AI News: They All Launched the Same Thing!

This week brought a significant wave of updates in the AI landscape, focusing on interactivity, autonomous agents, and advanced image manipulation. The primary highlights involve new features from Anthropic and OpenAI, the expansion of Perplexity’s autonomous "Computer," and innovative layering tools in Canva. ### Interactive Visualizations: Claude vs. ChatGPT A major trend this week is the introduction of interactive learning tools. Anthropic’s Claude now allows users to generate interactive charts, diagrams, and visualizations directly within the chat. Examples include a periodic table where users can click elements for details, and a compound interest visualizer with real-time sliders. While Claude builds these visualizations from scratch—allowing for custom creations like AI release timelines—the process is relatively slow and can occasionally produce errors, such as distorted maps or broken diagrams. OpenAI released a very similar feature for ChatGPT, though it functions differently. ChatGPT’s interactive visuals are faster because they are drawn from a curated library of pre-built models rather than being coded dynamically for every prompt. These include scientific and mathematical concepts like the Ideal Gas Law, the Pythagorean theorem, and Ohm’s Law. While OpenAI's visuals are more polished and include animations, they are currently limited to specific supported categories, whereas Claude offers more flexibility for unique, unscripted visualizations. ### The Rise of Autonomous Agents: Perplexity Computer Perplexity has significantly expanded access to its "Perplexity Computer" feature, now available to all paid subscribers. This represents a shift from simple chatbots to autonomous agents. The system runs on dedicated Mac Minis and can operate 24/7, even after a user steps away. It is designed to act as a digital proxy, moving across different tools and local files to complete complex tasks. In practical demonstrations, the Perplexity Computer was shown finding job candidates by scanning professional criteria, generating slide decks from revenue reports, and building custom financial terminals. One notable use case involved a marketing agent that performed over 200 micro-optimizations on an ad stack autonomously. While there is some ambiguity about whether the agent is making final decisions or simply building high-level data dashboards, its ability to connect to various APIs (like Plaid for financial data) and organize information into functional interfaces is a major leap in utility. ### Creative Tool Enhancements: Canva and Photoshop In the realm of design, Canva introduced "Magic Layers," a feature that uses AI to deconstruct a single image into multiple independent layers. This is particularly useful for AI-generated images, which usually arrive as a flat file. With Magic Layers, users can select individual elements—such as a character, a coffee mug, or the background—and move, resize, or replace them. This allows for total recomposition of an image after it has been generated. The feature also works on standard photographs, enabling users to separate subjects from their backgrounds with high precision. Adobe Photoshop also integrated a new AI Assistant for its web and mobile versions. This tool allows users to perform complex edits using natural language prompts. By simply describing a desired change—such as adding an explosion to a background—the software uses the "Nano Banana" engine to generate and integrate the new elements into the existing project. ### Technical Milestones and New Models Several high-performance models were released this week. Nvidia introduced Neotron 3 Super, an open-weight model with 120 billion parameters. Because it is open-weight, it can be run locally on powerful hardware or fine-tuned for specific enterprise needs. Google launched Gemini Embedding 2, its first natively multimodal embedding model. This tool is designed for developers to help AI understand and retrieve information across different formats—text, images, video, and audio—simultaneously within a single space. Anthropic also targeted developers with new updates to "Claude Code." This includes the ability to schedule recurring tasks, such as daily code reviews or weekly dependency audits. Additionally, they introduced an agent-based code review system that uses a team of AI agents to identify, verify, and rank bugs by severity before a human developer ever sees the pull request. ### Autonomous Research and Strategic Shifts One of the more experimental updates came from Andre Karpathy, who open-sourced a system called "Auto Research." This setup allows a large language model to optimize its own code autonomously. The agent runs experiments overnight, checks if the changes improved performance, and either keeps or discards the modifications. This creates a loop where the AI is constantly working to make itself faster and more efficient while the developer sleeps. In corporate news, Meta acquired the team behind Moltbook, a social network designed entirely for AI agents to interact with one another. This move suggests Meta may be exploring a future where agents handle content creation and social interaction, potentially reducing the reliance on human creators. This could also signal a shift in advertising, where companies begin targeting the AI agents that will eventually make buying decisions on behalf of human users. ### Practical Updates for Everyday Use Google and Microsoft continue to bake AI deeper into their core products. Google Maps now supports conversational queries, allowing users to ask for specific recommendations along a route, such as finding public restrooms or planning stops for a road trip. Google also integrated Gemini features into Docs, Sheets, and Slides for paid users, enabling instant drafting and data filling. Microsoft followed suit by rolling out a health feature for Copilot that aggregates medical records and wearable data to provide a coherent narrative of a user’s physical well-being. Finally, ChatGPT integration arrived for Excel, providing a sidebar assistant that can build models, run data analysis, and update spreadsheets using plain language instructions.