Day job
Keep it running
Windows, Linux, automation and the occasional mess that needs untangling.
Hi, I'm Gabin. Online as Opperiesen.
Gabin
Longoni
I keep systems running by day. After hours, I build overly specific tools for whatever I'm into.
- Online as
- Opperiesen
- Day job
- Systems administrator
- After hours
- Code, racing & games
- Known for
- Taking side projects too far
01 / Me
Mostly systems, side quests and games.
Side quests
Build what I miss
Most projects start with “surely this already exists”. Sometimes it doesn't.
After hours
Learn the system
Racing lines, football tactics and game economies. I like hobbies with something to figure out.
02 / Made
These ones made it out of my head.
01 Live
vim-quest
I kept forgetting Vim commands. Cheat sheets weren't helping.
So I built the practice loop I wanted: real commands, scored keystrokes, bosses and a dojo that brings back what I get wrong.
106levels
30dojo families
129/129cases matching real Vim
file:// · no build · no dependencies
02 Live
linux-quest
I wanted Linux basics to feel like using a shell, not reading a glossary.
So I built the full practice path: a safe shell, 24 focused chapters, one boss per category and manuals for every simulated command.
144missions
107command manuals
0host access
24 bosses · tested solutions · Tab-complete
03 Private
unifi-tools
My home network was becoming a pile of clicks I couldn't reproduce.
So I gave the gateway a Python CLI and a proper export, plan and apply loop. Much easier to trust.
Private, but still useful.
›uv run scripts/export_config.py --out config/network.yaml --force
›uv run scripts/apply_config.py --plan
›uv run scripts/apply_config.py --apply
export → plan → apply
04 Private
mpp-predictor
Everyone guesses football scores. I preferred to make the bookmakers do the maths.
Market odds go in. Shin probabilities, Dixon-Coles and expected MPP points come out. Then I still blame the model.
Runs on my machine. That's enough.
market odds→Shin probabilities
goal model→Dixon-Coles
every score→max E[MPP points]
E = P(exact) × points + P(result) × points
03 / Patterns
A few things I keep doing.
Automate it on the third time.
The first time is discovery. The second is confirmation. By the third, I'm writing a script.
Make it obvious.
Future me should be able to open it six months later and understand what's going on.
Polish the one-off.
I know it only needs to work once. I'll still fix the spacing.
Know when to kill it.
Dead project, deleted infrastructure, useful lesson kept. No guilt.