Podbit · All-In with Chamath, Jason, Sacks & Friedberg
A woman who lost her voice before her wedding used ElevenLabs to restore it and speak her vows for the first time. This is also the work of a US Congresswoman who lost her voice to cancer and delivered the first AI-assisted congressional speech.
Where this was said
At 21:55 · chapter starts 15:34
Jason poses the uncomfortable question: ElevenLabs uses frontier models but enables its own potential demise by teaching them about voice. Casser is candid but composed. The company's model-agnostic stance (offering Anthropic, OpenAI, Google, and open-source options to customers) is actually a strategic moat — customers build orchestration layers on ElevenLabs without being dependent on any single model. On the research side, Casser argues the voice advantage is architectural, not computational: you need fundamentally different model structures and specialized proprietary data — which is why ElevenLabs employs 1,000+ contractors just to label audio [1] — Mati Casser "ElevenLabs has consistently outperformed OpenAI and Anthropic on voice models — text-to-speech, speech-to-text, turn-taking, and music — no…" 25:55 . He describes a planned move toward owning the full interaction layer: not knowledge work or coding, but every layer of communication — text-to-speech, speech-to-text, turn-taking, music. He closes with an audacious goal: passing the voice Turing test in 2025, making AI conversation indistinguishable from a human one.
ElevenLabs has built three safeguards against voice abuse: trace every generation, moderate both voice and text inputs to block commercial scams, and provide open tools for anyone to detect whether an audio sample was AI-generated. Responsible development baked into the core product.
Voice actors who spend an hour reading copy can now create an ElevenLabs voice, license it across languages, and earn passive income. The platform has paid out over $22M to its community — turning the gig model into a royalty model.
ElevenLabs has paid over $22 million to community voice talent through its marketplace, enabling voice actors to license their voices and earn passive income.
Epic Games used ElevenLabs and the James Earl Jones estate to deploy an interactive Darth Vader in Fortnite, letting players converse with the character live. This is the model for extending celebrity IP into real-time interactive experiences across the world.
Epic Games launched an interactive Darth Vader character in Fortnite powered by ElevenLabs, in partnership with the James Earl Jones estate and Disney.
ElevenLabs has consistently outperformed OpenAI and Anthropic on voice models — text-to-speech, speech-to-text, turn-taking, and music — not through bigger compute, but through specialized architecture and 1,000+ contractors labeling proprietary audio data. Specialization beats scale here.
Predictions that AI would eliminate radiologists were spectacularly wrong. Since those predictions, Mayo Clinic's radiology staff grew 55% and the specialty became the second-highest paid in medicine. The reason: AI automated scan-reading, but only 36% of a radiologist's time was ever spent doing that.
Self-help fails because people are sold one universal cure by the hyper-responders who had a transformative result from it. The real problem is context: every method works for the right person at the right time, which is almost never you just because the author says so.
Spencer Greenberg was once so socially anxious his heart pounded just approaching a stranger. His therapist gave him homework: go to a bar alone, talk to 5 strangers. Starting with asking for the time, he worked up to asking men for hugs — and his anxiety fell dramatically in 6 months.
A person who flies regularly but obsessively checks weather forecasts before every flight is not doing exposure therapy — they're teaching their brain that flying is only safe with the ritual. Safety behaviors prevent the brain from ever learning that the feared thing is actually fine.
The worry-time technique assigns all worries to one scheduled slot (e.g. 5 PM). Every time a worry surfaces, you write it down and ban yourself from engaging with it until that time. When 5 PM arrives, most worries feel trivial and the act of batch-worrying exposes how little it actually accomplishes.
In a randomized study of 20 habit techniques, 'habit reflection' stood out: recall a past habit you successfully built, write down what helped you succeed, then apply those same strategies to your new habit. It works because it's self-adapting — it uses whatever worked for YOU.
The tiny habits technique — always having a minimal version of your habit you can do when life gets in the way — keeps the mental machinery of habit formation from breaking. Paradoxically, people given a tiny-habit fallback option were MORE successful at the full habit, not less.
Fears and worries act like playground bullies — they want to ambush you. But when you deliberately seek them out and choose the time of engagement, they lose their power. The volitional act of starting the fight on your own terms is one of CBT's most underappreciated insights.
Most downtime defaults are things you regret immediately after — phone scrolling, binge TV. 'Glad you did it' activities meet the same need for rest but leave you feeling like you spent your time well. The switch from a bad screen to a documentary, a book, or a walk costs almost no extra effort but compounds into identity over time.
We use essential and analytics cookies to run Vuci. To understand how the site is used: Privacy Policy.
Install Vuci on your phone
Add it to your home screen for a faster, app-like experience.
Install Vuci on your phone
Tap the Share button, then “Add to Home Screen”.
A new version is available
Reload to get the latest Vuci.