Tag: AI agents
-
OpenAI’s AI Agents Built 948 Working Linux Root Exploits in One Run: What It Means for Anyone Running Linux
OpenAI’s autonomous pipeline reproduced 1,724 Linux kernel bugs and built 948 root exploits for about $1.2 million. What was found, what is patched, and what admins should do now.
-
OpenAI LifeVal: A New Test of How Well AI Handles Childcare, Paperwork, and Other Everyday Tasks
OpenAI’s LifeVal measures AI on everyday tasks like childcare, paperwork, and groceries. What it tests, the reported scores, and why the numbers need caution.
-
SemiAnalysis GDPx: The Best AI Agent Scores Just 12 Out of 100 on Expert Work. Here’s What the New Benchmark Shows
SemiAnalysis GDPx tests AI agents on 206 expert tasks across 11 fields. Claude Opus 5.5 leads with 12.0 out of 100. What the results show and where they fall short.
-
Vercel’s agent-desktop Hits 1.0: Computer Use for AI Agents Without Screenshots
Vercel Labs released agent-desktop 1.0, a tool that lets AI agents use desktop apps through the accessibility tree instead of screenshots. How it works.
-
DeepSeek Super-Agent: An Open-Source Coding Agent That Watches Your Repo and Team Chat
DeepSeek released an open-source Super-Agent prototype that watches a repository and team chat and proposes code changes. What it does and how to try it safely.
-
Microsoft’s Mage-VL-Cowork: A 4B Open Model That Knows When to Click and When to Code
Microsoft’s open 4B Mage-VL-Cowork switches between clicking and running code, with a reported 66.0% on OSWorld-Verified. What it does and how to test it safely.