The more time passes, the more I feel the need, as someone who has always been passionate about computers, to take back control of the software I use.
I’m tired of increasingly bloated software, telemetry embedded everywhere, features I never asked for, and tools that decide what my workflow should look like.
I’m not necessarily looking for minimal software, but I’m looking for software that lets me choose.
In agentic development, for example, I want to start with a minimal harness and decide for myself how to extend it, which tools to add, and how to orchestrate my workflow.
I don’t want to adapt the way I work to decisions made by someone else.
That’s one of the reasons I find Pi such an exceptional tool. Pi starts from an extremely simple foundation and leaves the responsibility, and the freedom, of building on top of it to me. It provides the building blocks, but it doesn’t try to decide what the final system should look like.
And it’s also why Omarchy has completely won me over lately. Omarchy is definitely opinionated. It has very specific ideas about how a Linux desktop should work.
But paradoxically, those opinions make it a great starting point for building something deeply personal.
Hyprland, the tools the system provides, and especially the ability to easily create new functionality, including through dedicated LLM skills, constantly turn this:
“I wish my desktop could do this.”
into:
“Okay, now it does.”
And that has brought back some of the wow factor of computing that I feel I had lost over the years.
That feeling from when a computer wasn’t simply a device for consuming products built by someone else, but something to explore, modify, break apart, rebuild, and adapt to the way you think.
Over the last few days, for example, I found myself wanting a few things that simply didn’t exist in my desktop the way I wanted them.
So I built them:
And this is probably the part I find most interesting.
With today's AI tools, the distance between “I wish this existed” and “I built it” is becoming incredibly small.
That changes the relationship we can have with software.
I realized that, in the end, this is the same principle I’m trying to bring to Tabularis.
An opinionated solution, because good defaults matter, but not a cage. Software that can be extended, modified, and adapted.
That’s also why Tabularis has a plugin system: I don’t think I should be the one deciding every database, integration, or workflow the application will ever support. The core should provide good defaults and strong primitives, while leaving room for people to take it somewhere I never anticipated.
And maybe that’s the direction in software that excites me the most today:
Opinionated by default, hackable by design.
I don’t want AI to simply do more things for me. I want it to help me shape software around the way I think and work.
The future I’m interested in is one where software doesn’t decide for me, but gives me the tools to make it do exactly what I want.
And Omarchy reminded me how much fun that can be.