OpenAI built always-on personal agents. We spent a year building the same idea without the lock-in.

Disclosure: I'm one of the authors at Muvon, octomind is ours. When OpenAI launched Dots on Sep 29 I was glad. Always-on personal agents that take a job and do it while you're elsewhere is the shape I've been building toward for a year. Their entry point is ChatGPT Pro at $100/month and it's not available in Europe. So we kept going on ours. The octos are personal agents. You hand one a plain sentence, once or every morning, and the result lands in your chat even with the laptop closed. Nothing to install, working in sixty seconds. Mine watch two competitor pricing pages and only ping me when something moved. The team runs a morning brief on agent news, three things that matter, one line each, with links. Zapier moves data between apps by rules. This is the part that needs judgment: read, compare, decide, write. Each one is yours to shape, a name, rules in plain words, the model you trust, its own schedule and chat. Your rules apply to all of them, an octo's own rules win where they differ. They show their cost and stop when your balance runs out, ask before sending or deleting anything, and never see your password. Everything it makes downloads in a click. Cancel and your octos and files stay for thirty days. Devs get a Linux machine with sudo and Docker, live URLs, and a REST API behind one key. We just shipped the thing that removes the switching cost. Bring an agent over from ChatGPT, Claude Code, Codex, OpenClaw or Hermes, memories, instructions and skills import in one step. Free tier, no card and no invite. The runtime is Apache 2.0, same agents run on your own machine with the CLI. Links in the comments. What's the job you'd hand an agent that has to run whether your machine is open or not?