AI Parallel Time Multiplier (AI PTM): a method for measuring parallel time with AI agents, version 1.0

The AI Parallel Time Multiplier (AI PTM) measures how much parallel time a person acquires from AI agents: the average number of AI work streams (agent sessions and the sub-agents they launch) active in each minute of active work. For each clock minute m, n(m) is the number of streams active, where a stream is active between consecutive events no more than an idle cut-off tau apart (tau = 30 minutes in version 1.0). AI PTM = parallel hours / clock hours = sum of n(m) / number of minutes with n(m) ≥ 1. This note gives the definition, its properties, a worked example, its limitations, a family of companion metrics (peak parallelism, delegation ratio, autonomy span, prompt leverage, first-pass rate, framework adherence and time collapse factor) and how to cite it. It credits METR's February 2026 analysis of agent concurrency as prior art. The reference implementation is the open-source buildstats collector.