It is very unfortunate to share that today I had some important project files deleted by Claude's code. I wouldn't completely blame Claude, as I was the one approving all the command requests one by one. I should have been a little more careful. I wanted to merge two repositories and create a mono repo to make it easier for Claude to work in the future. In this process, I was following whatever Claude was suggesting to me. Only then did this blunder happen.
If you look at the image above, you can see how cool it is to deliver this bad news to me. One thing is very clear with this experience. We cannot fully rely on Claude to do all the activities for us. We have to keep our eyes wide open when Claude is doing some activities. This will help us be safe from such blunders that are happening. Another lesson is to have a backup done before all these activities. Even if something goes wrong, I should be able to recover. We shouldn't rely completely on the AI to take care of everything. They make mistakes too, and if they make a mistake, the toll is on us. So we have to be clear in validating all the actions that AI is taking.
Yes, fortunately, I have Syncthing running on my system, where all the files are backed up to a remote server for safety. Thankfully, I was able to recover my files from the remote server, where a backup and restoration option was available. I still have to validate if the restoration brought back all the files or not. Having a backup system like this is very useful, but it is essential to make sure mistakes like this do not happen in the future, and I have to be very careful when using AI tools.
If you like what I'm doing on Hive, you can vote me as a witness with the links below.
@balaz as a Hive Witness 
@kanibot as a Hive Engine Witness|
|
|
