orbsh/wiki

This repository is a personal knowledge base written mostly in Chinese, organized as a wiki index of notes and design documents. The README is mostly a table of contents that links out to dozens of individual markdown files grouped under topic headings, with a few entries also offered in an English translation alongside the Chinese original. The topics covered span a wide range of the author's technical interests. There is a section on memory systems for AI agents, covering ideas like graph based memory, layered memory architecture, and how persistent memory changes what an AI tool can do over time. Another section covers AI agent architecture and usage patterns, including design principles, agent selection research, and critiques of multi-agent approaches. A large section is devoted to language models and cognition, discussing topics such as prompt caching behavior, hidden behavior patterns in model outputs, and cognitive psychology as applied to agent design. Other sections gather notes on infrastructure and architecture choices, such as comparisons between database and storage technologies, and a set of pointed critique documents examining tools like Redis, Nginx, MySQL, and cloud platforms from the author's own technical perspective. There are further sections on programming language design, workflow and skill automation, broader thinking about AI era business models and organizations, and notes on editors, operating systems, and networking tools the author has evaluated. A final section links to documentation for the author's own software projects, including a UI framework, a NixOS configuration, and a distributed systems design. Because the README is structured entirely as a linked index rather than prose, this project functions as a personal reference library or digital notebook rather than a piece of software with an install process. It does not state a license, and most of the linked content is not included in the excerpt available here, so this description reflects only the topics and structure the index itself reveals.