Episode Details

## Episode Summary In this episode, we cover: - **FLINT: Efficiently Leveraging High Bandwidth Flash for Capacity-Scalable LLM Inference Acceleration** (arXiv) - **Squeezing the Cache, Preserving the Truth: Monotonic Equipotential Allocation with Geodesia-KV** (arXiv) - **How Cache Coherency Simplifies AI Software** (semiengineering) - **Arm Reveals AGI Server CPU Architecture at Hot Chips, Targeting Agentic AI Workloads - finance.biggo.com** (google_arch) - **Security Researchers Find Current RISC-V CPU Implementations Coming Up Short - Phoronix** (google_riscv) --- *Sponsored by Ada, Ago Consulting, and Zen Semiconductor*