When I was in design school, I discovered something intoxicating: I could make almost anything I wanted. Design schools are essentially small-scale manufacturing plants, wood shops, laser cutters, CNCs, spray booths, print studios, screen printing setups, and painting studios. If you could imagine it, you could probably build it.

That feeling never fully transferred to my digital practice. When I moved into product design and creative technology, I made things to ship, things to sell, things for clients. Rarely did I make bespoke digital tools purely for myself. The economics didn't make sense. Why spend a week building something when you could subscribe for $10/month?

Something shifted recently.

The tools changed everything

Last week, Anthropic introduced Cowork, a desktop tool that lets you work alongside Claude on tasks across your computer. For me, it's become an incredible brainstorming partner for code. I describe what I'm trying to build, we work through the architecture together, and by the time I open Cursor (an AI-native code editor), I'm already unblocked. The gap between idea and working prototype collapsed.

This is what made me realize: I can treat software the way I used to treat the wood shop.

The experiment

My goal for 2026 is simple: every time I need a new piece of software, something I would normally pay for, subscribe to, or download, I'm going to try building it myself first.

By the end of the year, I want to have accumulated a collection of tools built purely for me. Software that does exactly what I need, nothing more. No features I'll never use. No dark patterns. No subscriptions.

Along the way, I want to develop my own consistent design language across everything I make. A personal aesthetic that ties it all together.

The exceptions

I'm not a purist about this. Some things don't make sense to build:

• Communication tools for work: Email, Slack, Browsers, and messaging platforms. These need network effects and a serious security infrastructure
• Core creative tools: Obsidian for notes, Cursor for code, Sketch for designing interfaces

This list will probably grow. That's fine. The point isn't to build everything, it's to build first, buy second.

What I don't know yet

There's a lot I haven't figured out:

• Maintenance: Do I need to worry about updates? What happens when dependencies break?
• Data: Am I running everything locally? Do I need databases? Where do they live?
• Design systems: How do I keep a consistent language across dozens of small tools without it becoming a burden?

I genuinely don't know the answers. That's part of what makes this interesting.

Questions I want to answer

Beyond the practical unknowns, there are bigger questions I'm hoping this experiment helps me explore:

What does it mean to have my own design language across all my software? Not a design system in the corporate sense more like a personal aesthetic. The way a craftsperson's work is recognizable across different pieces. What would it feel like if every tool I use looked and felt like mine?

Where's the line between building software and building an operating system? At some point, enough custom tools start to feel like something more. Is there a meaningful distinction? Does it matter?

Does the math actually work? At the end of 2026, I want to add up every dollar I spent on LLM API costs, Cursor subscriptions, and compute and compare it to what I would have spent just buying the software I needed. I genuinely don't know which number will be bigger.

Phase two

If this pace feels sustainable, I have a second goal: start redesigning every tool I currently use but don't enjoy. The software that works but feels hostile. The apps that technically do the job but make me slightly miserable every time I open them.

That's the vision anyway. We'll see how it goes.

---

This is a making-in-public experiment. I'll be writing about what I build, what I learn, and what breaks along the way.

Tools built: