I built a working YouTube title generator in ChatGPT Codex — from a single prompt, using three bounded agents instead of one that does everything.
Not a time-lapse of a chat window. This is the actual decisions: how the work was split, what each stage added, what got verified before I called it done, and what the tool still cannot do.
HookLab turns one video idea into 10 titles across three strategies, 3 opening hooks, and 5 thumbnail phrases. It runs as plain HTML, CSS and JavaScript in your browser: no account, no API key, no hidden usage bill. Codex helped build the tool; the tool itself is deterministic.
▸ FOLLOW ALONG — the exact prompt, copy-paste ready:
https://www.aicreatorsroundtable.com/...
WHAT YOU GET IN THIS VIDEO
• The single Codex prompt the whole build came from, verbatim
• Why the work was split into three roles — one orchestrator, one read-only reviewer, one scoped implementer
• The six build stages, and why a smaller useful scope beats a giant prompt
• The verification checklist that was actually run before calling it done
• The limits, stated plainly: no fact-checking, no analytics, no proof any title will perform
CHAPTERS
0:00 The claim, not the process
0:35 What HookLab is — and what it is not
1:18 The four inputs
1:47 Three title directions, one human call
2:16 Sol, Luna, Terra: three bounded roles
2:54 The six stages (and why scope matters)
5:08 Verification: QA and content quality
6:33 The actual content workflow
7:07 Limitations, and who should skip it
8:37 Build it yourself, or hire me
WHO THIS IS FOR
Creators, freelancers and small teams who want a faster way to explore packaging ideas
without opening a blank document. If you are looking for a button that replaces research,
taste or honesty, skip it — this is not that.
AI can speed up the draft. It cannot care about the work for you.
▸ More practical AI content that shows the process, not just the reveal:
https://www.aicreatorsroundtable.com/
#ChatGPT #Codex #AIAgents #YouTubeTitles #AIWorkflow #ContentCreation