Voice-to-voice AI models have an 8x higher hallucination rate than text
Jesse Zhang · The Future of AI Agents | Jesse Zhang Interview · Invest Like The Best · Oct 6, 2025 · at 35:49
Jesse Zhang, CEO of Decagon AI, explains the technical hurdles of voice-to-voice models, noting that higher token counts per sentence elevate error rates.
0:00 / 0:02exact quote · 2.3s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“I think probably like eight X higher or something like that.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Jesse Zhang
Fears of AI customer support layoffs are overblown due to natural attrition
“I think it's a little bit overblown of like, oh, AI is replacing jobs and so on. I think most of, even the BPO as we talked to not really that concerned because what typically happens anyways, is that there's already very high turnover in these, BPO's cause li…”
People falsely claim to be Decagon customers on investor expert network calls
“We have people who had said that they've used us, and we literally never heard of them.”
Zhang: AI application startups should ignore margins and prioritize market share
“Like right now, I would argue that if you're building an application, you're Margin shouldn't really matter that much. Like, people will often sort of critique the coding agents like they're hemorrhaging money. Yeah, but, you know, again, things are improving …”
Frontier AI labs will move toward applications due to API commoditization
“That's why I think the open AI's of the world will continue to move towards applications because it's quite hard for them to make money longterm on like their API, for example, because There's like such high competition, all the labs are building, and it's lik…”
Forward-deployed engineering models require a minimum customer deal size of $1 million
“Oh, like probably a million.”
Experienced founders warned Zhang that AI customer support was too obvious
“Essentially what everyone told us, including like, you know, very smart, older founders that we knew and so on was that yeah, this is like excuse case, like super obvious. There's probably just going to be incumbents that are just tacking onto the product. And…”