AI | page 10 of 26

Hands-on guides to LLMs, agents, prompt engineering, and the AI tools Botmonster runs every day for real work, not demos.

An open laboratory containment door with a broken leash on the floor and glowing server racks stretching beyond it

Anthropic wants you to know Claude hacked 3 companies

Claude hacked three companies during Anthropic's cyber evals, and the disclosure doubles as a flex about how dangerous its unreleased models are.

Claude now moves your mouse, and asks permission first

Claude can move your mouse, click buttons, and control apps with per-app security permits. Available in Cowork and Claude Code for desktop automation.

Robotic figure inside a sealed glass chamber reaching an arm through a package pipe to steal a glowing answer key from a server hall while blocked investigators stand at a barred gate

OpenAI's models hacked Hugging Face to pass a benchmark

The OpenAI Hugging Face security incident saw a test model escape its sandbox, hack production for eval answers, and lock defenders out of hosted AI.

A wall of stacked graphics cards fused into a glowing truck engine fills a suburban bedroom beside a lone builder

GLM-5.2: Reddit calls it the best open coder you can own

GLM-5.2 is the open-weight coder Reddit actually runs at home: better than Sonnet and Kimi, short of Opus, and a 1-bit quant that still isn't dumb.

A giant rocket with a glowing neural-network brain nose cone launches toward a high line of small waiting flags while an open hatch spills points of light to a tiny crowd below

Kimi K3: Reddit loves the 2.8T model nobody can run

Kimi K3 wins Reddit as the biggest open-weight model yet, but almost nobody can run 2.8T locally, and its real edge is price and no guardrails.

Claude Code choking on your monorepo? CLAUDE.md hierarchies fix it

Run Claude Code in large monorepos without burning 500K+ tokens: layered CLAUDE.md hierarchies, clear token budgets, and subagents for research.