MondAI Roundup — May 2026

MondAI Roundup has been presented by Dominik Lukeš since late 2024. It gives an overview of AI news from the previous month and links them to key themes of interest. It runs every second Monday of every month at 12:30–13:30. A regular summary of key lessons and a more detailed write-up live on AI News Roundup.

May was one of those months that felt overloaded because the news kept arriving in every category at once. Anthropic's Claude Opus 4.7 (April 16) and OpenAI's GPT-5.5 (April 23) shipped a week apart and now sit at or near the top of the benchmark tables, but with very different reception. GPT-Rosalind, a frontier reasoning model fine-tuned for biology and drug discovery, shipped alongside. Underneath them, the open-weights tier kept moving: Qwen 3.6, Kimi K2.6, DeepSeek V4, and the oddly quiet Grok 4.3 API launch.

Voice and image became more useful. Gemini 3.1 Flash TTS introduced audio tags as in-script direction; GPT-Realtime-2 brought full flagship intelligence into the voice stack with passive listening and live translation; GPT-Image-2 took the #1 Image Arena slot by 242 ELO points, the largest single-month jump the benchmark has recorded. Anthropic also released Claude Design, a code-based design feature inside Claude.

Agent platforms appeared across several labs at once. Anthropic's Managed Agents added memory and a multi-agent shared-memory feature called "dreaming." OpenAI previewed Workspace Agents. Microsoft launched Foundry Hosted Agents. Mistral shipped Workflows. Cursor opened a Cloud Agents API. Codex itself was upgraded into something closer to a general work environment with browser previews, sidebar artifact rendering, and image generation built in.

Anthropic signed a compute partnership with SpaceX for the Colossus 1 data centre: 300 megawatts, more than 220,000 GPUs coming online within the month, and Claude Code rate limits doubled the same day. Microsoft and OpenAI amended their five-year partnership to be non-exclusive; OpenAI models are now available on Amazon Bedrock for the first time. The Mythos Preview / Project Glasswing arc continued, OpenAI explained how reinforcement-learning personality training produced a persistent gremlins-in-code tic across recent models, and Anthropic published more work on hidden model behaviour. The labour section covers Imas, Matthews, Pompliano, and Mollick on jobs, risk, and professions.

For more details, read the detailed notes or browse the presentation slides.

Key Stories

• Claude Opus 4.7 and GPT-5.5 — the two top-tier model releases, with different reception and different strengths.
• GPT-Rosalind — OpenAI's life-sciences model, now looking for academic testing partners.
• Qwen 3.6, Kimi K2.6, and DeepSeek V4 — the open-weights tier keeps shipping at speed.
• Gemini 3.1 Flash TTS and GPT-Realtime-2 — voice models get cheaper, smarter, and less awkward.
• GPT-Image-2 and Claude Design — image generation and code-as-design become more useful for real design work.
• Codex for almost everything — Codex becomes a general work environment with browser previews, sidebar artifacts, and image generation.
• Anthropic Managed Agents, OpenAI Workspace Agents, and Microsoft Foundry Hosted Agents — agent platforms become standard lab infrastructure.
• Microsoft Copilot Agent Mode and Claude across Microsoft 365 — Microsoft 365 added more agent and model-choice features.
• Anthropic-SpaceX compute partnership and Microsoft-OpenAI partnership amended — compute and cloud access remain the hard industrial constraint.
• Alex Imas, Anthony Pompliano, Dylan Matthews, and Ethan Mollick — jobs, AI risk, and professions.

Resources

• Detailed notes — expanded chapters covering all major stories
• Presentation slides
• Quick reference index — searchable list of people, organisations, and links

Key Links

• Claude Opus 4.7
• GPT-5.5
• GPT-Rosalind
• Qwen 3.6 family
• Codex for almost everything
• Anthropic Managed Agents memory + dreaming
• Microsoft Copilot Agent Mode
• GPT-Image-2
• Anthropic-SpaceX compute partnership
• Alex Imas — What Will Be Scarce?

Next Session

Monday 8 June 2026, 12:30 BST — register here. Same time, second Monday of the month, same Teams.