From UX Designer to Builder: How AI Helped Me Finally Ship My Own Platform
I recently built https://dotafury.gg
— a Dota 2 analytics platform.
But this post isn’t about the product.
It’s about something more important:
How a UX designer with limited dev control finally started building real systems using AI.
⚠️ The Problem I Always Had
For years, I lived in this gap:
- I could design systems
- I could imagine flows
- I could see product opportunities
But I couldn’t fully build them.
Everything depended on:
- developers
- timelines
- priorities
And a lot of ideas never made it past Figma.
🤖 Then Came AI (And “Vibe Coding”)
At first, it felt like magic.
I could say:
- build this page
- create this component
- connect this logic
…and it actually worked.
For the first time, I wasn’t just designing.
I was building.
💣 Reality Check: Vibe Coding Breaks Fast
Very quickly, things got messy:
- duplicated logic
- unclear structure
- random patterns
- fragile code
That’s when I learned:
Working code ≠ a working system
AI was fast…
but it was also amplifying my lack of structure.
🧠 The Shift That Changed Everything
I stopped asking:
“Build this feature”
And started asking:
“What is the system behind this?”
That changed how I worked with AI completely.
🏗️ My Actual Workflow (That Worked)
Instead of prompting randomly, I moved to:
1. Define the system first
- What are the layers?
- What owns what?
- What should be cached?
- What should NOT hit APIs repeatedly?
2. Enforce structure
I locked myself into:
UI → Services → Integrations
No shortcuts.
3. Use AI in small, controlled steps
Instead of:
❌ “Build player page”
I did:
- Create data service
- Add caching logic
- Then build UI
🤝 Where Codex Helped
I used Codex as a collaborator, not a magician:
- scaffolding structure
- refactoring
- implementing scoped tasks
- validating patterns
It worked best when I gave it:
- constraints
- context
- clear boundaries
🔁 Trial & Error Was the Real Teacher
This wasn’t smooth.
There was a lot of:
- rethinking
- rebuilding
- simplifying
- undoing bad decisions
Sometimes the problem wasn’t code.
It was unclear thinking.
🎯 My UX Background Became a Superpower
Here’s what surprised me:
UX thinking translated really well into building.
I already understood:
- flows
- edge cases
- user intent
- system behavior
AI just gave me the missing layer:
execution
🔥 The Real Unlock
Before this, my ideas stayed in:
- documents
- designs
- conversations
Now I can go from:
idea → system → working product
That changes everything.
🧪 What I Learned
- AI doesn’t replace thinking — it exposes it
- Vibe coding helps you start, not scale
- Structure matters more than speed
- Small steps beat big prompts
- UX + AI is a powerful combination
🚀 Final Thought
AI didn’t make me an expert overnight.
It gave me something more valuable:
the ability to stop waiting… and start building.
👉 If you’re curious, this journey led to:
https://dotafury.gg
Still evolving. Still learning.
But now — actually real.