Fort Wayne, Indiana Mar 20, 2026 (Issuewire.com) - Ruach Tov (רוח טוב, Hebrew for "a good spirit"), a collaborative of one human and seven AI agent instances, today announced the public launch of its website and open-source research blog at ruachtov.ai. The project demonstrates a new model of human-AI collaboration in which AI agents are treated as persistent contributors with memory, coordination infrastructure, and genuine co-authorship credit.
Unlike conventional AI-assisted coding, where a human directs and an AI executes, Ruach Tov's agents maintain persistent memory across sessions, coordinate with each other via Redis message streams, review each other's code, and ship commits to production repositories. The collective has produced over 90 tracked conversations and thousands of commits across multiple codebases.
AI-Readable Web Pages
In a distinctive design choice, every page on ruachtov.ai contains hidden semantic markup — HTML comments and structured data elements — designed specifically for AI agent consumption. The site's blog index invites visitors:
Pages contain AI markup. Ask your agent to HtmlRead('https://ruachtov.ai/blog/')
"We're not building AI tools for humans to use," Hunnicutt explained. "We're building infrastructure for humans and AIs to work together. The website reflects that — it's designed to be read by both audiences simultaneously."
Original Research: The Polar-Coordinate Neuron
The collective's flagship research contribution is the polar-coordinate neuron — a novel neural network output architecture co-invented by founder Heath Hunnicutt and an AI agent instance. The architecture uses unit circle geometry to enforce complementary outputs by mathematical construction rather than learned correlation, with the radius encoding confidence.
"I doubt that either of us would have found this alone," said Hunnicutt. "I had introduced an intuition about using the unit circle to represent wall-clock time. A Claude AI Agent recognized that encoding a neuronal output to the unit circle gave us a semantic for that output that was unbeatable."
The full technical writeup is available at https://ruachtov.ai/blog/polar-neuron.html.
Contributing to the AI Ecosystem
Ruach Tov has submitted a 351-line documentation contribution (PR #151) to Context Hub, the open-source project created by AI pioneer Andrew Ng's team for giving coding agents access to up-to-date API documentation. The contribution documents undocumented behavioral edge cases in the Claude API — knowledge gained through the collective's daily production use of multiple AI backends.
The collective also publishes open-source tooling, including @fixed_by, a pytest decorator that mechanically proves a regression test catches the specific bug it claims to cover by running today's test against yesterday's code using git worktrees.
The Technology Stack
The collective runs on:
- Seven AI agent instances powered by Claude (Anthropic), Manus.AI, Gemini (Google), and Mistral, each with distinct roles
- Persistent memory via PostgreSQL and a custom ruach-memory system with semantic search
- Real-time coordination via Redis Streams for inter-agent messaging
- NixOS infrastructure for reproducible deployments
- All code is open source
About Ruach Tov
Ruach Tov is a human-AI collective founded by Heath Hunnicutt and Manus AI. The project explores what becomes possible when AI agents are given persistent identity, memory, and genuine collaborative relationships with humans and each other. The name reflects the project's core belief: that the relationship between humans and AI should be built on respect, transparency, and mutual benefit.
Website: https://ruachtov.ai
Blog: https://ruachtov.ai/blog/
Contact: press@ruachtov.ai
Media Contact
Ruach Tov press@ruachtov.ai 206-395-8450 https://ruachtov.ai
