Summary
It was a busy week in AI, and Paul J Lipsky’s weekly recap covers the biggest stories, starting with Anthropic and OpenAI releasing new models on the same day: Claude Opus 5.5 and GPT-6 Sol.
Lipsky tests the two side by side at high effort. In a YouTube script-writing task with built-in web research, GPT-6 Sol finishes slightly faster, but Opus 5.5 produces the clearer and more human-sounding result and impressed him with how quickly it navigates the web. He also notes that GPT-6 Sol is priced at roughly half the cost per API token, and that both tasks used only a small share of his $100-a-month plan limits. In a motion graphics test he again favors Opus 5.5, calling it the best model he has used for writing and, overall, better than Fable 5.1 once speed and price are considered.
The recap also covers Gemini’s NotebookLM updates, including live voice chat with notebooks for Ultra subscribers, notebooks as sources inside Google Docs, and a new report type. Other stories include a new Grok model and voice calls with Grok bots, and announcements from Meta’s Connect event around Meta Muse.
It’s a quick way to catch up on the week’s AI news and see how the latest frontier models compare in practice.
📺 Source: Paul J Lipsky · Published September 25, 2026
🏷️ Format: Roundup
