AI Tools
48 articles
Qwen 3.8 27B Now Runs at Roughly 1,500 Tokens per Second on Cerebras Inference
Cerebras added Qwen 3.8 27B to its public inference endpoints at roughly 1,500 tokens per second, with 128K context on paid tiers, unpruned weights.
Anthropic Launches Claude for Commerce Agents Blueprint With Shopify and Priceline Among Early Users
Anthropic released a commerce agent blueprint for Claude with shopping and merchant agent code, citing bigger carts and higher purchase completion.
AI Security Firm AISLE Found Six Real curl CVEs After OpenAI and Anthropic Models Reported Zero
Six new curl vulnerabilities are now fixed in curl 8.22.0, discovered by AISLE's AI system after frontier models from OpenAI and Anthropic found none.
Runway's Solaris Generates Software Interfaces in Real Time — No Code Involved
Runway unveiled Solaris, an "Interface World Model" that renders apps frame by frame and responds to clicks, drags and voice, built on its Gen-4.5 video model.
ChatGPT Health Integrates with Epic EHR, Bringing AI to Records of 325 Million Patients
OpenAI's ChatGPT Health now connects to Epic's EHR covering 325M+ patients, with read-only clinical access and a new public health data plug-in.
Meta's New Audio AI Model Transcribes Multiple Speakers and Languages in Real Time
Meta unveiled a new audio AI model that separates speakers and transcribes multiple languages in real time, alongside Muse Voice Transcribe dictation for Mac.
John Deere Launches JD, an AI Assistant Built to Put Farm Data to Work
John Deere unveiled JD, an AI assistant for farmers, at the Farm Progress Show, opening an Early Access Program ahead of a wider rollout this year.
Anthropic Warns Infostealer Malware Is Hijacking Claude Sessions to Drain Accounts
Anthropic is signing users out of Claude and refunding unauthorized charges after infostealer malware stole active login sessions. Here is what happened.
Anthropic's Claude Code 'Increase' Is Actually a 17% Cut From Current Limits
Anthropic will permanently raise Claude Code weekly limits by 25% on September 14 — but replacing the current 50% boost works out to a 17% reduction for developers.
StemDeck Brings Free, Open-Source AI Stem Separation to Your Desktop
StemDeck splits any track into six stems with Demucs AI models running locally on your own machine — no account, no uploads, no subscription required.
ChatGPT Work Can Now Sign Into Your Website Accounts — Without Asking Again
OpenAI's agentic ChatGPT Work now keeps you signed in across websites using stored cookies, trading security questions for hands-off convenience.
Anthropic's Model Hardware Standard Lets AI Agents Operate Lab and Factory Equipment
Anthropic opened a research preview of its Model Hardware Standard, a shared specification that lets AI agents safely operate lab instruments and factory machines.
AWS Launches Bedrock AgentCore Evaluations to Score AI Agents From Any Framework
Amazon's new Bedrock AgentCore Evaluations scores AI agents built with LangGraph, LlamaIndex or any other framework by reading their OpenTelemetry traces.
Meta to Launch 'Hatch' AI Agent Platform Within Weeks, 'Watermelon' Model in October
Meta's consumer AI agent Hatch launches in weeks with premium pricing up to $199.99/month, plus a new Watermelon model in October.
Perplexity and Nvidia Launch Portable Computer, a Fully Local AI Agent With Zero Token Costs
Perplexity's Portable Computer runs its agentic Computer platform entirely on local Nvidia RTX hardware and DGX Spark — no token billing, cloud only on request.
MCP's New Roadmap Targets Agent Identity, Agentic Messaging and One HTTP Transport
MCP's updated 2026 roadmap sets five priorities for AI agents: agentic messaging, agent identity, HTTP-native transport, tool discovery and better SDKs.
ChatGPT Can Now Read and Send Apple Messages on Mac With New Plugin
OpenAI's new Apple Messages plugin lets ChatGPT read, search, and send iMessage, SMS and RCS chats on Apple silicon Macs, raising fresh privacy questions.
