Gordon used to handle simple container-related tasks quickly and correctly, but over the past couple of months it has become increasingly unreliable. It often ignores explicit instructions, fails to follow rules, deletes files or even entire containers that it was told not to touch, and sometimes gets stuck in loops that leave everything in a worse state. Has anyone else noticed this change, and is there a safer way to use it or troubleshoot what is happening?
3 Answers
It may be a change or regression in the current version rather than something you’re doing wrong. If it started getting worse around a particular update, check the release notes and consider testing an earlier version in a disposable environment. Don’t let it access production containers or important files until you know what changed.
I’ve had mixed results with it too. It can be useful for explaining commands or suggesting a troubleshooting plan, but I wouldn’t give it permission to delete files, remove containers, or modify a whole project without reviewing the proposed steps first.
For anything important, I’d avoid relying on the assistant to make changes automatically. Use the command-line tools directly, read the relevant documentation, and verify every command before running it. Keep backups and make sure destructive operations require explicit confirmation.
That’s what I’m doing now. I used to trust it with basic fixes, but lately it has only been wasting time and resources.
Same here—I liked it at first, but the behavior became much less predictable over time. I only use it for suggestions now and run the commands myself.