- Apr 7, 2009
- 5,215
- 2,567
- Awards
- 10
- First Name
- Joe
FULL TRIAGE IS COMPLETE
Total tabs triaged this session (including prior batches): ~90 of 100. The remaining ~10 are the Gemini tabs we're holding for batch processing.
Next step: I regenerate THS_Key_Chat_Reference.docx with every single entry from the entire session — all KEEPs, all ARCHIVEs, all standalone artifacts, properly organized. One file. You save it to 00_PROJECT_HQ/ and close every tab.
Want me to generate it now?
Task: create videos without needing my business partner in Maryland (not Jeff - Tom Harsha) to be available.
Fed a little data into Perplexity and asked it to write a script for a video that would be less than 2 minutes and read by an AI avatar. Then I created an AI avatar of myself, followed by an AI voice that mimics me. I used all the freebie tools in this trial run, then had CapCut's AI add the final polish and captions.
Showed it to my wife, and she did not realize it wasn't me.
Task: create videos without needing my business partner in Maryland (not Jeff - Tom Harsha) to be available.
Fed a little data into Perplexity and asked it to write a script for a video that would be less than 2 minutes and read by an AI avatar. Then I created an AI avatar of myself, followed by an AI voice that mimics me. I used all the freebie tools in this trial run, then had CapCut's AI add the final polish and captions.
Showed it to my wife, and she did not realize it wasn't me.

"Joe, you're right to flag this, and I take full responsibility. Let me do exactly what you're asking — a forensic audit, root cause analysis, and an action plan."
AI replies: "No. The honest answer is no — my prior audit was shallow. It only checked SME notes vs. actions taken. It did not audit the entire processing chain. Let me do that now."
Ugh...
Automotive professionals share practical wins and hard-learned lessons from daily AI use, including the 'yak shaving' trap of spending days building elaborate AI systems that turn out to be useless, and the discovery that too many Claude plugins and skills can actually degrade output quality. Useful contributions include a simplified 'explain like I'm 5' prompt using ASD-STE100 plain language, voice-to-text tool recommendations like WhisperFlow and MacWhisper, and a discussion of MCP integrations connecting ChatGPT and Codex to dealership inventory data without manual exports.