The author describes building DxC, an Agent-Native WeChat writing system, explaining why they chose a Skill+CLI architecture over a complete Web prototype, and how removing forms shifted determinism burdens to the agent architecture, causing issues like skipped steps, parameter errors, and idempotency risks.
This article analyzes why LLM inference often yields nondeterministic results, explains how floating‑point addition order, GEMM tiling, Split‑K, RMSNorm, FlashAttention, and NCCL all contribute to batch variance, and details the engineering steps vLLM takes to enforce batch‑invariant execution across GPUs.
The article analyzes Claude's natural‑language SKILL.md approach, highlights its flexibility and nondeterminism, and explains how adding a declarative skillflow.json graph enforces deterministic execution, auditability, lower token cost, and better consistency for high‑frequency, compliance‑critical tasks.
Martin Fowler argues that software engineering has moved from a deterministic world to a nondeterministic one driven by LLMs, outlining how this paradigm shift reshapes development practices, introduces new risks, and demands a harness‑based engineering approach to manage uncertainty.
This article examines whether the world is governed by chance or necessity, discussing quantum uncertainty, classical determinism, the dual role of knowledge, and how mathematics, probability, AI, and big‑data analytics together shape our ability to understand and forecast complex systems.
