The problem, the solution and how to try it yourselfhttps://mindwrapper.nethttps://mindwrapper.net/en/posts/your-ai-conversations-as-memory-and-contexthttps://mindwrapper.net/en/posts/your-ai-conversations-as-memory-and-contextMichael works with many AI tools at once: Claude Code, DeepSeek Harness, OpenCode, Codex, Hermes (a personal assistant), and Gemini and Claude in the browser. Each keeps its conversations in its own place, in its own format. Something worked out a month ago is hard to find again.Wed, 30 Sep 2026 00:00:00 GMThttps://mindwrapper.net/en/posts/build-your-own-audio-travel-guidehttps://mindwrapper.net/en/posts/build-your-own-audio-travel-guideMichael planned a business trip to Málaga with the Claude and ChatGPT apps. The apps were a deliberate choice: voice mode works perfectly when you're driving or walking through a city centre. It went surprisingly well, but a chat like that couldn't do four things. It didn't know who was travelling,…Mon, 28 Sep 2026 00:00:00 GMThttps://mindwrapper.net/en/posts/from-20-to-2-percent-wronghttps://mindwrapper.net/en/posts/from-20-to-2-percent-wrongMichael wears a Pebble Index 01: a ring you speak a short note into. The first word decides where it goes: "taak" (task) becomes a task, "gedachte" (thought) a journal note, "boodschap" (shopping) a line on the shopping list. But speech recognition doesn't always hear right. "Taak" becomes "paak",…Mon, 28 Sep 2026 00:00:00 GMThttps://mindwrapper.net/en/posts/prompt-profile-let-the-ai-answer-at-your-levelhttps://mindwrapper.net/en/posts/prompt-profile-let-the-ai-answer-at-your-levelI asked Claude to review my prompts, based on my past conversations. One tip: state your knowledge level of the topic in your question. That gets you an answer at your level instead of a generic one.Sat, 25 Apr 2026 00:00:00 GMT