New on Claude Sonnet 5's pricing is now permanent, Electric joins Databricks for Neon's agent data layer, and how to run parallel coding agents with Orca's Git worktrees.

New on Claude Code v2.1.227 ships a CI/Bash fix for GitHub-hosted runners and a /tui rewind fix, and Anthropic begins watermarking AI-generated text and images across Claude products.

New on Meta open-sources Muse Glimmer 30B under Apache 2.0, the UAE launches a national agentic AI project, and Zuckerberg makes the case for personal agents over superintelligence.

New on SynapsCLI is a Rust-based multi-agent terminal runtime delivering a 20ms cold start for parallel agent dispatch.

New on Claude Code v2.1.225 adds gateway spend limits and workspace trust prompts, Gradient open-sources Symphony-Coord for multi-agent routing, plus a LangSmith deep agent deploy how-to.

New on Prime Intellect open-sources Prime Agent, a self-improving RLM coding harness, plus how-to guides on spotting AI-generated pull requests and designing safer agent command approval workflows.

New on Sierra launches AI voice personas for brand-specific agents, Claude Opus 5 tops the BenchLM agentic leaderboard at 80.7, and Claude Code's Auto Mode becomes default August 14.

New on Agent Plugins 1.0 launches — Vercel, Amazon, Cursor, Microsoft, OpenAI, and Google back a portable skills standard. Claude Code ships v2.1.224 and v2.1.225 with self-hosted runners.

New on Goldman Sachs details its playbook for running AI coding agents alongside 12,000 engineers, and xAI ships Grok Build 1.0, the terminal agent’s first stable release.

New on Fable 5's biology classifier update opens healthcare pipelines, Claude Enterprise adds enterprise skill and plugin controls, and CrewAI 1.15.12 ships URLReadTool plus IBM Db2 search.