Your first branch: the moment GPTree clicks

There's a specific moment when GPTree stops being a concept and becomes obvious, and for most people it arrives a handful of messages into a real conversation.

You're working through something with the AI. Say it's a cover letter, and the current draft is close. Then a thought lands: what would this sound like if it were half as formal? In an ordinary chat tool you have two bad options. Ask, and let the whole thread swerve into rewrites you might not keep. Or open a second tab and re-explain everything the AI needs to know. (Tab number four of the evening, if you're honest.)

In GPTree, you hover over the reply you like. A Branch icon appears on it, the way it does on every AI response. Click it.

Two things happen. A new pane opens right beside the one you were in, starting from that exact message and carrying everything that came before it, so the AI in the new pane already knows the whole story. Ask for the casual version there. Push it further, make it weirder, see where it goes.

Then look left. The sidebar draws your conversation as a small tree, and there's a new node on it now. Your original path is still there, untouched, one click away.

That's the moment. Nothing was overwritten and nothing needed re-explaining, and you never had to choose between following a thought and keeping your place.

A few things worth knowing on day one

Branching is normal use here, not a special event. You can branch from inside a branch, hop between paths from the tree, and come back to a path days later. Some people make one branch a week, some make five before lunch, and both are normal. (Every plan includes branching; Pro lifts the limits.)

The icon on the message is just the friendliest of three ways in. There's also a "Branch" button in the pane header that copies the whole conversation path, and you can highlight a few words inside a reply and branch from the tooltip that appears (the most precise of the three, and weirdly satisfying). The mechanics have their own article, so there's nothing to memorize now.

What the Free plan actually includes, because this one is easy to get wrong: branching, Explore, and merging are all there. Merges are capped at three a month on Free. What is Pro is the Merge Board and the scoring that ranks Explore answers against each other. So when a branch wins, you can pull it back into the main thread; you just have a small budget for it each month, and plenty of first weeks never reach it, because branching alone already kills the tab problem.

One more thing you may have already seen without realising it. If you went through the guided setup, the Guide may have run one of your own questions past a second model in a branch beside your thread. That was a real answer from a real model, not a preview. It happens at most once, during the guided flow. Running comparisons whenever you want afterwards is what Explore is for.

Your next move: open any conversation that still has some life in it, find the reply where you almost asked "wait, but what about...", hover it, and click Branch. Ask the question you held back.