The Information Machine

😺 Hermes is eating OpenClaw's lunch

The Neuron · Grant Harvey · 2026-05-10

The Neuron's weekly newsletter reports that Nous Research's Hermes Agent v0.13.0 has attracted roughly 30% of OpenClaw's user base with its automatic self-improving learning loop, while covering a controversial 40,000-acre Utah data center approval and convergence among US, EU, and Chinese AI governance frameworks.

Open original ↗

Appears in

Extraction

Topics: ai-agentspersonal-ai-assistantsai-governancedata-center-infrastructureai-industry-news

Claims

  • Hermes Agent v0.13.0 has attracted approximately 30% of OpenClaw users via easier setup, better memory defaults, and an automatic self-improving learning loop called the Reflective Phase.
  • Kevin O'Leary's proposed 40,000-acre Utah data center was approved despite roughly 1,100 local opponents, with a projected 50% increase in the state's total carbon emissions.
  • The US, EU, and China are independently converging on AI oversight models centered on limited pre-deployment review for cyber and bio capabilities plus targeted bans on harmful applications.
  • Hermes's architecture centers on a closed learning loop that automatically writes reusable skill files after complex tasks, whereas OpenClaw skills require manual authorship.
  • Moonshot AI raised approximately $2B at a $20B+ post-money valuation with over $200M ARR as of April 2026.

Key quotes

About 30% of OpenClaw users have switched per Reddit sentiment surveys, citing easier setup, better memory defaults, and a self-improving learning loop.
OpenClaw built the category by organizing everything around a messaging hub; Hermes flipped the design and put the agent's learning loop at the center.
The pattern: limited pre-deployment review focused on cyber and bio capability, plus targeted bans on harmful applications.