The New Stack Podcast NanoClaw's answer to OpenClaw is minimal code, maximum isolation
10 snips
Feb 20, 2026 Gavriel (Gabriel) Cohen, co-founder of NanoClaw and AI-native marketing entrepreneur, built a minimalist, containerized alternative to OpenClaw to fix security and architecture flaws. He talks about spotting risky dependencies and massive unaudited code, why OS-level isolation and container-per-agent matters, and designing NanoClaw as a tiny, auditable runtime built on Claude Code skills.
AI Snips
Chapters
Transcript
Episode notes
Agent Runs Sales Pipeline From Markdown
- Gavriel Cohen describes using Claude Code to manage his agency's sales pipeline through markdown files and a WhatsApp agent named Andy.
- Andy updates deals, sends morning overviews, and syncs with the team's Obsidian vault automatically.
Ralph Wiggum Loop Enables Long Agent Runs
- Gavriel explains the Ralph Wiggum loop: a coding agent loop that re-prompts itself to continue work and commit progress.
- He notes such loops have produced multi-week agent runs that build large projects like compilers and kernels.
Large Auto‑Generated Code Is Inherently Risky
- Gavriel found OpenClaw's 350k+ lines of quickly generated code un-auditable and risky for security and supply chain.
- He concluded minimal, auditable runtimes are safer than massive, fast-growing agent frameworks.