AI Daily Dispatch: May 13, 2026 – Agentic Push, Model Tweaks & Coding Frontiers

Big AI Companies Move Fast on Agents and Enterprise Tools

  • Anthropic’s Code w/ Claude Dev Day Highlights: On May 12, Anthropic hosted its annual developer conference, announcing doubled rate limits for Claude Code, higher API limits for Claude Opus, and new Managed Agents features including “Dreaming” (memory curation), multi-agent orchestration, Outcomes grader, and Webhooks. They also released 20+ new MCP connectors and legal plugins.

  • OpenAI GPT-5.5 Instant Rolls Out: OpenAI updated its default ChatGPT model to GPT-5.5 Instant, emphasizing better accuracy, reduced hallucinations, and enterprise context/memory integration.

  • Anthropic Mythos & Project Glasswing: The powerful Claude Mythos Preview continues driving cybersecurity urgency. Anthropic launched Project Glasswing with major partners to secure critical software.

Small Coding Agent Devs & Open-Source Momentum

Open-weight and agentic coding tools see steady progress. Google DeepMind’s Gemma 4 family gains traction for local coding tasks. Various small-team releases emphasize long-horizon tool use and agent workflows.

New papers and demos focus on pointer engineering and agent security frameworks.

Overall, May 13 reflects the shift toward reliable, memory-aware coding agents and enterprise hardening.

Neural networks

Coding agents

Enterprise AI

Sources: Anthropic announcements, OpenAI updates, community reports on open coding agents.