OpenClaw 2.0

41 comments (HackerNews + Product Hunt; top 25 shown) reveal a deeply split user base. On the plus side, real workflows are reported: one HN user's claw 'scrapes some APIs, receives emails with job listings, and matches it all to my profile' and converts inbound emails to PDFs for bookkeeping software; another uses agent harnesses to 'manage most of my homelab (scoped Proxmox tokens and Portainer)'; another praises the organizational structure and MCP exposure ('audit this', pull thread context, search). Product Hunt voices love the self-hosted, messaging-native model ('Local & Private', 'texts you back on WhatsApp, Telegram, Signal', 'Results as a Service', 'hackable... a platform waiting for crazy ideas'). On the minus side: (1) Security is the dominant HN theme — 'Almost 600 [CVEs] in less than a year, roughly two per day', 'open door to a remote privilege escalation potentially granting root access... to any text your model ingests', and a Product Hunt commenter saw 'many people started complaining about attacks and hack attempts'; one expert-sounding comment notes 'chat ingress gets probed as soon as it is public' and recommends OPA allowlists, gVisor sandboxing, and audit logs. (2) Churn: 'It quickly turned into a lethargic mess which I really didn't see myself wanting to maintain long term' — that user moved to picoclaw (<10MB); others mention nanoclaw, the Pi harness, Hermes, or piclaw. (3) Redundancy for pros: 'if you already use a coding agent then openclaw adds very little... you exchange a low quality messaging interface for that ease of access.' (4) The 'it's just scripts/cron' critique is countered by the Dropbox-vs-SFTP argument: it 'enables more people to do the thing that was limited to fewer people.' (5) Cost/privacy friction: token costs ('feeding the lobster') scare non-developers, and background autonomy forces cheap models plus discomfort with sending personal data to 'a random inference provider.'

No aggregate ratings were found for this product during the last harvest.