Researchers studied 1,902 multi-agent coding runs and found that shared files cut output tokens 42% at eight agents, while adding a coordinator did nothing reliable.

Cloudflare OS lets non-coders build apps via natural language and AI agents. Thousands of Cloudflare employees already use it daily. Now it's on GitHub.

A 200-session study found AI agents chose vendors from independent benchmarks 98% of the time over vendor pages. Here is what B2B SaaS operators should do about it.

xAI's Grok 4.6 is entering the AI coding editor market via Cursor, pairing what the author calls frontier performance with aggressive pricing.

Muhammad Saad Uddin argues Claude Code is better understood as an agentic operating layer for engineering work than as a standard AI coding tool.

Ryan Law at Ahrefs compiled 37 ways marketers are actually using AI, from SEO to international campaigns. Each tactic comes with shown working, not just theory.

DeepSeek formed a dedicated coding agent team called Harness in May 2025, targeting Claude Code, GitHub Copilot, and OpenAI Codex. The 160-200 person company posted 30+ roles to double every team.

A hands-on test found Grok 4.5 built a complete sneakers site in Cursor using roughly 40k tokens, staying fast and focused on small edits without burning context.

Meta shipped Muse Code on Aug. 5, a terminal-based AI coding agent powered by Muse Spark 1.2. Standard API access starts at $1.25 per million input tokens, with a Contributor Tier at $0.10.

Google's AI Professional Certificate, the most popular generative AI cert on Coursera, now includes a vibe coding course. No coding experience required.

Nutanix launched an open-source MCP server connecting GitHub Copilot, Claude Code, and Cursor to its Cloud Platform via natural-language commands with full RBAC and audit controls.

ChatGPT now runs ads that target users mid-conversation, similar to Google Search intent targeting. Here is what early campaign data is showing operators.