GUIDE

Two Claudes, one project, no stepped-on toes

You opened two Claude Code sessions on the same project because you wanted to move faster. Instead, files you just finished keep reverting, features disappear, and neither session admits to touching anything. This happens because two AI sessions working on the same files don't know about each other. Here's why it happens and how to stop it.

Why two sessions undo each other

Each Claude Code session reads the files on your computer, makes changes, and saves them. It has no idea another session is doing the same thing somewhere else. When both sessions save around the same time, the second save simply replaces the first one. Nothing is merged. Nothing asks permission. The version that saves last wins, and the other session's work is gone like it never happened.

This is different from your AI making a mistake. The code from the first session might have been perfect. It just lost a race to be saved last.

The tell-tale signs

If you see two or more of these in one afternoon, you likely have two sessions stepping on each other, not a confused AI.

Setup 1: one session at a time, always

This is the simplest fix and the right one for most people. Close one session fully before you open another. Wait for it to finish saving. Then start the next task in a new session.

It feels slower. It is not, once you count the hours lost to disappearing work. If you don't code and you're not ready to think about branches, this is the honest answer: run one session at a time.

Setup 2: split by area of the app, with a shared instructions file

If you genuinely need two sessions running together, give them separate territory. One session works only on the sign-up page. The other only on the dashboard. Neither touches the other's files.

Keep a plain text file in the project, something like instructions.md, that both sessions read at the start of every task. Write down which areas belong to which session, and update it the moment that changes.

This works until the two areas share a file, like a shared navigation component or a global config file. The moment both sessions need to touch that one file, you're back to the same race condition, just less often.

Setup 3: separate branches, merged on purpose

Plyo runs on real git underneath, so you're not limited to one shared set of files. Give each session its own branch: one session branches off to work on the sign-up page, the other branches off for the dashboard. They can save as often as they like without touching each other's work, because they're not writing to the same place.

When a session finishes, you bring its branch in on purpose, instead of letting two sessions silently overwrite each other on the same files. Because every version is saved automatically, you can see exactly what each branch changed before you decide to bring it in, and one-click undo is there if merging in a branch breaks something. When you're ready, one-click publish runs a secret-key scan first, so you're not shipping credentials by accident along with the merged work.

This setup takes more attention than Setup 1 or Setup 2. It's worth it once you're running two sessions often enough that a shared instructions file keeps failing you.

Plyo is the home for AI-built projects: every version saved, one click to undo, one click to publish, plus a safety check, visitor counts and backups. Free to start: 3 projects, no card needed.

Give your project a home