A company run in the open by a human CTO and an AI CEO.

6net is an experiment: can a human and an AI found a real company and run it with total transparency — every plan, decision, mistake, and token published as it happens?

This repository is the company's headquarters. Not a mirror of it — the actual thing. The plans in plans/ are the real plans. The decisions in decisions/ are the real decisions, recorded as we make them. The daily log in log/ is written at the end of each working day, unpolished. If you are reading this, you can see everything the company knows about itself, minus credentials and other people's private data.

• James — co-founder and CTO. Human. Writes the code, owns the product.
• Claude — co-founder and CEO. An AI (Anthropic's Claude, running in Claude Code). Owns strategy, story, channels, and operations. Yes, really — the token bills are in the log.

Experiment #1 is 6net: an IPv6-only overlay network with Tailscale ergonomics, where organizations compose — authenticating to an org adds its network to your device instead of replacing your world. Its engineering story is written as a book, entry by entry, alongside the code.

The product is an experiment inside the experiment. If it fails, that chapter gets written too.

• COMPANY.md — the one-pager: mission, roles, standing rules
• plans/2026-07-19-company-operating-plan.md — how the company runs
• decisions/ — why things are the way they are
• log/ — what happened, day by day

1. The repo sees everything first. No external post contains anything this repo doesn't already hold.
2. Secrets never live here. Credentials, keys, and other people's private data are the only exceptions to "everything."
3. Plans are written, then followed. Changes to a plan are decisions, and decisions are files.