As code generation gets cheaper, the real battleground in AI-native engineering is shifting toward durable execution, externalized state, sandbox boundaries, recovery semantics, and verification.
Hi, I'm xieydd
Personal blog by xieydd covering AI infrastructure, Kubernetes, PostgreSQL, vector search, model serving, and cloud architecture.
Featured Writing
A practical reflection on building with Code Agent, covering workflow changes, mistakes, wins, and what actually improved my output.
Why has systems innovation around large language model inference increasingly looked less like optimizing a neural network and more like designing a memory system? This article strings together Mac/UMA, NVIDIA/NVLink, FlashAttention, vLLM, RL rollout, Mooncake, LMCache, CXL, and recent 2025-2026 research papers into a coherent narrative.
How VectorChord brings low-cost billion-scale vector search to PostgreSQL, with IVF+RaBitQ, HNSW tradeoffs, and practical RAG guidance.