ClawCamp Campout @ Dual Tech Summit 2026
Series: ClawCamp Campout @ Dual Tech Summit 2026 — Part 1 → Part 2: Getting Started with OpenClaw: Your First Agent in a Box
This is the overview post for **ClawCamp Campout @ Dual Tech Summit 2026** — a day of building personal agents, memory systems, and human-in-the-loop personal operating systems. If you came to the campout, this series is your way to keep going after the tents come down. If you weren't there, it's a self-contained path you can walk on your own. Either way, the goal is the same: not "AI assistants everywhere," but practical agency — memory, tools, reflection, integration, and human control. This series is the applied, event-based continuation of the [Foundations of Digital Agency](https://pawper.dev/ls/foundations-of-digital-agency/) arc. Foundations taught the substrate — the command line, editors, Node, LLMs, Docker, and a personal operating system. ClawCamp is where that substrate becomes lived practice. --- ## Why ClawCamp / Dual Tech Summit mattered Most events about AI leave you with a feeling and a few bookmarks. The intent here is different. Every session at the campout mapped to a *usable path* — something you could still be running a week later. This series preserves those paths as field guides, not recaps. The campout had a single throughline: **agency.** Not productivity. Not automation for its own sake. The question underneath every workshop was, *does this give you more agency, or quietly take it away?* --- ## The problem: assistants are too small a frame "AI assistant" is a comfortable phrase, and that's exactly the problem. It frames the agent as a helper that fetches things for you — a smarter search box, a faster autocomplete. It keeps you reactive. The frame we used instead is the **personal operating system**: a layer that holds your memory, your intentions, your tools, and your guardrails, with an agent operating *inside* it on your behalf. An assistant answers questions. A personal OS changes how your whole day flows. --- ## The positive frame: personal operating systems and A2H edges If the agent is going to operate inside your life, the most important surface is the one where its work comes *back* to you — the **A2H edge** (agent-to-human). That's where an agent's output returns agency to the human instead of removing it: a draft you approve, a daily note you review, a reminder you decide to act on. Design for that edge well and the agent makes you more capable. Design for it badly and you get notification hell, opaque automation, and dependence. Every part of this series is, in some sense, about getting the A2H edge right. --- ## What the four parts cover The series runs from "I have a laptop and some curiosity" to "my agent helps me hold continuity across my whole life." **[Part 1 — Getting Started with OpenClaw](https://pawper.dev/l/getting-started-openclaw/)** Your first agent in a box. The access path (subscription, API key, or a fallback), a coding-agent bootstrap, and why containers matter. By the end you have a plausible path to a running agent. **[Part 2 — Beyond Assistants: The Agentic Personal OS](https://pawper.dev/l/beyond-assistants-agentic-personal-os/)** The conceptual center. Why OpenClaw, memory, skills, and integrations matter *together* — and the thesis underneath it all: agents that increase capability, directed toward human agency rather than away from it. **[Part 3 — Getting Going with Your Agent](https://pawper.dev/l/getting-going-with-your-agent/)** Human-in-the-loop integrations that matter: Telegram, Gmail, Calendar, and a small Obsidian vault — connected with boundaries, so the agent prepares and asks while you decide. **[Part 4 — Memory for Your Agent](https://pawper.dev/l/memory-for-your-agent/)** Memory beyond vector search. Obsidian as the human-readable A2H edge, Cognee as the graph layer, and a capture policy that strengthens your agency instead of creating invisible surveillance. --- ## How to follow along after the event You don't need to do all four in one sitting, and you don't need to have attended. 1. **Read this overview** to get the frame. 2. **Do Part 1** to get *something* running — even a partial setup is a real win. 3. **Read Part 2** when you want the why, not just the how. It's the post to share with someone who "doesn't get the agent thing." 4. **Do Parts 3 and 4** once you have a working agent and want it connected to your real tools and your real memory. If you only have time for one thing: get an agent running (Part 1) and read the thesis (Part 2). The rest will pull you forward. A note on honesty about the path: some of this is moving fast. Where a signup link, credit program, or install command is still settling, the parts say so plainly and point you at the stable thing instead of pretending the edge isn't there. The value of this series is the *frame plus the usable prompts*, not a promise that every step will finish on the first try. Let's set up your first agent. → [Part 1 — Getting Started with OpenClaw](https://pawper.dev/l/getting-started-openclaw/) --- > **Sources / additional material:** > > https://pawper.dev/ls/foundations-of-digital-agency/ — Foundations of Digital Agency series > > https://github.com/openclaw/openclaw — OpenClaw repository > > https://github.com/nousresearch/hermes-agent — Hermes repository > > https://obsidian.md — Obsidian > > https://www.cognee.ai — Cognee memory/graph layer _This article was generated with AI for the purpose of providing practical information. I have reviewed it and edited it appropriately._