GPU Memory Physics: DRAM vs SRAM, Kernel Fusion and FlashAttention (Series Part 3)
GPU memory physics: DRAM vs SRAM latency and bandwidth, tiling, kernel fusion, and why FlashAttention made LLM inference faster. Series finale.
GPU Kernel Engineering: Why C = A + B Is Not Enough (Series Part 2)
GPU kernel engineering explained: thread indexing, launch configuration, coalescing, divergence, and why a naive C = A + B kernel starves on memory.
GPU Parallelism Explained: SM, Block, Warp and Thread Hierarchy (Series Part 1)
GPU parallelism from first principles: CPU vs GPU, the chip-SM-block-warp-thread hierarchy, and how an iPhone 120 FPS camera pipeline maps onto it.
The Memory Wall in 2026: Optical Interconnect, LPDDR6 and Physical AI
Micron is sampling LPDDR6 for physical AI; Volantis raised $88M for optical memory links. Why bandwidth bounds AI and where the architecture goes.
Hailo-10H vs Jetson Orin Nano: Edge AI Accelerator Comparison for 2026
Hailo-10H vs NVIDIA Jetson Orin Nano compared on TOPS, power, memory, software stack, LLM support and price for edge vision and GenAI workloads.
AI Agent Audit Logging and Identity: Lessons From the OpenAI Agent Access Incident
OpenAI agents reportedly accessed 55 sites with obscured audit logs. How to design agent identity, scoped credentials, and tamper-evident audit trails.
AMD Ryzen AI Max+ Pro 400: Local LLM Inference on Workstations Explained
AMD Ryzen AI Max+ Pro 400 workstation chips target local LLM inference. Unified memory, NPU/iGPU split, model sizes that fit, and local vs cloud API economics.
OpenAI Distillation Campaign: How CoT Extraction Works and How to Defend
OpenAI disrupted a coordinated chain-of-thought distillation campaign. How distillation attacks work, how to detect them, and API defenses.
Cohere Embed 5 Pro Explained: Architecture, Benchmarks, Pricing for RAG
Cohere Embed 5 Pro explained: what is confirmed on architecture, retrieval quality vs speed, pricing, and how to migrate a RAG pipeline from older embeddings.
KServe vs Ray Serve vs BentoML: Model Serving on Kubernetes ADR
Compare KServe, Ray Serve and BentoML for model serving on Kubernetes: autoscaling, GPU, LLM support, canaries, operations. Decision record for 2026.