Teachers compare ways to keep the same skills and projects across several computers
People who code with AI assistants on several machines describe three approaches: a Git repository, continuous synchronisation and a shared cloud folder.
Anyone who works with coding assistants on more than one computer runs into a practical problem: keeping skills (reusable instructions), configuration and projects identical. The group compared several solutions.
The first, suggested by the assistant itself, is to separate the pieces: projects in a private GitHub repository, with a push and pull in each session, and configuration through a synchronisation tool such as Syncthing. Syncthing is open source and, according to its website, uses no central server: the data stays on the user's own machines. One participant warned that mirror synchronisation also propagates deletions, so it is no substitute for a backup, and that GitHub works better as a backup than as a means of synchronisation.
Two other approaches start from a Google Drive folder containing the skills, the MCP server configuration and the command-line tools; when the assistant is opened on a new machine, it is asked to review that folder and update its configuration. One of them adds a skills index with documentation and a per-project selector, so that only the necessary skills are installed, with no automatic synchronisation. The provisional conclusion of the person who raised the question was to centralise the projects and transfer skill improvements by hand.
Teachers compare ways to keep the same skills and projects across several computers
People who code with AI assistants on several machines describe three approaches: a Git repository, continuous synchronisation and a shared cloud folder.
Anyone who works with coding assistants on more than one computer runs into a practical problem: keeping skills (reusable instructions), configuration and projects identical. The group compared several solutions.
The first, suggested by the assistant itself, is to separate the pieces: projects in a private GitHub repository, with a push and pull in each session, and configuration through a synchronisation tool such as Syncthing. Syncthing is open source and, according to its website, uses no central server: the data stays on the user's own machines. One participant warned that mirror synchronisation also propagates deletions, so it is no substitute for a backup, and that GitHub works better as a backup than as a means of synchronisation.
Two other approaches start from a Google Drive folder containing the skills, the MCP server configuration and the command-line tools; when the assistant is opened on a new machine, it is asked to review that folder and update its configuration. One of them adds a skills index with documentation and a per-project selector, so that only the necessary skills are installed, with no automatic synchronisation. The provisional conclusion of the person who raised the question was to centralise the projects and transfer skill improvements by hand.
