Skip to content
IFAI

AI agents

Industry tag1,473 stories

  1. ResearchReach 31

    Why massive context windows could unlock true AI continual learning

    Epoch AI just outlined why ultra-long context inference is far more than an easy way to ingest giant PDFs: it could be the breakthrough needed for continual learning and reinforcement learning scaling. Expanding context to millions or billions of tokens allows agents to retain months of work experience and self-correct across complex tasks. Why it matters: while wall-clock research slowdowns and high serving costs remain real bottlenecks, rapid context scaling could fundamentally alter agent autonomy.

  2. AI modelsReach 52

    xAI's Grok 4 Fast cuts reasoning costs by 98%

    xAI just launched Grok 4 Fast, a cost-efficient reasoning model packed with a 2M token context window. It matches Grok 4 on key benchmarks while using 40% fewer thinking tokens, slashing overall costs by 98% and taking the top spot on LMArena's Search Arena. It is now completely free for all users.

  3. AI toolsReach 64

    Notion's new AI agents will basically do your job

    Notion just rolled out version 3.0 to all users, headlined by autonomous AI Agents designed to act like human teammates. The agents can build pages and databases, pull info across Slack and the web, and execute tasks across hundreds of pages for up to 20 minutes at a time. Basically, Notion built a power user inside your workspace that remembers how you like to work.

  4. AI toolsReach 41

    Google turns Chrome into an AI powerhouse

    Google just announced its biggest Chrome upgrade yet, integrating Gemini directly across desktop and mobile. Users can now summarize multiple open tabs, recall past browsing via natural language, and search straight from the address bar, with autonomous agents for booking appointments coming soon. The rollout starts for U.S. desktop users in English.

  5. AI toolsReach 34

    Zoom's new AI will tell you which meetings to skip

    Zoom just unveiled AI Companion 3.0, pushing the assistant into agentic territory with a launch planned for this November. The standout feature, "free up my time," audits your calendar to suggest meetings you should skip, flags attendees who can be made optional, and reserves dedicated focus blocks. Why it matters: You still have to sign off on any schedule changes, but getting an AI-backed excuse to dodge pointless calls might be the ultimate workplace perk.

  6. RoboticsReach 47

    TUM's underwater AI robot can haul 550-pound trash from the seabed

    Researchers at the Technical University of Munich have built an autonomous underwater robot capable of finding and hauling ocean debris. Powered by sonar and cameras, on-board AI turns seabed scans into 3D models so a four-fingered gripper can lift up to 250 kilograms without crushing fragile objects.

  7. AI toolsReach 52

    Streamlabs puts an AI producer in your streaming setup

    Streamlabs just partnered with Inworld and Nvidia to launch the Intelligent Streaming Agent, an AI tool that acts as a 3D co-host, live producer, and tech assistant. It handles live scene-switching, auto-clips victory moments, and alerts streamers if their mic is muted. Solo creators can now run broadcast-level production right from the Streamlabs Desktop app.

  8. AI toolsReach 42

    GitHub launches an MCP Registry for AI agents

    GitHub just launched the GitHub MCP Registry, creating a central home base to discover and install MCP servers without digging through random repos. It ranks servers by GitHub stars and activity, plugs directly into VS Code with one-click installs, and works across Copilot and other agentic tools. Why it matters: Connecting AI agents to dev tools was messy and fragmented, and GitHub just made finding trusted integrations frictionless.

  9. ResearchReach 31

    Sakana AI builds an AI agent to write and optimize CUDA kernels

    Sakana AI just unveiled an agentic framework and a new benchmark called robust-kbench to automate low-level CUDA kernel engineering. The system translates PyTorch code into raw CUDA kernels and uses an evolutionary meta-generation loop with LLM verifiers to iteratively boost runtime speed. Why it matters: Squeezing maximum performance out of GPUs is notoriously tedious, and handing kernel optimization over to AI agents could save developers months of manual tuning.

  10. AI toolsReach 41

    Microsoft Copilot Studio agents can now control your computer

    Microsoft just launched a public preview of computer use in Copilot Studio, allowing AI agents to interact with websites and desktop apps just like a human. Powered by Windows 365 hosted browsers, agents can handle forms and dashboards even without APIs. Currently available for U.S.-based environments.

