The bottleneck moves from the agent to your decisions
Before and after
Before: you ask, then wait while the agent works, and your attention drifts. After: cards build in the background and tweaks land in seconds, so the next thing that holds you up is deciding what to build next.
What that changes for you
Your attention works at a higher level: what to build and whether it's right, not watching it being built.
The new limit
Once the bottleneck is you, your attention is the scarce thing. Five cards finishing means five pull requests and possibly five questions. That's why stride caps how many cards run at once, the parallel-sessions setting doing the job of a Kanban limit on work in progress. It's also why cards bring you only the decisions that are yours.