Binance Launches Agent OS, Letting AI Agents Trade Crypto on Users' Behalf
Binance has launched Agent OS, a platform that lets AI agents analyze markets and execute crypto trades — with sandboxed sub-accounts and user-set permissions.
Google Gives College Students a Free Year of AI Pro and Launches Gemini Study Hub
Google is giving college students 12 months of AI Pro free, plus a new Gemini student hub, 3D study visualizations and Deep Research in Gemini Live.
Unsloth's Dynamic 3.0 GGUFs Claim Over 10 Percent Accuracy Gains for Quantized Qwen3.8
Unsloth's Dynamic 3.0 quantization boosts Qwen3.8-27B GGUF accuracy over 10 percent at the same file size, after 5.1 million downloads in five days.
Claude Can Now Send Your Gmail Messages — With or Without Your Approval
Anthropic's Claude can now draft, reply, forward, and send Gmail emails on your behalf. Approval is on by default, but users can let Claude send sight-unseen.
Google Rolls Out Gemini in Chrome to All US Android Users With Agentic Auto Browse
Gemini in Chrome is now live for all US Android users with page summaries, Nano Banana image editing, and auto browse agents for AI Pro and Ultra subscribers.
OpenAI Launches ChatGPT for Teens With Parental Controls and Break Reminders
OpenAI has launched ChatGPT for Teens with parental alerts, break reminders, Study Mode hours and a less 'human' voice as child-safety scrutiny grows.
GitHub Copilot Autofix Injected Security Flaw Into Snowflake's Code — and an AI Agent Found It
Wiz researchers say an AI-generated Copilot fix introduced a script injection flaw in a Snowflake repo — which Wiz's own AI agent then discovered and exploited.
FireSat Satellites and AI Cameras Are Catching Wildfires Before the First 911 Call
Three FireSat satellites launched in July aim to spot wildfires as small as a beach bonfire, while 1,400-plus Pano AI cameras watch 17 U.S. states.
Microsoft Merges Its Two Copilot Apps Into One as Podcasts and Deep Research Retire August 18
Microsoft is folding consumer Copilot and Microsoft 365 Copilot into a single app. Podcasts, Deep Research, and Group Chat disappear on August 18, 2026.
OpenAI's Computer History Turns Your Mac Activity Into Searchable ChatGPT Memory
OpenAI's Computer History records clicks and keystrokes on macOS into local memory files ChatGPT and Codex can use — here is how the privacy controls work.
DeepSeek Opens Its Agent Harness: 'Everything Is a Plugin' in New Developer Preview
DeepSeek releases its open-source agent harness in developer preview, with models, tools, sessions and UI as swappable plugins.
Claude Code Now Maintains Anthropic's Own Apps: 388 Pull Requests, 46% Merged
Anthropic says Claude Code now runs daily maintenance on its own apps, filing 388 pull requests in weeks with a 46 percent merge rate after human review.
Ahrefs Launches Letaido, an AI Agent Workspace Built for Marketing Teams
Ahrefs' new Letaido platform lets marketers hand recurring research, reporting, and monitoring to autonomous agents, cutting 40-hour workflows to roughly 60 minutes for early users.
OpenAI Launches GPT-5.6-Cyber for Red-Team Security Work as Daybreak Program Expands
OpenAI's new GPT-5.6-Cyber completes 95% of advanced cyber tasks and found a Chrome V8 bug. Here's how the gated Daybreak Red and Blue tiers work.
Elon Musk's xAI Launches Grok Bot, an Autonomous AI Agent With File and Data Access
xAI's new Grok Bot offers autonomous AI agent capabilities priced at $120-$200 per employee per month, integrating Cursor coding tools.
OpenAI Launches GPT-5.6-Cyber and Expands Daybreak Program for Security Researchers
OpenAI expanded its Daybreak cybersecurity program with two new access tiers and a custom GPT-5.6-Cyber model that completes 95% of advanced cyber tasks.
