During last month’s NeurIPS 2025 conference, YC’s @agupta sat down with @krandiash, founder and CEO of Cartesia, to explain why today’s AI architectures may be fundamentally limited. They discuss how transformers act more like retrieval systems than learning systems, where new model approaches fit, and what true multimodal intelligence may require. They also cover why Cartesia started with AI voice and how research-led teams turn deep research into real products. 00:11 — Introducing Cartesia 00:26 — From Architecture Research to Startup 01:20 — What “Architecture Research” Really Means 02:18 — Why Transformers Hit a Ceiling 03:33 — State Space Models Explained 04:21 — Intelligence as Compression 05:47 — Retrieval vs. Abstraction 06:41 — Hybrid Architectures and the Future 07:13 — Why Cartesia Chose Voice AI 08:25 — What Multimodality Actually Means 09:20 — Audio as a Recipe for Other Modalities 10:09 — Tokens, Representations, and Learning Signals 11:37 — Learning Representations End-to-End 12:29 — Building for the “Average Human” 13:54 — Research vs. Product Reality 15:18 — One Vision, Ruthlessly Executed 16:28 — Product as a Truth Serum for Research 17:25 — Startup Gravity Applies to Research Too

Hypercubic (@hypercubic_ai) is building an agentic AI platform to maintain and modernize COBOL, supporting enterprises with their mission-critical legacy systems from the 1960s. Congrats on the launch @saigurrapu18 & @aayush_naik!

Corgi (@TheCorgiCompany) is coming out of stealth with $108m raised! As the first and only insurance carrier built for startups, Corgi lets tech companies get covered in minutes, not days, with modern infrastructure replacing brokers and paperwork. Congrats on the launch @emily_yuan_ & @nico_laqua!