I have been thinking about this a lot this week. The last 3 days or so I have been working on setting up new Postgres servers with replication running over Wireguard and migrating existing connections/apps, which requires they become wg peers. I worked this out with Claude Code, but I ended up making about 15 versions of the plan file to do it. This happened as I started executing it and I found problems with the plan, details left out, steps out of sync. It got pretty exhausting, but I wanted my plan to end up representing not just execution, but also good documentation on how to reproduce everything.
I thought about the old days were I would follow a guide like this,
https://autoize.com/migrating-... it is a lot easier to understand this at a glance and certainly gives me the concepts of Wiregaurd a lot faster than learning them via Claude mistakes. However, it also leaves out a GREAT DEAL that is actually important to my setup. Overall, the end result and the documentation I made is much better than if I had followed some combination of existing articles this way. Because frankly, I wouldn't have made the documentation if I had taken this off the shelf path.
Claude Code did some unbelievably stupid things especially in regards to debugging networking issues, however, it also pointed me towards a clue that when my local Wifi router changed(I have two) that caused my Windows peer to break. Of course Claude kept digging at my Windows firewall
:)
So frankly I probably didn't gain any speed on this approach, but I did actually do a better job. The plan file on Google Docs is 21 pages now. But I can easily expand my start here to be a far better (cheaper, more secure, better performance) alternative to something like AWS RDS.