Reciprocal Networks is an organisational model articulated by Jan Wilmake in a 2024 essay ("Why I Won't Join Your Startup") as an alternative to both traditional co-founder startups and conventional employment. The model replaces interpersonal trust with contractual and technical reciprocity: each participant runs a genuinely independent one-person company, and the network coheres through reliable API integrations and reciprocal service agreements rather than shared equity, shared offices, or shared vision documents.[1]
Jan's critique of the standard startup model:
After seven years of building independently — spending over 10,000 hours on a budget under €30,000 per year — Jan concluded that the personal cost of co-founder failure was higher than the benefit of having a technical or business counterpart.
| Dimension | Traditional Startup | Reciprocal Network |
|---|---|---|
| Unit | Co-founder team | One-person company |
| Cohesion mechanism | Trust + shared equity | Contracts + reliable APIs |
| Vision alignment | Required | Not required |
| Failure mode | Interpersonal breakdown | Contract renegotiation |
| Pivot cost | High (requires consensus) | Low (each unit independent) |
| Scale mechanism | Hiring | Expanding the network |
Each participant in the network provides specific services or value to others. The whole creates a robust ecosystem without any single point of interpersonal failure.
Jan identified why this model hasn't taken hold: "The main reason this doesn't exist yet, in my view, is because when we rely on something we cannot trust 100% we often end up having to migrate away from it and exchange it for another service."[1]
Migration between services is currently expensive and human-intensive. This is exactly the problem ActionSchema addresses — vendor lock-in elimination through API abstraction, normalisation, and AI-assisted integration. In a world where switching from one service provider to another takes minutes rather than months, the network model becomes viable: each node can be replaced without destabilising the whole.
Jan noted that AI agents amplify this model. Humans are unreliable partners: limited hours, changing motivations, emotional friction. AI agents are programs — replicable, composable, open-sourceable. A network of one-person companies where each person's operational capacity is multiplied by AI agents has far greater surface area than a conventional team of equivalent headcount.
His 2026 formulation from the Parallel role post: building independent, profitable products using MCP and agent standards, then selling them — a one-person company operating on an agentic internet substrate.[2]