MondAI Roundup - July 2026

MondAI Roundup has been presented by Dominik Lukeš since late 2024. It gives an overview of AI news from the previous month and connects individual announcements to the wider changes in models, tools and professional work. The full July roundup is available on AI News Roundup.

The July roundup was dominated by model announcements. Not just GPT-5.6 Sol and Claude Fable 5 but many others.

Coding agents were also continuing their march toward becoming general work environments. The new ChatGPT desktop app brought Chat, Work and Codex together, while the previous app could remain installed as ChatGPT Classic. Same trends were happening across the industry: Claude Tag added delegation from Slack, while Kimi Work coordinated parallel agents for research and knowledge tasks.

We also saw further progress in voice technology with Gemini Live Translate and GPT-Live. We also talked about new coding benchmarks and Anthropic's global-workspace research.

For more detail, read the written notes or open the presentation.

Key Stories

• Claude Fable 5 and Mythos 5 and GPT-5.6 Sol — two frontier releases with different strengths, prices and working styles.
• Grok 4.5 and SWE-1.7 focus on faster coding and agent work, while GLM-5.2 releases MIT-licensed weights.
• ChatGPT Work, Claude Tag and Kimi Work — agents become work environments for research, documents and team delegation.
• NotebookLM's research update — code execution, source discovery and downloadable documents, spreadsheets and presentations.
• Gemini 3.5 Live Translate and GPT-Live-1 — simultaneous interpretation, better interruption handling and delegation during live conversation.
• Gemini Omni Flash and Seedream 5.0 Pro — conversational video editing, selected-region changes and layer-level image control.
• OpenAI's SWE-Bench Pro audit and FrontierCode — coding benchmarks face broken tasks, saturation and questions about maintainable solutions.
• Claude Science and OpenAI's study of agent use — coordinated research agents and a 137-fold increase in weekly non-developer individual users between August 2025 and early June 2026.
• OpenAI Daybreak and Jalapeño — defensive cybersecurity workflows and OpenAI's first custom AI chip.
• Anthropic's global-workspace research — J-space interventions connect selected internal representations to later reasoning while leaving major interpretability limits unresolved.

Resources

• Written notes — expanded chapters based on the session transcript
• Presentation

Key Links

• Claude Fable 5 and Mythos 5
• GPT-5.6 Sol, Terra and Luna
• ChatGPT Work
• Claude Tag
• Kimi Work
• Gemini 3.5 Live Translate
• GPT-Live-1
• OpenAI's SWE-Bench Pro audit
• FrontierCode
• Claude Science
• OpenAI's study of agent use
• Anthropic's global-workspace research

Next Session

There is no MondAI Roundup in August. The next session is on 14 September 2026, 12:30–13:30 — register here.