About
📌 Technical Origin of OpenML - A Quantum Physics Problem
The conceptual foundation of OpenML dates all the way back to 2013 when a college boy in Physics major was
reading and implementing the Matrix Numerov Method for Solving Schrödinger Equation
2452 words
|
12 minutes
从吴思的《潜规则》看中国职场
2026-11-18
正式条文是明规则,大家私下默认的利益玩法就是潜规则。书中“合法伤害权”这个概念,直接对应职场:领导不需要硬整人,只需要在考核、资源分配上拿捏尺度,
就能控制下属。完美解释很多制度解决不了的办公室博弈。
1877 words
|
9 minutes
Practical BM25 - Part 1: How Shards Affect Relevance Scoring in Elasticsearch
2026-11-17
Elasticsearch (ES) has shifted from a traditional keyword/lexical search engine into a core foundational layer for
Retrieval-Augmented Generation (RAG), hybrid search, and AI agents. This post discusses practical BM25 about
similarity ranking (relevancy).
219 words
|
1 minute
AI Architect
Traditional software architecture is built on deterministic components (if A, then B). Large Language Models introduce
a probabilistic component into the middle of the system. The defining job of an AI Architect is to wrap a stochastic,
unpredictable engine inside deterministic boundaries so that the overall system is reliable, auditable, cost-bounded,
and safe.
1354 words
|
7 minutes
Process and Reality
A deep dive into Whitehead's Process and Reality, replacing classical substance ontology with dynamic occasions of
experience, relational prehensions, and the metaphysical foundations of quantum mechanics.
793 words
|
4 minutes
Agent Workflow
The post "软件行业的 AI 思维" ignited my to study CrewAI whose documentation starts with "Flow", which is why this post
was initiated to help me systematically study a concept I had been encountering a lot of times
2260 words
|
11 minutes
RAG & Vector Database
The conceptual foundations of RAG and its backing vector database rests on decades of research into Approximate
Nearest Neighbor (ANN) search, designed to solve the "curse of dimensionality" where exact nearest-neighbor searches
degrade to linear scans. As deep learning models like Word2Vec, BERT, and convolutional neural networks began mapping
text, images, and audio into dense mathematical embeddings, engineering teams required ways to search these spaces
efficiently. This produced the first wave of standalone vector search libraries in the mid-2010s. Around 2019, the
industry shifted toward purpose-built systems, popularizing "vector database" as a distinct database category.
3336 words
|
17 minutes
《人情与面子 - 中国人的权利游戏》
2026-11-09
帮助我们理解中国职场行为逻辑的源头,学术但不晦涩。讲清楚了中国人社交里:面子、人情、关系、回报这套体系到底怎样运转
472 words
|
2 minutes