Who needs working restores when you can just do 10 more backups? 😵‍💫.

But I realise that I don't use most of the Mac features and only stay in the console most of the time. I hate the finder, I hate time machine, I hate the app bar, . not sure what is left to be honest? Why should I stay on Mac? Only reason right now is focus, which I need for.

What do you want to know?.

Should I move back to Linux as a desktop? There is such a hype around it right now :D. I was using Linux for many years as my sole desktop/dev OS but switched to Mac after I realised that I put too much time into customisation. I enjoyed the golden cage for ~4 years now. .

I keep going back to SQL. no matter how much I hated it in the past :D (I thought for a day that maybe the key-value store is enough for this one use-case. but nope. I need transactions again. ).

IMHO, Claude Code and friends are way too cheap for what they deliver. Just today, it did work worth hundreds of $ for me while I only paid ~10€. I'm absolutely willing to pay more for it, when at the same time I get the guarantees that it keeps working without bottlenecks or.

I'm considering implementing this as a standalone open source project. Would allow cloud agnostic incremental backups/sync of volumes.

Thin provisioned volumes and snapshots with LVM2 and ext4 on top of them perform nearly as good as plain ext4. Thin provisioned snapshots can also be used to effectively find the differences between two snapshots, so a highly performant incremental volume sync is maybe realistic.

Aaand it's not going to be btrfs. Looks like it's still having serious performance issues when it comes to databases. Did a lot of benchmarking today and best case was still a 50% reduction in performance compared to ext4. And yes, even with copy-on-write being disabled.

Here is the initial patchset: I had to learn git via mailing lists to get this accepted into the Linux Kernel :D. Still extremely proud of this work, even though I then never actually used it in production by myself.

And maybe, 13 years after I initially contributed the btrfs send/receive code to the Linux kernel, it finally pays off that I crunched through the kernel for multiple weeks :D.

The movement between machines is meant to be cloud agnostic and should even work between cloud providers or even to and from on-prem/bare-metal. I'm considering using btrfs internally as it natively supports snapshots and incremental backups (btrfs send/receive).

I assume persistent storage support is the next thing to implement :) A "box" is a virtual/abstract unit that defines docker compose workloads, which run on machines. The box can move between machines, and the storage must move as as well. Backup/restore should also be a breeze.

Big milestone achieved today. "Boxes" are auto-reconciled now, meaning that services restart and get added/deleted when the box spec changes. Getting closer to the the first open-source release. After that, it's time to revive the cloud version and onboard the first users.

"implement darkmode" and it just finishes it in 30 seconds. Something that would have taken me at least a full day in the past. I'm getting 🤯 day after day. .

A "box spec" is a spec for a box (👋 from captain obvious). A "box" is the abstraction workloads that will be run on a machine (cloud/metal). A box can move between machines, including persistent storage. A box also has a virtual network (using attached.

Back to building features now, much more fun and rewarding than building boilerplate SQL stuff. Current feature: Automatic pulling of updated "box specs" which are then re-executed on the target machines. Using as the technical basis for communication.

ORM was the end-boss for me in the last 5 days. My journey went from sqlc (SQL -> go code gen.), to GORM (a full blown object rel. mapper) and finally back to a self-built SQL->object scanning and plain SQL generation. Using sqlx now and hopefully won't have to switch again 🤪.

RT @kluctl: Kluctl v2.27.0 is finally ready :). It's mostly a maintenance release with all dependencies being updated and minor fixes. htt….

What would you recommend when it comes to AI assistance when preparing slides for a talk you want to give? I need help in regard to visuals and shortening of information.

First time ever I got complimented on my UI/UX skills, never imagined this would ever be part of my reality 😅. Thanks to AI.