You work in a session for an hour, fork it to try an alternative, work inside the fork, then reopen the original the next morning and carry on there. Afterwards, what is true of the two session identifiers?

The principle — a fork is a copy taken at a moment in time, and the two halves stop influencing each other immediately afterwards.

Why the key is correct — Anthropic separates the two operations explicitly. Resuming a session reopens it under the same session ID and appends new messages to the existing conversation; forking copies the history into a new session ID, leaving the original unchanged. Trace the sequence with those two rules and the result falls out. The first hour belongs to identifier one. The fork produces identifier two, which begins life holding a copy of that hour. Work inside the fork lands on identifier two only, because the original was left unchanged at the moment of copying. The next morning you reopen identifier one and append to it, which touches nothing belonging to identifier two. Both now hold the same first hour, and each holds its own work since.

Why the others are wrong — a single identifier would require the fork to rename rather than copy, and there would then be no original left to return to. Having the fork's later work appear in the original describes a link that is severed the instant the copy is taken, and it would remove the whole reason to fork before something risky. Reopening pruning later messages inverts what resuming does, which is to append; it also has no reach into another identifier's history.

Remember this — fork to branch, resume to continue. After the fork the two never speak again.

Sources — Anthropic's How Claude Code works and the Claude Code glossary.