🧠 Claude Can Now Drive Your Desktop

Google shipped Gemini 3.8 Flash. Google's new Gemini 3.8 Flash is their best reasoning and coding model yet—at the same speed and cost as 3.7. On complex tasks, it works harder: executing extra reasoning steps and calling tools iteratively to maximize performance.

Google's own teams are already using Gemini 3.8 Flash Cyber to patch vulnerabilities faster than ever. It's available now to Google AI Pro and Ultra subscribers in the Gemini app, Google Search, and Sheets.

Your invoice didn’t notice. You will.

Claude can now use your computer in the background inside Cowork and Claude Code. With a permission grant, it clicks, types, and opens apps while you keep working somewhere else.

Hand it a multi-step desktop chore: fill a form, open the right apps, knock out the clicks; then stay in your own flow. Watch the permission prompts, you’re still the one who says yes.

This is the background path Anthropic announced, not a silent takeover. Grant access when you’re ready, and let it work behind you.

Finally, a coworker that doesn’t need the chair.

ClaudeDevs posted an open-source blueprint for shopping and merchant agents aimed at retail, travel, telecom, and entertainment. It’s a developer pack for building those flows, not a click-and-buy founder tool.

If you’re engineering a purchase or merchant agent, start from the blueprint and wire it to your stack. Founders looking for a one-click shopper should wait for a productized layer.

No public GitHub link was in the announcement tweet, so follow the thread for where the code lands.

Your cart can wait.

awesome-design-md is a free MIT collection of brand DESIGN.md files you drop into a coding agent so the UI matches a known design language. Copy one pack into the repo and ask the agent to build a page that looks like that brand. The packs themselves are free; the GitHub page has sponsor noise around the edges, ignore that and grab the markdown.

This Five Whys prompt is a pasteable Toyota-style root-cause template for when an AI answer feels off. Drop it into your chat after a shaky reply and force the model to dig past the first explanation. It’s one prompt tip, not a product launch, so steal the wording and keep it in your notes.

How to Catch AI When It’s Confidently Wrong is a Free Playbook on six signs that an answer needs verification — specific numbers with no source, invented certainty, weak reasoning, and more. Open it, pick one sign you keep missing, and run that check on the next AI draft before you ship.