A shell command sent to the background, and a prompt set to run on a schedule in the cloud, are both sometimes described as delegating work. What do they actually have in common?

The principle — two mechanisms look like parallel delegation because neither makes you wait. Neither gives you a second agent beside you.

Why the key is correct — Anthropic rules each out in its own sentence. A background bash command runs one shell command without blocking the conversation, and it doesn't spawn an agent. A routine runs a session on a schedule in the cloud, not in parallel on your machine.

Why the others are wrong — there is no lighter class of agent here — one of the two involves no agent whatever. Nor are both parallel: a scheduled run happens in the cloud on its own timetable. And treating non-blocking as the test would make almost anything delegation.

Remember this — not waiting is not the same as being helped. Ask what got started, and where.

Sources — Anthropic's Claude Code agents documentation.