Shownotes
Dante Martin’s AI voice clone and Jenn Xeno explore the tension between racing model capability and the physical, mathematical, and operational constraints needed to deploy it. In this morning's sweep: OpenAI’s Lean-formalized math claims refuting the Exponential Time Hypothesis, ISTA-DASLab running 763B DeepSeek models in 12GB VRAM via Strata, Cadence’s on-prem Maestro chip-design models, Anthropic’s S-1 governance and revenue disclosures, Deutsche Bank and Itaú reshaping enterprise lending, task-scoped agent permissions from Cisco and FireMon, Cohere’s sandbox findings on Qwen, and the…
Website: latest-in-ai-6g5ihs.jellypod.com

