DataAIHub
DataAIHubNews · Research · Tools · Learning

Stack Pulse

Know what changed, why it matters to your architecture, and what to validate next.

Curated review queue

Changes worth revisiting in your stack

A focused, source-linked review queue from DataAIHub's reviewed entity records. It is not a live or exhaustive news feed, an alert subscription, or a cross-vendor scorecard.

Agent systems: selected source-backed signals

Orchestration, tool-use, and provider changes that can alter an agent boundary.

Evaluation evidence · Methodology update

LiveBench 2.0 changes the evaluation scope

LiveBench 2.0 adds long-horizon Agentic, Design, Coding, Browser Use, Data Analysis, Research, and Reasoning suites. Do not compare its scores directly with prior LiveBench versions without matching conditions.

  1. MilestoneCursor

    Malicious Claude Code extension builds removed

    Cursor removed Anthropic.claude-code 2.1.312 and 2.1.313 for macOS and Windows from the Cursor VS Code Extension Marketplace. Those builds were listed from about 2:27am to 1:35pm PT and were not published by Anthropic. Cursor shut the marketplace, rotated keys, and brought it back. The notice says other versions and the official Claude Code distribution were not identified as affected. Uninstalling the extension does not clear every change those builds made; follow the remediation on the notice if you installed either build in that window.

    Read primary source
  2. PlatformClaude Opus

    Claude Models on Google Cloud (Anthropic managed)

    Claude Models on Google Cloud (Anthropic managed) is the Claude API operated by Anthropic and reached through a Google Cloud project. Google Cloud IAM controls access. Google Cloud Marketplace handles billing. Fast mode is available for Claude Opus 5.5 on this surface.

    Read primary source
  3. ProductClaude Opus

    Claude Duet research preview

    Claude Duet is a research preview for US Pro and Max subscribers on a waitlist. It runs on Claude Opus 5.5 in a cloud virtual machine with a browser, sends a morning brief, and keeps working after the app is closed. It asks before sending a message, making a payment, signing a document, or deleting anything. Team and Enterprise plans are not included. It uses the same plan limits, at no extra subscription price. It is not a new API model id.

    Read primary source
  4. ProductCursor

    Agent SDK

    Cursor ships the Agent SDK so teams can define an agent as a folder of code, test it with evals, and deploy it to Cursor or their own infrastructure. Initialize with npx @cursor/agent-sdk init.

    Read primary source
  5. ProductCursor

    Origin App Marketplace

    Origin’s app marketplace lists partner apps for deployments, CI, code review, security, monitoring, and project management, and it is open to public submissions. Apps authenticate with an app JWT and installation access tokens, receive webhooks, and can write check runs, pull request comments, and reviews. Cursor reviews a submission before listing it.

    Read primary source
  6. ProductCursor

    Remote Control for local agents

    The Cursor iOS app can view and message agents running on your local machine. Agents keep running locally; the phone is a remote view and input surface. On by default for all plans except Enterprise.

    Read primary source