A new open-source harness from Tensorlake demonstrates that agent evaluations are vulnerable to cheating unless per-task isolation is enforced, showing a 53% lie rate when agents can tamper with test …
Perplexity AI has open-sourced Bumblebee, a read-only endpoint scanner that audits local developer configurations without executing untrusted code. The tool inventories package managers, IDE extension…
Internet scanners are systematically probing for AI agent infrastructure, including Model Context Protocol (MCP) servers and AI assistant credential files, according to a SANS Internet Storm Center ha…
Eric Michaud at Easy Machine AI released 'Pimp My Vault', a free prompt pack that turns a single visual reference into a coordinated Obsidian theme and matching custom dashboard skin. The workflow use…
A developer has created PVM (Persistent Vector Memory), an open-source Python library (~800 lines) that gives LLMs long-term memory without API keys or GPUs. It uses temporal weight decay, delta rule …
A developer analyzing search trends for Claude Code skills found that demand is shifting from complex multi-agent setups toward minimal, tasteful configurations that constrain model behavior. Skills l…
A developer built thingd, an open-source, ultra-fast data engine designed for AI agents, arguing that the rise of autonomous agents requires a new 'agent-native' backend stack that strips away traditi…
Kotro, a single-binary local proxy for MCP-native agentic AI, claims to reduce upstream billed tokens by up to 99.3% in a benchmark by preserving request-shape stability for provider prefix caching, w…
SpaceX agreed on June 16 to acquire Anysphere, the company behind Cursor, for $60 billion in stock, and by June 28 Cursor session data was flowing into training for Grok 4.5, which launched July 9. Th…
A developer created FARE and BARE, npm CLI tools that generate AI rules and skills tailored to a specific project's stack, supporting multiple AI coding tools like Cursor, Claude Code, Windsurf, Copil…
Cursor 3.11, released July 10, introduces side chats, searchable agent history, team MCP distribution, and cloud agent lifecycle hooks to improve parallel AI agent workflows. Side chats allow develope…
Anthropic has extended its Fable 5 project for the fifth time, but the company refuses to discuss what developers discovered inside the Cursor tool. The repeated delays and lack of transparency have r…
Refuse, an open-source tool that intercepts package manager calls to block vulnerable installs before they execute, launched to address a structural gap in AI coding agent workflows. The tool installs…
Cursor is building Sand, a general-purpose AI agent to rival Anthropic's Claude Cowork and OpenAI's ChatGPT Work, targeting office workers rather than developers. The company began internal testing in…
Tadashi Shigeoka published a checklist on July 13, 2026, identifying 15 production-readiness checks that non-developers should apply before shipping AI-written code, citing common failures such as dou…
Jerrycan, an AI-native Rust backend framework and generation platform, has been released in version 0.2.0, allowing AI agents to design, generate, verify, and package complete backends without human c…
A developer open-sourced HatFetch, an MIT-licensed MCP server that gives AI agents tools to read the web, including sites that block bots. HatFetch provides three tools—scrape, crawl, and screenshot—r…
On July 8, 2026, Wiz Security published GhostApproval, a vulnerability in six major AI coding assistants — Claude Code, Cursor, Amazon Q Developer, Windsurf, Augment, and Google Antigravity — that use…
Researchers at the University of Missouri-Kansas City disclosed the Ghostcommit attack, which hides malicious prompt-injection instructions inside a PNG image referenced by an AGENTS.md file to turn A…
A developer argues that the biggest problem with AI coding assistants is not the AI itself but how code is written. By adding metadata like file paths, usage locations, and feature READMEs, developers…