Some important rules when considering adding ML to your app:
1. Start building without machine learning
2. Continue without machine learning
3. Be blissfully unaware of concept drift, model versioning, AUC, and feature engineering
4. Turn off the computer
5. Go outside and buy a small child an ice cream cone
6. Sit with your beloved under the stars and, wonder at the vastness of the universe and your small role in it
7. Turn the computer back on, you’re now ready to add only linear regression
This week for #projects — in the vein of what @simon has been posting — I have been partnering with @miek on his https://github.com/miekg/gitopper, which a small, simple agent to manage GitOps on a node.
It's a therapeutic project to work on. I have a lot of configuration management sins from earlier in my life to atone for, so I want to do right. But in all seriousness, I find the problem statement compelling: https://miek.nl/2022/november/15/provisioning-services/.
Happy hacking!
Well, that wasn't hard at thanks to fuse-go
https://github.com/miekg/mutfs
damn it; it's just this: https://github.com/hanwen/go-fuse/blob/master/example/winfs/main.go
*slightly* modified... hmmpf
Ha, rest over ssh, with public key auth https://github.com/miekg/gitopper/pull/17
With the Twitter reinstatement of the former US President who attempted to incite a fascist coup after losing an election, now would _really_ be a good time to stop cross-posting your Twitter RTs here.
If you feel that cross-posting is the only way to maintain a presence here, please consider limiting it to original posts without mentions or RTs, to avoid cluttering timelines with Twitter URLs. Thank you kindly for your consideration.
https://github.com/miekg/gitopper is feature complete, I just added a way to bootstrap the whole thing.
It's amazing simple still considering what you can do with it