Q&A with ElevenLabs co-founder Mati Staniszewski on how audio models work, the company's business model, the conversational Turing Test, voice agents, and more (John Collison/Cheeky Pint)
THE SO WHAT
Voice is graduating from feature to substrate — ElevenLabs talking openly about a “conversational Turing Test” is a statement that synthetic audio will sit between users and most software. If your product assumes text as the primary interface, you’re already behind the interaction curve.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIHow Legora hired Jude Law
Celebrity alignment is becoming a go-to wedge for consumer AI brands trying to differentiate in a sea of similar capabilities. If you're building in consumer AI, your moat is narrative, IP, and distribution — not model access — so budget and hire accordingly.
Applied AI'Trusted access for the next era of cyber defense': OpenAI reveals its Mythos rival, designed for cybersecurity pros to spot the next level of attacks
Dedicated GPT-5.4-Cyber means frontline defense is shifting from dashboards to model-native workflows — the SOC is becoming an LLM-first environment. If you're a security vendor, assume your UI is a thin layer over someone else's model and start deciding whether you are data, workflow, or infra in that stack.
Applied AIWhat if World Models and Quantum Computing complemented LLMs?
The 'AI of everything' narrative — LLMs plus world models plus quantum — is a reminder that today's architectures are interim, not end-state. Don't over-optimize org structure or infra contracts around a single paradigm; design for swap-ability at the model and accelerator layer.
Applied AIByteDance launches its Seedance 2.0 video model to enterprise clients in 100+ countries, excluding the US amid legal disputes, after a February launch in China (Juro Osawa/The Information)
Seedance 2.0 going to 100+ countries — with the US carved out — shows AI video is globalizing while regulatory blocs harden. If you're an enterprise content or marketing leader, you now have serious non-US options for video generation and need a jurisdiction-by-jurisdiction risk map, not a single global vendor.