My DevOps goals for 2026
Well, would you look at that - we survived 2025.
It’s a great time to reflect on what has been a very pivotal year in my tech
journey, and decide what I want from 2026.
Here are my goals for 2026, a DevOps Roadmap, if you will.
Summing up 2025
This year, I have completed my first year with Dream, Getting down in the trenches of what has probably been the most interesting product I have worked on.
Dream was awesome enough to send me over to Las Vegas to attend my first ever re:Invent!
I finally kicked off the Zero-Reach Stack post series, detailing the tools I use for ditching the mouse and controlling my Macbook/Linux machines with nothing but the keyboard.
Shortly after that, I moved this blog from Wordpress to Hugo, something I can’t wait to write about once I’m done with moving to our new apartment!
I got to work with amazing people and for that I’m grateful, yet I feel like in 2025 I haven’t reached my true potential.
Here are 5 things I want to achieve in 2026:
1) Update my tooling
I’ve been using the same tools for years. The same NeoVim configuration (which extends Rafi’s),
the same terminal emulator - Kitty, and the same keybindings on KMonad.
However, I am not the same person I was 2-3 years ago, and my needs have changed.
What I need from my machine has changed when I joined Dream, and I think it’s time to revisit each choice I made regarding tooling and config. Maybe I need a new keybinding in NeoVim? Maybe a different directory tree
plugin?
Heck, maybe I need some workflows automated, or new syntax highlighting colors.
All I do know is that I need change, and it’s not a bad thing.
So here’s to finding new tools and making my machine closer to what I actually need.
DOD:
- Having my own NeoVim config instead of extending Rafi’s (If you’re seeing this, Rafi, you’ll always be the GOAT!)
- Iron out my KMonad config
2) Automate the sh*t out of everything I can
Automation is the dream, man. I believe that, Ironically, the role of a DevOps
engineer is to automate himself out of a job. Now with the age of Agentic AI,
it’s possible to supercharge my projects in a way that was not possible before.
No, I don’t think AI can solve all of my problems, and it’s not always the right
tool (especially when you want a deterministic outcome), but it’s an amazing
tool for debugging, brainstorming, and learning.
DOD:
- I want to set up an AI agent to send me daily digests on recent infra changes
- Script out 3 repetitive tasks in Bash/Python
3) Get a better feel for MLOps
I work at an AI Security company.
The fact that my only real experience with AI
internals can be summed up to “I implemented Langfuse” is
borderline a criminal offence. This year, I want to play around with running
local LLMs, understanding system prompts, temperature, and fine-tuning processes
in AI.
Given I don’t have much time for gaming nowadays, I think it’s time to put my RTX 4070
Super and 32GB of RAM to good use, as RAMageddon doesn’t seem to be going
anywhere.
DOD:
- Run ollama on my Linux machine
- Be able to understand how models are fine tuned
4) Resist the “Black Box” Workflow
Vibe coding might be cringe, but it’s more addictive than anyone realizes - we are killing our brains with convenience, and AI psychosis is no joke.
I’m ashamed to admit that I too have sinned in “just letting AI do it” - deadlines are tight, production is down, sometimes you just need a quick solution.
That old golang project I hit a wall in 3 years ago? AI broke that wall down.
That NeoVim plugin I needed? AI wrote it.
It enabled me to push the project to completion, but over time, it took the joy out of coding.
If you’d ask me what that code does, the most appropriate anseer would be “fuck if I know”.
The temptation to let the model drive is real, but my goal this year is augmentation over replacement.
I’m treating AI as an intellectual sparring partner - a way to refine my patterns and catch edge cases, while maintaining full authorship of the code.
In DevOps, ‘it just works’ isn’t enough; I need to own the logic to own the troubleshooting.
I don’t get fired for saying this, but… Efficiency isn’t everything.
Sometimes, art is worth pouring time and care into, and computers are just as
much art for me as paintings are to the general public.
DOD:
- Go over my old dotfile manager project and understand the code Claude wrote
- Learn Lua to be able to write NeoVim plugins myself
5) Embrace proactivity over reactivity
This one is a bit special, because it’s professional just as much as it’s personal.
Ever since October 7th, we all became much more anxious, short tempered and
angry.
The whole nation suffers from undiagnosed PTSD. I want to be grounded
again, be patient, open, and calm towards the people around me.
That extends to work.
First of all - because I’m a human, surrounded by other humans. Before AI we
actually bounced ideas off of each other.
Second of all - because observability is king over monitoring.
I don’t want to simply know production is down, I’d like to prevent it, plan for
the infra’s shortfalls, and patch them before they even happen.
DOD:
- Rightsize pods and nodes
- Recover from alert fatigue - only get relevant alerts
The way forward
2025 was about surviving and adapting. 2026 will be about intentionality - reclaiming my tools, owning my logic, and I know it’s gonna sound dumb - protect my peace.
DevOps is often seen as just a series of pipelines, but for me, it’s about the humans behind the screens and the art of the build.
I don’t host a comment section here to keep things lean, but if any of this resonated with you (especially the fight against “vibe coding”) - I’d love to hear from you.
You can contact me on LinkedIn or Email.
Let’s get to work.