AI Pulse 🚨
A bite-sized curation of this week's most important AI news.
🛠️ Cursor 2.0 introduces Composer and multi-agent interface
Cursor 2.0 launched with Composer, a new coding model that delivers faster and more accurate code generation. The release includes a redesigned interface built for managing multiple AI agents working simultaneously on different project aspects.
🛠️ Cursor launches Cloud Agents for remote task execution
Cursor introduced Cloud Agents, allowing developers to run AI-powered coding tasks remotely without keeping laptops online. Users can assign bug fixes, todos, and feature development to agents that run independently in the cloud.
🛠️ Microsoft 365 Copilot now builds apps and workflows
Microsoft 365 Copilot launched App Builder and Workflows, allowing users to create full-stack applications and automate business processes using only natural language prompts. Users can describe the app they need—such as a project tracker with dashboards—and Copilot generates a working application complete with database backend, UI, and security controls.
🤖 GitHub launches Agent HQ for seamless AI agent integration
GitHub introduced Agent HQ, a platform designed to integrate AI agents directly into development workflows. The system enables developers to use any AI agent within GitHub’s ecosystem, creating a unified environment for agent-powered development.
🤖 OpenAI introduces Aardvark, an AI security researcher agent
OpenAI launched Aardvark in private beta, an autonomous AI agent designed to identify and fix software vulnerabilities. The system can conduct security research without human supervision, scanning code for weaknesses and suggesting fixes.
🛒 PayPal becomes first payment wallet in ChatGPT
PayPal integrated its digital wallet directly into ChatGPT, allowing users to purchase items without leaving the AI chat interface. This makes PayPal the first payment processor embedded in a conversational AI platform, marking a shift in how e-commerce might function.
💼 Claude expands financial services offering with Excel integration
Anthropic expands Claude for Financial Services, featuring an Excel add-in, real-time market data connectors, and pre-built Agent Skills for financial tasks.s
🚀 OpenAI adds reusable characters and video stitching to Sora
OpenAI updated Sora with the ability to create reusable avatars for AI-generated videos, plus new video stitching and leaderboard features. Users can now maintain character consistency across multiple video generations, a major limitation in previous text-to-video models.
💰 Nvidia reaches $5 trillion valuation on AI demand
Nvidia became the first company to achieve a $5 trillion market valuation, driven by surging demand for AI chips and infrastructure. The milestone reflects massive capital flowing into AI development, with Nvidia’s GPUs serving as essential infrastructure for training and running large language models.
🔍 Perplexity launches Patents for AI-powered patent search
Perplexity introduced Perplexity Patents, an AI-powered search tool that makes patent databases accessible and easy to query. Users can search for patents using natural language instead of complex boolean queries and legal terminology.
🔧 Google launches Jules extension for Gemini CLI
Google introduced the Jules extension for Gemini CLI, enabling developers to manage coding tasks asynchronously through the command line. The extension allows developers to assign work to Gemini agents and check back later for results, similar to Cursor’s Cloud Agents.
🚀 Cognition releases SWE-1.5 with 13x speed improvement
Cognition launched SWE-1.5, a fast agent model for software engineering that achieves near state-of-the-art performance while running 13 times faster than previous versions. The model is designed for autonomous software engineering tasks like debugging, feature development, and code refactoring.
🛠️ OpenAI releases gpt-oss-safeguard for content moderation
OpenAI launched gpt-oss-safeguard, open-weight models for safety classification tasks that allow developers to customize content moderation policies. Unlike closed models, developers can fine-tune these models for specific use cases and deploy them on their own infrastructure.