🧪 Lab Notes
When I'm curious, I just try it. Logs from throwing the same question at multiple AIs, tweaking prompts, and comparing the results.
-
Experiment 11
GPT-5.5 Was Overkill: Three OpenAI Models, One Job, Real Bills
gpt-5.5 vs gpt-5.6-luna vs gpt-5.4-mini: how $0.72 per article became $0.11, and where the money really went.
Jul 17, 2026 -
Exp 10
What Happens When You Let an AI Write Its Own Grading Rubric
"Write the grading criteria, then grade your own answer" — is self-evaluation ever fair?
Jul 3, 2026 -
Exp 9
Korean vs. English Prompts: Comparing the Results
The same request in Korean and in English. Is the quality gap real, or just a rumor?
Jun 30, 2026 -
Exp 8
The Question Order That Makes AI Answers Sharper
Asking for the conclusion first vs. building up step by step. The order of your questions changes the depth of the answer.
Jun 26, 2026 -
Exp 7
Iterating on a Work-Automation Prompt, Version by Version
From v1 to v5 — how my weekly-report automation prompt evolved.
Jun 23, 2026 -
Exp 6
The Moments AI Writing Loses Its Human Touch
I dissected 20 AI-written pieces and found 6 patterns that give the machine away.
Jun 19, 2026 -
Exp 5
Short Prompts vs. Long Prompts: How Different Are the Results?
A 10-character request vs. a 500-character one — where's the break-even point on length?
Jun 16, 2026 -
Exp 4
How Well Do Image AIs Understand Abstract Instructions?
I threw vague prompts like "a lonely Tuesday afternoon" at three image generators.
Jun 12, 2026 -
Exp 3
Ask an AI for 30 Blog Titles — What Patterns Emerge?
I generated 30 titles and sorted them, and the formulas AI loves became obvious.
Jun 9, 2026 -
Exp 2
I Compared 5 Prompts for Summarizing Long Documents
From a bare "summarize this" to role and format instructions — same document, five summaries.
Jun 5, 2026 -
Exp 1
Same Question, Different Answers: ChatGPT vs. Claude
I asked both AIs the exact same 10 questions. I didn't expect their personalities to differ this much.
Jun 2, 2026