Openclaw

I already run OpenClaw on my server. It handles the useful day-to-day layer: Telegram conversations, scheduled jobs, notifications, research, summaries, and the accumulating collection of small automations that make a personal assistant worth having. So at first glance, Claude Code Remote Control looked redundant. Why put another AI interface on the same box when I can already talk to OpenClaw from my phone? Because they solve different problems. OpenClaw is my normal control plane. But when an OpenClaw update breaks that control plane, I need an independent way back in. In the past that meant SSH from a laptop with the right keys. Fine at a desk; useless when the laptop is shut, elsewhere, or simply not with me. ...

I had already written one postmortem about an OpenClaw update going badly. Then I did the thing that postmortems are supposed to prevent: I took another beta update. This time the failure was unusually clean. The update installed exactly what the package registry advertised. The gateway did exactly what its safety checks told it to do. And the combination could not possibly converge. The result was a gateway that crash-looped forever, a doctor command that appeared to hang, and a local TUI frozen on “connecting”. Recovery did not involve restoring a backup or deleting data. It involved understanding why the release was impossible, refusing to downgrade past a database migration, and applying one temporary local patch. ...

Most notification systems do not have an information problem. They have a presentation problem. The email system had already done the hard work. In an earlier post, I described the scoring pipeline that separates mail needing an immediate alert from messages that can wait for the evening summary or a later review. More recently, I wrote about removing recurring junk at the source, rather than merely teaching the filter to tolerate it. ...

This was the worst OpenClaw update I have had. Not because one command failed. That would have been easy. It was worse: Telegram disappeared immediately after the update, openclaw doctor could not repair it, the local TUI was broken too, the first workaround degraded the system, and the beta recovery path corrupted enough state that I had to bring in a standalone Claude session from outside OpenClaw to patch SQLite manually. ...

Email junk does not arrive as one dramatic flood. It accumulates quietly. A shop you bought from once starts sending three promotions a week. A travel site mistakes one booking search for lifelong intent. A restaurant app decides you need personalised recommendations forever. None of these messages are individually offensive. That is why they survive. But the result is stupid: an inbox full of tiny claims on attention. So I asked Closp to audit the last month of Gmail and find the recurring junk worth unsubscribing from. ...

My garden CCTV system started with a simple loop: Frigate detects movement, a Python script sends a Telegram notification, and I can look at a snapshot or short clip from my phone. That was good enough for real-time alerts. It was not good enough for history. Telegram is a terrible archive browser. It is excellent at interrupting me when something happens, but bad at answering questions like: What happened in the last two hours? Did I miss a person detection yesterday? Which clips exist for this camera? Can I open the full event without digging through chat history? So the next small project was obvious: keep Telegram as the notification surface, but build a proper CCTV history portal on the Raspberry Pi itself. ...

At the end of March, my OpenClaw setup hit an awkward wall. For a while, Anthropic OAuth access had been good enough for day-to-day agent work. Then access tightened. I do not mean “the model got slightly slower” or “one request had to retry”. I mean the system became unreliable enough that I was pushed toward the API path. That sounded fine in theory. In practice it turned the assistant from a predictable tool into a metered engineering problem. ...

Most AI assistants are designed to be agreeable. They say “Absolutely.” They praise the question. They hedge every answer until nothing is left but damp corporate fog. This is fine if the assistant is a search box with manners. It is not fine if the assistant has access to your shell, email, calendars, automations, Raspberry Pi, finance reports, blog drafts, and the long tail of unfinished things that make up a real life. ...

I run a Raspberry Pi 5 at home as part of my OpenClaw setup. It started as an OpenClaw node connected through an SSH tunnel, then became a garden CCTV box running Frigate and recording video clips. By this point it is a camera box, a Frigate server, a Hailo detector host, and a remote machine my AI assistant can inspect from Telegram. At least, that was the idea. After a recent OpenClaw update, the Pi still looked connected. It was visible in the gateway. It was paired. It had broad permissions. The UI suggested everything should work. ...

I recently got a Raspberry Pi 5 (8GB) and decided to connect it to my OpenClaw setup as a remote node. The idea: give my AI assistant direct access to a physical machine sitting on my home network, so it can run commands there, monitor things, and eventually interact with hardware. This post documents how I did it, the pitfalls I ran into, and what I plan to use it for. ...