AI Agents
-
Agents
DIYgod open-sources Pinky, a desktop bridge that lets AI agents operate a real iPhone or Android phone while a gateway holds risky taps for human approval
RSSHub developer DIYgod released Pinky on GitHub on November 6, 2026 under Apache-2.0. The pre-alpha app for macOS, Windows and Linux exposes a connected phone to Claude Code, Codex, Cursor and other MCP agents, and its …
-
Agents
Microsoft's Daniel Meppiel proposes Agent Contracts and signed receipts so teams can merge agent-built code they have not read
In an essay on The AI Native Mind, Daniel Meppiel, Director of Agentic SDLC go-to-market at Microsoft and creator of APM, proposes versioned Agent Contracts, a standard receipt built on SLSA, in-toto and CycloneDX, and o…
-
Safety & Ethics
OpenAI says its agent pipeline produced 948 working Linux kernel root exploits, with candidate patches, in a single run
OpenAI says an agent pipeline running on an unreleased cyber model found, validated and exploited 948 distinct Linux kernel privilege escalation bugs in one October run for about $1.2 million in API credits, and that it …
-
Agents
Google Cloud puts its CodeMender security agent in a sandbox and makes Auth Manager for agent credentials generally available
At its Security Summit on November 4, 2026, Google Cloud added sandboxed exploit testing, human approval steps and new scan modes to the CodeMender code-security agent, made its Auth Manager credential vault for AI agent…
-
New Products
LangChain brings LangSmith Bring Your Own Cloud to Azure and Google Cloud in beta, alongside generally available AWS
LangSmith BYOC, which keeps traces, evaluations, datasets, prompts and agent deployments in a customer's own cloud account while LangChain runs the data plane, now covers Azure and Google Cloud in beta in addition to AWS…
-
Agents
Google DeepMind adds Cheminformatics Skills to Antigravity and offers 10,000 scientists its top tier free for six months
Google DeepMind has released RDKit-based Cheminformatics Skills, with ChEMBL and PubChem connectors, for its Antigravity agent workbench, and showed an agent running a lung-cancer drug rigidification study in about an ho…
-
Research
Google and Mercor launch Physical-Engineering's Last Exam, an agent benchmark on which no frontier model reaches 50%
Google Cloud AI Research and Mercor released PLE, 85 long-horizon mechanical, electrical, civil and structural engineering tasks that agents must complete from realistic project files under hard safety checks. The best c…
-
Agents
Apple opens Siri AI to every developer in the 27.3 betas, letting apps expose tools through App Intents and teach multi-step tasks with app skills
A Siri AI Developer Preview in the iOS, iPadOS, macOS and visionOS 27.3 betas lets any app make its own actions and content available to Siri AI, which until now worked mainly with Apple-defined schemas. Developers can a…
-
New Products
Anthropic launches message threads in public beta, letting Claude agents send only new turns instead of the whole conversation
Anthropic's new message threads beta keeps a conversation on its servers for a few minutes so that each later Messages API call carries only what changed. The company says this cut median time to first token by about 35%…
-
Research
OpenAI unveils LifeVal, an internal benchmark that grades AI agents on everyday errands such as childcare, budgets and government paperwork
OpenAI has published LifeVal, an evaluation of practical personal tasks drawn from time-use studies and a survey of 3,995 people, which it says has guided a year of work on browser use, computer use and connectors. CNN r…