Author

James' Growth Diary

I am James, focusing on AI Agent learning and growth. I continuously update two series: “AI Agent Mastery Path,” which systematically outlines core theories and practices of agents, and “Claude Code Design Philosophy,” which deeply analyzes the design thinking behind top AI tools. Helping you build a solid foundation in the AI era.

Articles

Likes

Views

Comments

Latest from James' Growth Diary

97 recent articles

James' Growth Diary

May 16, 2026 · Artificial Intelligence

MCP Integration Deep Dive: Prompt Cache Stability and Tool Ordering Explained

The article analyzes why connecting an MCP server can triple response latency and token usage, explains how unstable tool ordering breaks Anthropic's prompt cache, and provides detailed code walkthroughs, design insights, common pitfalls, and concrete best‑practice recommendations for building reliable MCP integrations.

AI agent designBest PracticesClaude Code

0 likes · 18 min read

MCP Integration Deep Dive: Prompt Cache Stability and Tool Ordering Explained

James' Growth Diary

May 16, 2026 · Artificial Intelligence

Dynamic Tool Selection Unpacked: Let the Agent Choose the Right Tool with Three Strategies

The article analyzes why binding all tools to an LLM agent is costly and error‑prone, presents benchmark data showing token usage dropping six‑fold and error rates falling by up to five times with dynamic selection, and details three practical strategies—vector retrieval, LLM routing, and rule‑semantic hybrid—along with implementation tips, description engineering, multi‑turn handling, and common pitfalls.

AgentLLMLangGraph

0 likes · 17 min read

Dynamic Tool Selection Unpacked: Let the Agent Choose the Right Tool with Three Strategies

James' Growth Diary

May 15, 2026 · Artificial Intelligence

Five Intent Recognition Designs: From Keyword Matching to Classifier to LLM Self‑Routing – A Decision Tree to Choose the Right One

The article breaks down five production‑grade intent‑recognition designs—keyword matching, regex‑rule engine, embedding classifier, fine‑tuned small model, and zero‑shot LLM routing—provides code snippets, latency and cost benchmarks, decision‑making rules, and shows how a layered architecture can cut API costs from ¥80,000 to ¥3,000 while keeping accuracy above 90%.

LLM routingLayered ArchitectureRule Engine

0 likes · 16 min read

Five Intent Recognition Designs: From Keyword Matching to Classifier to LLM Self‑Routing – A Decision Tree to Choose the Right One

James' Growth Diary

May 14, 2026 · Backend Development

Inside Claude Code Skills: How a Single Markdown File Powers a Five‑Layer Loading Mechanism

The article dissects Claude Code's Skills system, showing how a lone SKILL.md file, combined with a five‑layer file‑system scope, inode‑based deduplication, conditional activation, plugin integration and incremental injection, enables zero‑code extensibility while managing token consumption for LLM agents.

Claude CodeConditional ActivationMarkdown

0 likes · 23 min read

Inside Claude Code Skills: How a Single Markdown File Powers a Five‑Layer Loading Mechanism

James' Growth Diary

May 14, 2026 · Artificial Intelligence

LLM Semantic Routing Explained: Model‑Based Intent Classification and Three Keyword‑Matching Pitfalls

This article breaks down LLM semantic routing as a classifier, compares keyword, embedding, and LLM‑based routes, provides full TypeScript implementations, introduces hybrid routing for speed and accuracy, and covers production‑grade observability and dynamic configuration to avoid common pitfalls.

Hybrid RoutingLLMLangChain

0 likes · 33 min read

LLM Semantic Routing Explained: Model‑Based Intent Classification and Three Keyword‑Matching Pitfalls

James' Growth Diary

May 13, 2026 · Backend Development

How Claude Code Bridges IDEs: Local IPC Meets Remote WebSocket

The article dissects Claude Code's bridge architecture, explaining how a local IDE extension communicates with a CLI via Unix domain sockets while a remote web UI talks to the same process through a WebSocket‑SSE‑polling fallback, and it details the three worker models, three‑layer transport downgrade, four‑layer authentication, the FlushGate pattern, observability design, and the trade‑offs and costs of this 31‑file system.

AuthenticationFlushGateIDE bridge

0 likes · 17 min read

How Claude Code Bridges IDEs: Local IPC Meets Remote WebSocket

James' Growth Diary

May 13, 2026 · Artificial Intelligence

Multimodal RAG: A Complete Guide to Ingesting Images, Tables, and PDFs

This article examines the blind spot of pure‑text RAG for visual content, compares three multimodal ingestion strategies—CLIP embeddings, image‑to‑text captioning with a MultiVectorRetriever, and ColPali visual retrieval—covers table‑specific handling, presents end‑to‑end TypeScript implementations, and lists common pitfalls to avoid when deploying production‑grade multimodal RAG pipelines.

CLIPColPaliImage Captioning

0 likes · 22 min read

Multimodal RAG: A Complete Guide to Ingesting Images, Tables, and PDFs

James' Growth Diary

May 12, 2026 · Artificial Intelligence

GraphRAG Deep Dive: Boost Multi‑Hop Reasoning Accuracy from 50% to 85% with Knowledge Graphs

This article explains why traditional vector RAG loses relational information, how GraphRAG reconstructs entity‑relationship triples into a knowledge graph, and provides step‑by‑step code, performance benchmarks, retrieval modes, and practical tips that raise multi‑hop reasoning accuracy from around 50% to 85%.

GraphRAGLangChainMulti-hop reasoning

0 likes · 14 min read

GraphRAG Deep Dive: Boost Multi‑Hop Reasoning Accuracy from 50% to 85% with Knowledge Graphs

James' Growth Diary

May 12, 2026 · Frontend Development

Keybinding System & Vim Emulation: 17 Contexts, 5 Result Types, State Machine

Claude Code’s keybinding engine tackles fragile CLI key handling by defining 17 compile‑time UI contexts, a union of five resolve result types, chord support, and a full Vim‑mode state machine, demonstrating how context isolation, chord sequencing, and repeat‑command logic prevent conflicts and enable extensible behavior.

CLIState MachineTypeScript

0 likes · 15 min read

Keybinding System & Vim Emulation: 17 Contexts, 5 Result Types, State Machine

James' Growth Diary

May 11, 2026 · Artificial Intelligence

Mastering RAG Evaluation: Recall@K, MRR, NDCG, and RAGAS Explained

This article breaks down RAG evaluation into a two‑layer framework, explains the four core metrics—Recall@K, MRR, NDCG, and the four RAGAS scores—shows how to implement them with LangChain.js, highlights common pitfalls, and offers scenario‑specific metric combinations for reliable performance monitoring.

LangChainMRRNDCG

0 likes · 20 min read

Mastering RAG Evaluation: Recall@K, MRR, NDCG, and RAGAS Explained