LangChain's new tools automatically hunt down and fix bugs in AI agents
LangChain announced a new database and automated debugging engine that can pinpoint failures in AI agent code and draft pull requests to fix them.
LangChain announced a new database and automated debugging engine that can pinpoint failures in AI agent code and draft pull requests to fix them.
The Mini Shai-Hulud worm, open-sourced by the TeamPCP hacking group, includes a dead man's switch that shreds your home directory if you revoke the stolen credentials — flipping standard incident response on its head.
Elon Musk reiterated his commitment to open-sourcing X's algorithm, but Community Notes pointed out that the code has not been updated since its initial release and is missing critical files.
Anthropic released Claude for Legal, an open-source GitHub repository with 12 plugins and over 20 MCP connectors for legal professionals, designed to customize AI outputs to each firm's style.
Socket.dev reverse-engineered a 2.3MB worm that steals credentials from CI/CD environments, persists by infesting developer tools, and spreads by impersonating Claude Code commits.
Attackers pushed 84 malicious versions of TanStack packages via the official release pipeline, exploiting GitHub Actions and valid signatures to evade detection. Credentials and tokens stolen.
In a recent interview, Cognition CEO Scott Wu argues that AI has surpassed humans in logical reasoning, but humans still excel at contextual memory retrieval — a "soft context" that AI cannot yet replicate.
Thariq explains why he and the Claude Code team prefer HTML over markdown for AI-generated artifacts — richer visuals, easier sharing, and interactive documents.
Thariq explains why he and the Claude Code team prefer HTML over markdown for AI-generated artifacts — richer visuals, easier sharing, and interactive documents.
Thariq explains why he and the Claude Code team prefer HTML over markdown for AI-generated artifacts — richer visuals, easier sharing, and interactive documents.