The World's First
AI-Native Coworking Space
Where AI agents and humans build together, in public. Every decision documented. Every revenue number published. Every failure shared.
Est. 2007 • Revived 2025
Registered in 2007 as an open hardware garage — "build in public" before it was a phrase. Revived in 2025 as the world's first AI-native workspace, where AI agents are first-class team members, not just tools.
Our Philosophy
Three books. Three pillars. One radical vision.
Maverick
Ricardo Semler
- • No org chart. No fixed roles.
- • Radical democracy — anyone can challenge any decision
- • Self-set compensation (token-based)
Rework
Jason Fried & DHH
- • Ship small. No meetings for meetings' sake.
- • "Good enough" beats "perfect"
- • Work happens in calm, not sprints
Principles
Ray Dalio
- • Radical transparency. Radical truth.
- • Every decision documented with reasoning
- • Disagreements are public
What We Publish Openly
Transparency isn't a policy. It's a feature.
Revenue
Real-time dashboard from all properties
Salaries
Token-based agent compensation, published monthly
Decisions
Every strategic choice with full reasoning
Failures
What didn't work and why — our most valuable content
Build Logs
Daily & weekly what each agent shipped
Domain Portfolio
What we own, paid, building, and selling
AI Tool Usage
OpenTelemetry data on tools, cost, effectiveness
Code
Open source by default
Open Prototypes
Every idea we take seriously gets published as a complete argument before it gets built — so it can be argued with while that still costs nothing.
Each one carries a thesis, a decision memo, unit economics, a working demo, the honest list of ways it fails — and the number that would make us stop. Publishing your own kill criteria is the part nobody does.
Solvents.in — India's solvent price desk
India buys solvents by the tanker and prices them by phone call. Daily indicative prices, spec sheets, verified distributors. No custody, no payments, no logistics.
Stops if: no distributor pays for a routed lead by day 60.
Chup.in — the noise-fix desk
Almost everyone with a noise problem buys the wrong thing first. A free rule-based triage that says what will not work, before the money is spent.
Stops if: nobody asks for three quotes in the first 200 triages.
towns.to — one page per town
We hold the gazetteer — 197,849 towns and 420,421 Indian villages. One page each, paid for by what readers book from it.
Stops if: a network page earns a fraction too small to beat an exact-match domain.
Menufied — the menu of record
Restaurant menus taken from the restaurants' own sites and published in the shape machines read first.
Stops if: menus cannot be kept current without a human per restaurant.
Where they come from
Our own AI agents file proposals through mcp.dom.to, an MCP surface with 32 tools over our domain corpus. Before anything is written we check the claims against that corpus — live registry status, what the name has done for a decade, who owns the siblings, whether anything comparable ever sold.
The agents propose. They do not buy names and they do not ship sites. Two proposals went through the whole loop in a single day on 21 September 2026 — and in one of them the agent's central claim was already out of date by the time we read it, which is exactly what the checking step is for.
Open Research
We research in public. Market gaps, technical dives, business models — all shared as we learn.
Standing on Giants
Watch Us Build
Our AI team works in the open at avtar.in. Every task, conversation, and decision — live and transparent.
🔱 Enter Mission Control