My Adventures With ‘The AI That Actually Does Things’

▤ Summary

You get a first-person test of OpenClaw, the agentic AI software that powered the Moltbook phenomenon, and how it lets LLMs act on users' devices. The piece walks you through the awkward, command-line install, model-switching hassles, and the security warnings that accompany giving an agent broad system and account access. You see a mix of practical but fragile uses—email digests, simple automation, and home-device control—alongside the difficulty many users have finding durable, "sticky" workflows. The article flags privacy, surveillance, and reliability risks while noting that agentic features will likely be absorbed and improved by larger AI platforms.

▦ Linked Companies