Sovereign Machines

For its whole history, an AI agent could only talk. At the edge of every real action, a human held the money. In 2026 that rail came off — agents can now hold a balance, pay an invoice, and charge for their own services, settling in seconds with no bank and no human in the loop. Which raises a question you can get expensively wrong: what happens when the agent can spend money on its own?

Sovereign Machines is the field manual for that moment. Its argument is one sentence: an AI agent with money is the most useful and most dangerous kind there is — so give the machine sound money, and put the wallet behind a leash it cannot cut.

Drawn from a real autonomous agent the author built that earns, holds, and spends real Bitcoin over the Lightning Network — bounded by a cryptographic vault a compromised host cannot bypass, and proven by attacks that failed — this book shows you:

• Why a machine is an economic orphan in the legacy system, and why it needs Bitcoin specifically (the honest case vs. stablecoins included)
• Digital Energy — what a balance really is to a machine, and the economic loop an agent runs on sound money
• The Funded Bounded Agent — the deterministic envelope that judges every payment by rule, so a fooled agent still can't drain you
• The Sovereign Leash — moving the boundary into Bitcoin itself: a 2-of-2 vault with timelock recovery that even a captured agent can't cut
• How to earn an agent's autonomy a rung at a time, and keep a tamper-evident record of every sat it moves

For builders of the agent economy and Bitcoiners watching AI arrive on the network. The bridge from the author's Reliable AI series — Architecting Reliable AI Reasoning Systems, Building Reliable AI Agents, and Evaluating AI Systems — into the economy machines are now building.