, and learning from results in physical sciences. Unlike traditional AI models trained solely on finite internet data, Periodic Labs focuses on generating novel scientific knowledge through AI-operated autonomous laboratories that produce vast amounts of high-quality experimental data, including rare negative results. Their goal is to accelerate scientific discovery in areas such as high-temperature superconductors, materials design, and semiconductor manufacturing by integrating AI agents that analyze experimental data and optimize research iterations. Backed by top investors including a16z, NVIDIA Ventures, and prominent tech leaders, Periodic Labs aims to revolutionize science with verifiable simulations and AI-driven experimentation, starting with physics—the verifiable natural environment where rapid and accurate experimentation is possible. Their team includes leading experts who contributed to foundational AI and scientific advancements, positioning Periodic Labs as a forefront innovator in AI-powered scientific research and autonomous laboratories.

Hand Picked Video

In this video, we'll look at China just dropped QwQ-32B, and it’s making waves in the AI world! This model is being called the smallest yet most powerful reasoning model out there. But what makes QwQ-32B so special? How does it compare to Deepseek r1 and other AI giants?

Top AI Products from this week

• Terramind - AI search engine with deep research and multi-provider analysis

• MailoReply – Generate polished emails and smart replies with advanced AI — fast, secure, and seamlessly integrated into your workflow through our Chrome extension. Your privacy comes first: we never store your emails or replies, only the templates you choose.

• FillWise – Fill Any Web Form with AI in one click. Transform how you complete web forms with our AI-powered Chrome extension. Insert text from your CRM, spreadsheet, email, or any source. AI analyzes and matches your data with form fields, then automatically fills them.

• Illustria - Illustria is your go-to tool for generating high-quality, customizable illustrations in seconds. Whether you’re designing a website, presentation, or app, our AI crafts unique SVG illustrations tailored to your needs - no design skills required

• Cognitia - Cognitia is the AI that remembers you. No more starting over every chat. It learns, adapts, and grows with your workflow. Connect your tools, recall past convos, and get things done faster with an AI that finally feels personal.

• Cafécode - An AI-powered coding platform where you build projects you love with guided support. Stop more coding alone or getting confused vibecoding. AI helps you understand, create, and truly own every project you make.

This week in AI

• Alexa+: Smarter AI Assistant - Alexa+ is Amazon’s next-gen AI assistant, more conversational, personalized, and capable, controlling smart homes, making reservations, and managing tasks seamlessly. Free for Prime members.

• Opera Neon AI Browser - Opera Neon is an AI-powered browser that acts on tasks, manages tabs, automates workflows, and protects privacy. It offers AI agents, reusable templates, and local data processing, priced at $19.99/month.

• Replit Connectors - Replit Connectors enable seamless, secure app integration with tools like Google, GitHub, and Notion—no credential management, one login with reusable connections. Build faster, safer.

• Dreamer 4 AI Agent - Dreamer 4 is a scalable AI agent that trains inside a fast, accurate world model, mastering complex tasks like mining diamonds in Minecraft purely from offline data, advancing robotics potential.

• Measurement Trees for AI Evaluation - Measurement trees provide hierarchical, interpretable metrics for AI evaluation, integrating varied data (user feedback, expert input) to assess system validity, reliability, and risks

• KG & NLP for Crime Chat Data - This paper integrates knowledge graphs and NLP to semantically enrich instant messaging data for criminal investigations, enhancing search, entity extraction, and insights from audio transcripts

Paper of The Day

Large language models exhibit a notable self-preference bias, favoring their own outputs regardless of human evaluations. This bias is linked to familiarity, as models assign higher ratings to texts with lower perplexity. While some models show strong self-preference, others display reverse bias, undervaluing their responses. This phenomenon raises concerns about objective evaluations by AI and highlights the need to understand and mitigate biases in LLM assessments.

To read the whole paper 👉️ here.