AI News

Adobe, Canva, Figma and others are building ChatGPT plugin extensions

OpenAI Developers says Adobe, Canva, Figma, Shopify, Instacart, Atlassian, tldraw, and MagicPathAI are building with plugin extensions for ChatGPT.

Microsoft AI brings MAI-Code-1.1-Flash to GitHub Copilot for local execution

Developers will be able to run agentic coding tasks on-device using MAI-Code-1.1-Flash inside GitHub Copilot without paying inference fees for local model calls.

Polymarket puts the chance of U.S. and China agreeing to pace the AI frontier at 8%

Polymarket estimates an 8% chance that the United States and China will agree to pace the AI frontier, according to its prediction market.

Claude Python and TypeScript SDKs add built-in computer and browser use toolsets

Claude's Python and TypeScript SDKs now include built-in toolsets for computer use and browser use, running the execution loop and sending actions directly to drivers.

Study finds keeping agent harness prompts and tools small cuts costs by up to 3x

A study comparing Claude Code, mini-SWE-agent, and OpenCode with the same model on SWE-bench Verified found that minimizing harness prompts and tool schemas reduces costs by up to 3x without lowering accuracy.

Scale Labs introduces Humanity's Sixth Sense visual reasoning benchmark

Scale Labs launched Humanity's Sixth Sense, an evaluation benchmark spanning 522 open-ended tasks across images and video to test everyday intuitive visual reasoning.

SambaNova adds prompt caching for MiniMax M3 on SambaCloud

SambaNova has launched prompt caching for MiniMax M3 on SambaCloud, reducing input costs by 90% to $0.06 per million cached tokens and improving time to first token by 35% to 88% without code changes.

CoreWeave introduces Agent Lens to group production agent failure traces

CoreWeave has detailed Agent Lens, a tool that analyzes an AI agent's full production trace corpus to cluster conversations with identical failure modes.

OpenCode uses agents to manage issues and pull requests

OpenCode is keeping its issues and pull requests active by deploying agents to respond to users, open pull requests for issues, and guide items to closure automatically.

OpenCode adds Claude 5.5 Haiku to OpenCode and Go

Claude 5.5 Haiku is now available across OpenCode and Go.

GitHub Copilot makes local sandboxing generally available

GitHub has made local sandboxing generally available for GitHub Copilot, isolating command execution with granular controls over files, network access, credentials, and system capabilities.

Artificial Analysis scores Claude Haiku 5.5 at 43 on its Intelligence Index

Artificial Analysis scored Claude Haiku 5.5 at 43, up 26 points from the previous Haiku, and reported a 1 million token context window. Its results also cover token use, terminal tasks, factual accuracy, and an AutomationBench score the firm says is likely understated.

OpenAI introduces ChatGPT for Teens and previews College Planner

OpenAI has announced ChatGPT for Teens, an experience with default protections automatically enabled for users under 18, alongside a preview of an upcoming College Planner for U.S. high school students.

Cline Cloud Agents supports any model with free tier and discounted ClinePass subscription

Users can run Cline Cloud Agents with any AI model, including free trial access using models such as DeepSeek-V4.1-Flash. Cline is also offering ClinePass, a subscription providing approximately 5x discounted access to open-weights models.

Cline introduces Cline Cloud Agents to run coding tasks in the browser

Cline has launched Cline Cloud Agents, allowing users to run the AI assistant directly from a browser. Tasks run in a secure cloud sandbox to write code, test implementations, and open pull requests on GitHub repositories.

Runway integrates directly into ChatGPT Astra

Runway is now available directly within ChatGPT Astra, allowing users to submit briefs, run generations, and provide revision notes in the same chat window.

deepagents now loads tools dynamically when a skill is loaded

Harrison Chase says deepagents can now load tools dynamically when a skill is loaded, so tools needed by only some skills do not have to stay available all the time. With OpenAI and Anthropic models, this can be done without breaking the prompt cache.