Frank
1. The Gap Is Wider Than You Want to Admit
I spun up a Trilium Notes instance over a year ago to catalog my entire music library, complete with lyrics, style tags, and media files. Proper metadata, proper structure, the whole setup.
I log into it maybe once a month, if that.
Meanwhile, I’m in Emby every single night. I’m in Vaultwarden constantly because I’d lose my mind without a password manager. HomeBase, the asset tracker I built myself in React and TypeScript, gets opened several times a week. The tools I reach for instinctively are almost never the ones I put the most effort into configuring.
The gap between “things I set up” and “things I actually use” isn’t a planning failure. It’s something more interesting than that, and more inconvenient.
2. Daily Use Has Nothing to Do With How Hard You Worked to Deploy It
Authentik running SSO on my public-facing sites took me a solid weekend to configure properly. Authelia for the internal services took another stretch of time. Both run quietly in the background, and I almost never think about them, which means they’re working exactly right.
But I don’t interact with them. They’re infrastructure, not tools.
The thing I actually interact with most is Caddy, because I’m tweaking reverse proxy configs more often than I’d like to admit. Not because Caddy is complicated, but because I keep adding things to the homelab that need routes.
High effort to configure does not equal high daily value. Sometimes the opposite is true.
3. Some Tools Are Alive. Some Are Museums.
SunoHarvester sits half-finished on Scooby right now. Built it in Node.js with Puppeteer to bulk-download my Suno tracks with metadata intact. Works in bursts. Hasn’t been touched in a while.
Crumble, my PHP and MySQL recipe manager I built from scratch with Claude Code, gets used when I remember it exists, which is not often enough to justify the build time.
There’s a version of me that operates in pure hyperfocus mode, builds something start to finish, and then moves on. Those projects are time capsules. The tools that survived that cycle and still get used daily are the ones that solved a problem I bump into constantly, not just once.
4. The Hidden Cost Is Not Storage. It’s Context Switching.
Running a service you don’t use isn’t free. It’s not just compute overhead. It’s mental overhead. Every abandoned service is a small background process in your brain, using a little bit of memory. You know it’s there. You feel vaguely guilty about it. You wonder if it’s still running correctly.
My four NAS units, Rollo, Lamont, Grady, and FredG, each have a clear job. That clarity means I don’t have to think about them much. They just do the thing.
The containers that create friction are the ones with ambiguous purpose, the ones I set up to “try out” and never made a real decision about.
5. The Tools I Use Daily Have One Thing in Common
Emby for media. Vaultwarden for credentials. HomeBase for asset tracking. Caddy for routing. The Anthropic API baked into HookHouse-Pro. These all do one job, do it well, and put something in front of me that I actually need.
The tools that collect dust are almost always the ones I deployed because they seemed cool, or because a post on a homelab forum made them look indispensable, or because I thought I’d build a habit around them. I didn’t.
You can’t build a habit around a tool just by installing it.
6. Forgetting About a Tool Is a Data Point, Not a Character Flaw
I used to interpret a forgotten service as laziness or poor planning. I don’t anymore. If something I deployed six months ago hasn’t become part of my regular workflow, that’s information. It means either the tool doesn’t actually solve a real problem I have, or it solves one I’ve already solved a different way, or the friction to use it is just slightly higher than the value it returns.
None of those are disasters. They’re just signals worth listening to.
The honest move is a periodic audit, not to feel bad about the graveyard, but to decide which containers deserve to keep running and which ones should get a clean shutdown and a proper goodbye. Entropy is not a homelab strategy.
