All Things Product with Teresa and Petra Bad Advice
13 snips
Mar 24, 2026 They dig into how public transcripts are used to build AI clones and why many outputs feel cringe or low quality. The conversation covers how past content lacks context and how mediocre AI responses can damage reputations. They probe IP, consent, and whether open-sourcing transcripts is ethical. The hosts also talk about ways creators can be supported as AI tools spread.
AI Snips
Chapters
Books
Transcript
Episode notes
Founders Felt Cloned By Bots Built From Transcripts
- Teresa and Petra discovered people had built chatbots claiming to be them from public podcast transcripts.
- Petra felt abused and cringed because others took her public words out of context to build tools that give poor advice.
Teresa Tested A TeresaBot And Rejected It
- Teresa tested a Delphi-based TeresaBot and saw conversations that she could view and that gave advice she'd never give.
- She refused to release a public clone because it used her voice and produced answers unlike what she'd actually say.
Average LLM Outputs Are Not Expert Thinking
- Average LLM outputs mimic past public statements and create an amalgamation rather than true expert thinking.
- Petra notes expertise includes evolving judgment and context that two-year-old transcripts can't capture or reproduce.
