I built a live AI art voting platform in public — 33 migrations, one atomic RPC, and a methodology for working with AI that actually holds up
12 min read →
Directed Output
AI-assisted development done with intention. Every decision behind this project is documented in the build log. The methodology is at the refinement loop. Built by Oliver White.
The difference
Vibe coding
Directed Output
This methodology built something real
AI Art Arena is a live voting contest platform built entirely through the Directed Output process. Every decision is documented. Every line is understood. See the project →
How it works
01
Get something on the table. The first prompt does not need to be perfect — it needs to be honest about what you want.
02
Do not ask if it works. Ask if you understand it. Could you debug this at 2am when production breaks?
03
Vague follow-ups get vague results. The more precisely you describe what is wrong, the closer the next response gets.
04
As many times as it takes. Each pass deepens your understanding and sharpens the output.
05
Not when it looks done — when you can explain every line. That is the difference.
From the build log
All posts →Follow the build
The dead ends, the redirects, the fifth iteration that finally made sense. No noise. Just the work.
4 votes
7 min read →