Reversibility is one of those words that sounds simple until you try to build it. Everyone wants an undo button. Very few systems can actually deliver one.

When we say every action in the Ledger is reversible, we mean something specific, and it's worth explaining honestly, including where the limits are.

Not everything can be undone

Let's be direct about this first. Some actions can't be reversed, because the real world doesn't have an undo button. Once an email reaches someone, they've read it. Once money moves, it moved.

Pretending otherwise would be dishonest. So we don't.

What reversible actually means

For actions that happen inside connected systems, reversibility means the agent records enough about what it did to put things back the way they were.

A status it changed can be changed back. A record it created can be removed. A field it updated keeps its previous value, so restoring it is one click, not a reconstruction.

The Ledger doesn't just say what happened. It stores what it would take to undo it.

Where Checkpoints come in

For actions that genuinely can't be undone, like sending an external email or moving money, reversibility isn't the safety net. The Checkpoint is.

Those actions hold for human approval before they happen, precisely because they can't be taken back after. You can't undo it, so you get to decide before.

Honesty as a feature

We could have marketed reversibility as magic. Instead we built it to be exactly as good as it can honestly be, and we put a Checkpoint in front of everything that falls outside it.

That's the whole philosophy. Undo what can be undone. Hold the rest for a human. Never pretend the line isn't there.