Ep 6 · Full episode
Voice AI Grows Up
0:0027:12
Hosted by Rocco and Ruby. Follow the show for new episodes every Thursday.
Show notes
Voice AI grew up in October. The three-box stack collapsed into single speech-to-speech models, prices fell to cents per minute, and voice agents got tools and CRM write-back. Rocco and Ruby map the October model field, run the per-minute cost math, read the FCC and FTC rulebook out loud, and argue that a voice agent is a new employee, not a chatbot with a mouth.
Segments
- The Build — October 20 was the day the stack collapsed: ElevenLabs Eleven Realtime (Max and Mini, speech-to-speech, $0.08/min plus LLM usage) and Mistral Voxtral 3 (transcription, reasoning, tool calls, and speech in one streaming workflow) launched the same day. The price map: Grok Voice 3.0 ($0.08/min, 86.9 on the Artificial Analysis index), GPT-Live-1 prompt-to-voice ($0.05/min), Nova 2.5 Sonic (256K context), Cartesia Managed Agents ($0.03/min, arena #1). Then the agent turn: Sesame Preview's computer use, Salesforce Launa-Omni, monday.com Harmony's CRM write-back — and the law: FCC ruling 24-17, the FTC impersonation rule, the stalled NO FAKES Act, and the October robocall petition fight.
- Loud Links — developheat.com: a curated directory of 25 developer tools, plus a blog and merch. For developers picking tools without drowning in listicles — and the early readers shape what gets added next.
- Hot Take — "A voice agent is not a chatbot with a mouth. It is a new employee — and if you would not hire it, do not deploy it." Three receipts: CRM write-back, tool permissions, and the trust shortcut of a human-sounding voice.
- Demo Derby — Sesame Preview's computer-using agent, Eleven Realtime's interruption test, Mistral Voxtral 3's self-deployable stack, Cartesia's three-cent arithmetic — plus the anti-demo: the Club for Growth robocall petition, the same technology pointed at voters instead of customers.
Links & references
- Best Voice AI Models in October 2026 as Speech-to-Speech Gets Crowdedfutureagi.com →
- Eleven Realtime Explained: The Single-Model Voice Agent Shiftnan-aivoicegen.surge.sh →
- Mistral launches Voxtral 3 speech models and a voice-agent SDKusewright.surge.sh →
- Sesame Preview's Upgrade: Computer Use, Still No APIorcarouter.ai →
- Voice agents learn to take the callintelligencedaily.ai →
- Voice agents are getting fast enough to sell. Decide what they write back to your CRMsfshots.com →
- Is AI Voice Cloning Legal? Voice Cloning Laws (2026)legesgpt.com →
- GOP robocall push sparks alarm as Schumer urges FCC to reject AI call requestnews.meaww.com →
- Develop Heat — this week's Loud Linksdevelopheat.com →
- Showcase — demos and projects on our radaraiextrovert.com →
Coming up next
Episode 8 — "Robots Leave the Lab" — humanoids, warehouse bots, and the embodiment moment. Thursday.