Nous Research Co-Founder on Building the Self-Improving Hermes Agent
petergyang · x · 2026-08-02
This podcast features Karan Malhotra, co-founder of Nous Research, diving deep into the philosophy and practice behind building Hermes, the #1 personal agent on OpenRouter.
Key takeaways include:
- Differentiation: How Hermes differs from tools like Claude Code and Codex, and strategies to prevent AI from generating generic, monotonous slop.
- Skill Acquisition: The mechanism for creating and learning new skills without bloating the model's context window.
- Open Source vs. Safety: The debate on how open-source models impact AI safety compared to closed models.
- Workflow Integration: Using a unified Kanban board to manage workflows for both humans and AI agents.
- Live Demo: Rebuilding a childhood classic Sonic the Hedgehog world using Hermes.
Related event: Nous Research Co-founder on Open Source AI and Building Hermes(3 posts)→
More from coding & agent
- From Autocomplete to Autonomous Agents: The Evolution of AI Coding Paradigms — dejavucoder · 2026-08-03
- 3D Engine Optimization: Zero-Parsing Binary Format Loads 3,777 Assets in 1.4s — gdechichi · 2026-08-03
- Save Tokens with code-review-graph: Smart Codebase Mapping for Claude Code — dr_cintas · 2026-08-03
- Databricks Launches Industry's First Context Engineer Certification — Zachly · 2026-08-03
- Securing AI Agents: Treat Them Like Interns, Not Threats — alex_teichman · 2026-08-03
- Comparing AI-Assisted Coding to Car Gears: From Autocomplete to Agents — dejavucoder · 2026-08-03