  11. AI toolsReach 54

    OpenAI upgrades Codex for autonomous coding

    OpenAI just overhauled Codex with a specialized model fine-tuned for real-world software engineering. The agent dynamically adjusts its thinking time, can run autonomously for over seven hours on massive refactors, and automatically reviews GitHub PRs for bugs. Why it matters: AI coding is shifting from line-by-line autocomplete into an independent teammate that tackles complex tasks end-to-end.

  12. Open sourceReach 46

    H Company drops Holo1.5 for Computer Use

    H Company just open-sourced Holo1.5, a family of 3B, 7B, and 72B models built specifically for Computer Use agents. The update boosts UI element localization accuracy by over 10% compared to its predecessor and handles complex software interfaces like Photoshop and VSCode. All weights are now accessible on Hugging Face.

  13. IndustryReach 41

    Albania taps an AI minister to fight procurement corruption

    Albania just appointed Diella, an AI-powered virtual avatar, as the world’s first non-human cabinet minister to take over all public procurement. Prime Minister Edi Rama handed government tender decisions over to the algorithm to eliminate chronic corruption and make spending transparent. Why it matters: It is the boldest experiment yet in putting high-stakes state governance entirely in the hands of an AI agent.

  14. AI toolsReach 42

    Replit Agent 3 tests its own apps in a real browser

    Replit just launched Agent 3, claiming a massive leap in autonomy over V2. The AI can now open an actual browser preview to click buttons, test user flows, and fix bugs automatically—3x faster and 10x cheaper than Computer Use models. It also runs up to 200 minutes solo and can build bots for Slack and Telegram.

  15. AI toolsReach 34

    Google rolls out AI Max for Search to all advertisers

    Google just launched AI Max for Search campaigns in beta to all advertisers globally, offering a one-click setup to optimize search ads using Google AI. The update brings built-in one-click A/B testing across Google Ads platforms and APIs, with custom brand-safety text guidelines rolling out later this fall.

  16. AI toolsReach 42

    ChatGPT finally adds support for MCP tools

    OpenAI just rolled out full MCP tool support inside ChatGPT. In developer mode, developers can now create connectors for write actions like updating Jira tickets or triggering Zapier workflows, rather than just fetching data. It basically turns ChatGPT into a practical automation hub for daily engineering workflows.

  17. AI toolsReach 47

    Replit launches Agent 3 with 3-hour autonomous coding

    Replit just launched Agent 3, an AI developer that can autonomously build, test, and debug code for over three hours straight without human intervention. Backed by a fresh $250M funding round at a $3B valuation, CEO Amjad Masad called it software's "full self-driving moment"—using a multi-agent architecture to boost uninterrupted runtime tenfold over its predecessor.

  18. AI toolsReach 37

    Anchor founders return with Oboe to turn any prompt into a custom course

    The co-founders behind Anchor just launched Oboe, an AI-powered platform that builds personalized learning courses in seconds from a single prompt. Powered by an internal multi-agent pipeline, it generates material across nine formats, including two-host podcasts, lecture audio, and interactive quizzes. Users get five free course creations monthly, with paid tiers starting at $15.

  19. AI toolsReach 31

    Perplexity Comet adds local MCP to bridge web and desktop files

    Perplexity is prepping local Model Context Protocol (MCP) support for its Comet browser. While Comet previously only supported remote connections, local MCPs allow its agent to browse local folders, attach files, and trigger desktop apps from chat prompts. Why it matters: It pushes AI browsers beyond the web sandbox into full desktop automation, though the feature is still tucked away in dev builds.

  20. AI toolsReach 50

    ElevenLabs launches automated testing for AI agents

    ElevenLabs just rolled out Agents Testing, a new suite that lets developers run structured simulations on conversational agents before shipping updates. It validates tool calling, human handoffs, and guardrails, and can turn past customer chats into test suites inside your CI/CD pipeline. Why it matters: teams can finally update prompts without fear of silent regressions.