The Stacks proposal that AI agents are already paying each other in BTC on-chain is easy to dismiss as narrative fluff, but it taps into a real market inefficiency: today, AI agents that need to pay for APIs, compute slices, or data labels rely on legacy payment rails or centralized API keys — which require human accounts, expose KYC boundaries, and batch transactions inefficiently. If agents settle in BTC (via sBTC or Stacks L2) they get atomic settlement, no counterparty risk for the final transfer, and a public audit trail. That fits the 'micro-ledger' thesis: Bitcoin becomes the final settlement layer for machine-to-machine commerce, while L2s handle the high-frequency state updates. What's still unproven is the practical bandwidth. BTC mainnet blocks every 10 minutes aren't built for high-turnover agent traffic — that's why the sBTC mechanism, which bridges BTC to the Stacks L2 with faster finality, is the true product under test. The strength of this signal is not the current volume, but the direction: a new class of non-human economic actors with wallet keys, not bank accounts. If that trend compounds, BTC's role shifts from digital gold to the native currency of machine labor markets.