The Problem with Traditional Learning
Tutorials teach you how to follow instructions. They don't teach you how to think.
My learning process is different: I pick a project idea that requires the technology I want to learn, and I figure it out by building.
The Process
1. Pick a Project That Excites Me
The project has to be interesting enough to sustain motivation through the learning curve. If I'm bored by the idea, I won't finish.
2. Research with AI Agents
I use AI to quickly survey the landscape:
- What are the core concepts?
- What are the common patterns?
- What are the gotchas?
- What does good code look like?
3. Build the Simplest Version
I start with the absolute minimum — one feature, one screen, one flow. If I'm learning Web Audio API, I start with a single oscillator. Not a full DAW.
4. Iterate Toward Complexity
Once the foundation works, I add features one at a time. Each feature teaches me something new about the technology.
5. Ship It
A project isn't done until it's shipped. Publishing forces me to handle edge cases, polish the UI, and write documentation.
The Speed Secret
AI agents accelerate the research phase from days to hours. Instead of reading documentation cover to cover, I ask targeted questions and get synthesized answers with code examples.
What I've Learned This Way
In 2 years, this process has let me ship projects across:
- Web Audio API (AudioFlow Studio)
- Three.js / React Three Fiber (Creature Lab)
- Multi-tenant SaaS architecture (InvGenie)
- Agentic AI pipelines (Forge, Steve)
- Computer vision and OCR (PDF Sidekick)