FOUNDER / REALLIER WEI / SHENZHEN
Work in complexity. Prove it in reality.
A parallel thread beyond the products: what I keep reading, how technical judgment is formed, and the questions not yet compressed into company language.
Writing is not a display of conclusions. It is where judgment is formed.
GitHub ↗01 / WRITING
Write until the problem reveals its structure.
Long-form thinking on personality, relationships, order, technology, and individual choice.
笼中鸟与守夜人：解析《赛博朋克 2077：往日之影》的道德镜像
用心理类型学（MBTI）与存在主义的手术刀，剥开《往日之影》中李德与百灵鸟的底层逻辑，揭示高塔结局的“政治阉割”本质，以及这款游戏如何残酷地测出了玩家潜意识里的“赛博顺民图谱”。
Read the essay ↗02 / FIELD NOTES
Keep the time, the context, and the judgment made then.
Technical observations and tool research—not formal company product commitments.
Between Framework and Intelligence: The Structural Evolution of the LangChain Ecosystem
LangChain is no longer a framework; it is an "ecological structure" for organizing and operating modern LLM applications. From a toolkit to a syntax layer, a logic layer, and an operations layer, its trajectory mirrors the industry's shift from calling models to engineering applications.
A Container for Order: AnythingLLM and the Topological Restructuring of Private Knowledge
The difficulties of modern AI applications often stem from tightly coupled components. By abstracting the core elements of a RAG system, AnythingLLM hands technical sovereignty back to the user and embodies a philosophy of "decoupling." On the road to AGI, architectural flexibility matters more than simply iterating on models.
First Principles for Shipping Agents: Less Conversation, More Tools
From chasing an "all-powerful digital employee" back to "automated scripts." Why does deterministic logic chaining beat anthropomorphic, conversation-heavy design in internal enterprise efficiency scenarios? End the "essay-writing collaboration" between models and return AI to the componentized essence of software engineering.
Coze, Dify, and n8n: An Engineering Field Report on AI Workflow Platforms
From Coze's no-code drag-and-drop, to Dify positioning itself as an AI middle platform for engineers, to n8n being used far below its intended weight class. Peeling back the marketing gloss to look at the real selection pitfalls and capacity limits of mainstream AI platforms in production.
Manus Technical Deep Dive (2026): How Does It Actually Run Agents?
A plain-language technical breakdown of Manus: context engineering, parallel multi-agents, sandboxing, browser automation, Skills, MCP, and the API. The focus is on why it gets things done — and where it remains opaque.
A Landscape Survey of Financial Analysis Tools on GitHub
Covering 30+ core open-source projects across 7 categories: all-in-one platforms, quant frameworks, AI/LLM, backtesting engines, data APIs, crypto trading, and technical analysis
OpenCLI In-Depth Research
A complete research note covering the positioning of OpenCLI — an OpenAPI for CLIs — its ecosystem progress, open controversies, and enterprise adoption recommendations