concept
An AI system's confident output of things that are not true, here above all quotations nobody said. In episode 01 it is the failure the whole pipeline is built to catch, from Gemini's invented Krishnamurti quotes to the telephone effect of models quoting models. Start at about 28 minutes.
Across the episodes
01
Enters as the failure the pipeline is built around: invented quotations delivered with full confidence
Names what makes the failure dangerous: the errors come with no signal of uncertainty, so they cannot be caught by reading for hesitation.
"They're confidently wrong when they're wrong. They don't warn you that they guessed."Chad Nantais, Syntopic, Episode 1 (11 Oct 2026), 30:10
Restates the test every summary of a conversation has to pass, which is the question provenance is meant to answer.
"Is this truly what was said? Or is it distorted by AI hallucinations?"Chad Nantais, Syntopic, Episode 1 (11 Oct 2026), 37:17
Extends the risk from one bad quotation to a chain: as models summarize models, an error compounds the way it does in the children's game.
"sometimes you get a situation where one llm is quoting another llm and if it's wrong you get the telephone effect you know the telephone game"Chad Nantais, Syntopic, Episode 1 (11 Oct 2026), 38:35
Connections
exampleGemini
Invented quotations from Krishnamurti.