Back to Process
Process

Project Ideation & Planning

Every project starts the same way for me: a notebook and a pen. Before I touch any code, I need to think through the idea, figure out if it's even possible, and plan how I'm going to build it. This process has saved me from wasting weeks on things that wouldn't work.

The key is being honest with yourself early. Is this actually buildable? Can I afford the tools I need? What's the fastest path to something that works? I'd rather spend an hour sketching flows and crossing out bad ideas than discover problems three weeks into development.

The Process

01

Idea Generation

Start with a high-level concept. Doesn't need to be fully formed, just enough to get excited about. Write it down before it disappears.

02

High-Level Approach

Consider platform constraints and architecture decisions. Backend vs monolith? API design? What are the non-negotiables?

03

Viability Research

The reality check. Research costs, test tools, find affordable solutions. If it's going to cost $500/month to run, maybe rethink the approach.

04

Initial Build Phase

Set up the foundation. Pick your stack, scaffold the project, get something running. Speed matters here.

05

User Flow Sketches

Sketch out core flows on paper. Login, search, main actions. These don't need to be pretty, they need to be clear.

06

Complexity Testing

Build a minimal version to test the hardest parts first. Find out early where the dragons are hiding.

07

Iterate & Pivot

Adjust approach based on what you learn. Sometimes the original plan was wrong. That's fine. Pivot fast, don't get attached.

08

Solve Hard Problems

Tackle the identified challenges head-on. This is where most projects die. Push through.

09

Deep Dive & Refine

Hands-on testing, tool selection, polish. By now you know what works. Double down on it.

Example: LearnEngine

1

The Idea

Dynamically generated learning engine
2

High Level Approach

Avoid rebuilding app depending on OS
Probably organized a backend accessed using API calls
3

Viability Research

Research cost of models for backend
Start with open source models, ideally free
OpenRouter is the window into these models
Test models to find the right fit
4

Build Phase 1

Build Next.js app using Bun
Code with Antigravity
Add Supabase to App
sketchLogin flow
sketch"What to learn" search bar
sketchExisting lessons cards
OpenRouter for AI use
5

Backend & Complexity Testing

Create backend to test complexity of minimal requirements
Learning: storing User Data is hard
Create a frontend to access the backend
Pivot to Next.js build of front and back — still backend centric + faster to launch
Explore saving User Data!!! → Supabase to the Rescue?
6

Deep Dive & Model Selection

Supabase hands-on build
Test AI models for fit (must be free)
DeepSeek R1 T2 → meets criteria ✓
MiMo V2 Flash → same but faster ✓

Tools & Technologies

NotebookPenWhiteboardFigmaGoogle Docs