Do you use a personal agent?

Ben's Bites is soliciting reader input on who uses agent products as a "personal agent" — organising things outside of work, handling email, or doing tasks on their computer. Answers are to be revealed in Thursday's post.

The newsletter is sponsored by Cloudera, which cites its survey, The Great AI Re-Architecture, stating that 95% of organizations have delayed AI projects due to poor governance and compliance issues.

In headlines, Grok Bot is described as taking over much of the ex-OpenClaw crowd, with a social feed for Grok Bots that is not easily readable by a human. Hermes Desktop, described as another work-with-agent app, launched a Bot mode to mimic the UX of Grok mode.

A new Codex/ChatGPT feature turns desktop activity across apps and websites into memories and a timeline for reference when a user asks it to do something. It is opt-in via settings. Other Codex items include how to enable a 1M-token context window for Sol, and ultrafast GPT-5.6-Sol in the API for select customers.

In Claude Code, a /design skill creates editable UI artboards where users can pick a direction and tweak it visually, after which Claude Code can implement it, plus an auto-continue option for when the timer resets.

Google released Gemini 3.7 Flash three weeks after 3.6 Flash, with benchmark improvements described as better than GPT-5.6 Terra and Sonnet 5. Flash 3.7 and 3.6 are 50% off until the end of the year.

The roundup also notes Google Sheets Canvas, a Mac tool for scheduling tasks and headless agents, GitHub being down again yesterday, Cursor launching Origin as a competitor, Theo Browne's breakdown of his AGENTS.md and SKILLS.md files, and Factory opening Guild, described as a highly selective builder program. It adds that the ElevenLabs MCP is now available in Claude, and that Stripe has finalised an agreement to acquire OpenRouter for more than $7B.