Skip to workspace
Will Jennings

Three questions to revisit.

Earlier decisions, explored with what we can build now. These are new simulations—not recordings of shipped behavior.

Can it get started without cutting you off?

The original question

When someone says “set a timer for eight minutes… called…”, should the timer start, or should the assistant wait for the rest? Does starting imply that listening has ended?

The question now

Can progress and listening be independent? Start the reversible action, leave its name open, and make that distinction visible.

“Set a timer for eight minutes…”

08:00

Unnamed timer

Ready to compare

The timer has not started.

Use the same request with each treatment. No microphone or model is connected.

What to notice

At the pause, would you assume you can still speak? Does the acknowledgment give you room—or hurry you along? This sketch tests legibility, not whether a model can detect the end of a turn.