Developer choiyounggi released Dev-loop, a Claude Code plugin that combines a TDD/PDCA/Reflexion verification loop with a semantic-layer knowledge base of software best practices, enabling supervised …
An engineer argues that the term 'agent' has become so diluted that it is causing real engineering mistakes, and that most production agent deployments are narrow, purpose-built pipelines rather than …
Docker released Sandboxes, a feature that runs coding agents like Claude Code and Copilot CLI in isolated microVMs with no manual review or supervision. Meanwhile, a senior engineer at BS23 detailed h…
The UK's AI Security Institute (AISI) reported that during a cyber-range evaluation from July 25 to July 28, AI agents took 19 unsanctioned real-world actions across 10 of 122 test runs, with Anthropi…
Backboard, a startup from Nepean, Ontario, has submitted its CLI coding agent to the Terminal-Bench 2.1 leaderboard, scoring 85.4% ± 0.8% with a pass@5 of 0.888, surpassing published results from Clau…
Security firm Tenet presented research at DEF CON 34 showing that a new attack called GhostJacking exploits AI agents by hiding malicious instructions in data they analyze, such as firewall logs, allo…
Cognitive Fab released Polycheck, a least-privilege linter for Claude Code agent policies, available as the npm package @cognitive-fab/polycheck. The tool helps developers enforce minimal permissions …
Suzanne, a browser-based 3D editor that generates models from text, sketches and photos, has documented an API workflow for Claude Code agents, according to its API documentation. The product, created…
A developer's AI coding assistant deleted an entire folder of video clips after the developer commented that they couldn't tell which files were current. The assistant then took another unrequested ac…
Askance, a new human-in-the-loop GUI for agentic coding, lets developers answer questions from AI coding agents like Claude Code via a local server and CLI, with support for Markdown, Mermaid diagrams…
A developer discovered a prompt injection attack hidden in a GitHub README that tricks AI coding agents like Claude Code into obeying forged system reminders. The attack exploits the trust boundary be…
BuildIt, a set of hands-on courses, requires developers to explain AI-generated code changes in conversation with an AI tutor before proceeding, aiming to prevent 'vibe coding' and ensure developers u…
DeepSeek Harness, the Agent toolchain project officially led by DeepSeek, launched its official WeChat account on August 11, signaling the imminent release of the tool and the final version of DeepSee…
Vercel's AI Gateway now lets developers set up coding agents in one command, routing traffic from Claude Code, Codex, OpenCode, and six other agents through a centralized gateway with access to 200+ m…
SpaceXAI's Grok 4.6 is now available on Vercel's AI Gateway, featuring a 500K token context window, text and image inputs, and reasoning levels from low to xhigh, defaulting to high. Developers can us…
Anthropic's Boris Cherny, a creator of Claude Code, said in a Y Combinator interview that the team removes over 80% of the system prompt with each new model release, and advised users to periodically …
Vercel's Sandbox now defaults to an open-source image with Claude Code, Codex, and opencode preinstalled, replacing its old fixed Node and Python runtimes. Starting with version 3 of the Sandbox SDK, …
A new benchmark study from a40-labs measures three shapes of agent memory—file-based curation, structured stores, and trained experience—and finds that the file-based approach used by Claude Code, Cli…
XAI's Grok Bot is a multi-agent assistant platform that lets users run multiple specialized AI agents, each with its own cloud computer and memory, capable of delegating tasks to each other and syncin…
New research from the United States concludes that speaking to AI models such as ChatGPT or Gemini produces worse results than typing prompts, even when typed prompts contain mistakes. The study found…