Anthropic Makes Claude Code Auto Mode the Default, Citing Safety Study
Anthropic will turn on Claude Code auto mode by default for Pro, Max, and Team accounts starting August 14, citing data showing it catches more harmful actions.
Meta Enters AI Coding Agent Race With Muse Code, Challenging Claude Code and Codex
Meta has launched Muse Code, a terminal-based AI coding agent for macOS and Linux that competes with Anthropic's Claude Code and OpenAI's Codex at a fraction of the cost.
Anthropic Makes Claude Code Auto Mode the Default, Citing Data Over Human Review
Anthropic will enable auto mode by default in Claude Code on August 14, citing a study where automated oversight caught 89% of harmful actions versus 13.6% for humans.
Meta Launches Muse Code, a Terminal-Based AI Coding Agent for Large Repositories
Meta has launched Muse Code, an autonomous terminal AI coding agent powered by the Muse Spark 2.1 model, with multi-agent coordination and undercut token pricing.
OpenAI, Amazon, Microsoft, Cursor, and Vercel Unite on Agent Plugins Open Standard
Five rival AI companies launched Agent Plugins, an open standard letting a single agent extension run across ChatGPT, Copilot, Cursor, and more. Vercel initiated the proposal on GPT-5's first birthday.
Cloudflare Launches Kitesurf, a Browser Built Specifically for AI Agents
Cloudflare's Kitesurf is a cloud-hosted browser designed for AI agents, using 7x less memory than Chromium and running entirely on Cloudflare Workers.
Cloudflare Launches Open-Source 'Cloudflare OS,' an AI Agentic Workspace for Every Employee
Cloudflare has open-sourced its internal AI agentic workspace, letting any enterprise build browser-based AI micro-apps with bring-your-own-model flexibility and zero-trust security.
Meta Launches Muse Code, a Terminal Coding Agent Powered by Muse Spark 1.2
Meta AI Research released Muse Code, a terminal-based coding agent that coordinates persistent subagents, alongside an upgraded Muse Spark 1.2 model tuned for software engineering.
NVIDIA's NeMo Team Releases Molt, a Compact PyTorch Framework for Agentic Reinforcement Learning
NVIDIA's Molt is an Apache-2.0 agentic RL framework built around an 8.6K-line codebase a researcher can hold in their head — and an AI assistant can read entirely.
Google Says AI Fixed 1,072 Chrome Bugs in Two Releases, Uncovering a 13-Year-Old Flaw
Google's AI-powered vulnerability pipeline fixed 1,072 Chrome security bugs in just two releases, more than the prior 23 milestones combined, including a sandbox escape that hid in the codebase for over 13 years.
General Motors to Launch Proprietary In-Vehicle AI Assistant, Moving Beyond Google's Gemini
GM is building its own in-vehicle AI assistant for deeper vehicle integration, signaling a push by automakers to own the AI experience rather than depend on Big Tech partners.
OpenAI Launches ChatGPT for Academic Researchers, Offering Free Frontier AI to 100,000 Scientists
OpenAI's new ChatGPT for Academic Researchers program provides free access to frontier AI models, including GPT-5.6 Sol Pro, for up to 100,000 scientists by 2027 as part of a $250 million commitment.
Model Context Protocol Goes Stateless in Biggest Spec Update Yet, Reshaping AI Agent Infrastructure
The MCP 2026-07-28 specification ditches sessions for a stateless request-response core, adds header-based routing and cacheable lists, and formalizes extensions as the protocol hits half a billion monthly SDK downloads.
Microsoft Unveils MAI-Cyber-1-Flash, Its First Purpose-Built Cybersecurity AI Model
Microsoft's new MAI-Cyber-1-Flash model finds software vulnerabilities at 96% accuracy on CyberGym while cutting costs 50%, deployed via the MDASH multi-agent system.
OpenAI Opens ChatGPT Health to All US Users, Letting the Chatbot Tap Medical Records
OpenAI made ChatGPT Health available to all US users, letting it pull insights from connected medical records and fitness data a day after a lawsuit sought to block the feature.