Shared automation can speed up every agent
Boris as quoted by Theo argues that infrastructure and developer-experience automation now benefit every agent you run. Work that once sped up an engineer can also speed up the agents using that environment. Theo agrees with the argument that more automation can mean more output per unit of time.
Fix shared delays before adding more agents
Before increasing the number of parallel agents, examine the slow steps those agents share. Profile environment setup, test execution, and waiting for checks to identify where tooling effort belongs. Fix those shared delays first when deciding how to support more concurrent agent work.
Concurrent agents benefit from removing shared delays
Use this argument when allocating tooling time for a team running multiple agent sessions at once. Focus that investment on setup and checking delays shared by the agents using the environment. If you are not running concurrent agents, do not justify the investment through an unobserved agent-count multiplier.
The automation gains have not been measured
The video asserts the direction of the benefit without measuring how much extra output automation produces. Theo's agreement does not supply a measured return for your own tooling investment either. The idea still gives you a reason to examine shared delays before expanding parallel work.