It’s finally here… Get on the waitlist: https://claude.com/product/claude-duet Always on agent from Claude. This is their version of always on agents like Grok Bots, ChatGPT Dots and Meta’s Muse. I have been obsessing over Dots and Muse for the past few weeks. Let me know if you are interested:

Four tech giants want to give you an AI helper. They’ve even made them fuzzy and friendly. Apparently, the future of getting things done needs to be a little less intimidating. Welcome to Always On, a new series about always-on AI agents. So, what’s an always-on agent? Imagine asking a human or AI assistant: “Check my calendar and tell me what I need to prepare.” Then you add: “Do this every morning.” Once an AI agent has permission to see your calendar, it can make that a routine. You don’t have to open a fresh chat and ask the same question every day. A human assistant might occasionally need a reminder. With Dots, Bots, Muse or Duet, that’s what the routine is for. That’s the basic idea. An always-on agent is an AI helper that can keep working on a job, or return to it later, without you sitting there. It uses a computer or smart phone over the internet, so your laptop can be closed. It can remember useful details, work with apps you connect, and bring you results—or ask for help when it gets stuck. Depending on the agent and the access you give it, you might say: - “Give me a short briefing every morning before my first meeting.” - “Watch for a reply about that delivery and let me know.” - “Every Friday, pull together a list of what’s still unfinished.” - “Draft a follow-up to this customer so I can review it.” You choose the job and what the agent can access. It handles the steps it’s allowed to take and comes back to you for actions that need your approval. Think of it as a helper with a to-do list and permission to get started. For example, OpenAI calls its always-on agents Dots. Once you’ve connected your dot to your email and granted the necessary permissions, you can give it an instruction such as: “Every morning, check my inbox and summarize anything that needs a reply.” 🫡 Think of yourself as the general of a vast, very cute army of dots. You give the orders, explain when the work should happen, and decide what needs your approval. Your dot gets to work. No marching required.

You open a new AI conversation to work on something for your business. Before you can get to the work, you have to explain the business. Who you help. What you’re building. How you write. What you’ve already decided. Again and again and again... That repeated setup is the problem. This is what I wanted to address when I built my AI operating system, or AIOS. I wanted a place to keep the context that should carry forward. The rules I work by, the people I serve, the work underway, The work, the outcomes, and what I’ve learned. All in one convenient, highly functional single source of truth so I didn't have to hunt, peak, and constantly restart with base information each and every time. Then I could direct AI to the relevant information before asking it to help. The structure behind that system is TRACK: - Tenets: identity, business direction, and rules. - Reputation: positioning, voice, and what the audience tells us. - Actions: the work that moves things forward. - Commitments: agreed outcomes, who owns them, and what counts as complete. - Knowledge: lessons and reference material worth reusing. TRACK is a framework, a method that gives continuity a practical foundation to my AI OS. Now, important context has a place to live and a way to be found when the next piece of work starts. I still have to keep it current, give AI access to the right material, and check what it produces. Saving a note doesn’t guarantee an AI will read it or use it correctly. The aim is to spend less effort rebuilding the background so we can get to the work. In the next post, I’ll walk through TRACK and how its five parts work together.

You don’t always need a detailed prompt to discover what AI can do. Try it with the photos, blueprint, and screenshot of a poem attached to this post. 1. Open a new chat in a LLM that accepts images. 2. Upload the first image and type: “Describe what you see.” 3. Upload the next image in the same chat. Then the next. Read what it picks up: objects, colors, visible text, layout, and details you might have overlooked. Compare its description with the image—did it miss anything or make an assumption? 📸 Now upload your own photo or photos. Try a screenshot, a handwritten note, an item you want to sell, or something you’re curious about. 💬 Share what surprised you?

Level: Beginner Category: Working with AI OS Five terms that come up the moment you start building a home for your AI OS. Here’s each one, in plain English. 🪄 AI OS One organized place for your information that AI can read, so it already knows your work before you ask it for help. 🪄 Second Brain An organized collection of your notes, ideas, and files, kept outside your head so you can find and use them later. 🪄 Markdown A simple way to format plain text using a few symbols, so it’s easy for both people and AI to read. 🪄 Obsidian A free note-taking app that keeps your notes as plain files on your own computer. 🪄 Vault The folder on your computer where all your notes live. It’s the word the note app Obsidian uses for it. Want the full definition of each term, with real examples and a pro tips upgrade to Premium? ❤️ Learn with Your Whole Heart: Expanded lessons, special benefits and community recognition for $9 a month. And 30% of Premium revenue goes to our designated charities. 👉 Join AI Bits & Pieces Premium AI Bits & Pieces - helping people and businesses adopt AI with confidence.