AI Development
9 posts in this series
A broad AI development track for app building, API integration, model capabilities, automation, and practical implementation patterns.
AI-Powered Refactoring of 10,000 Lines: A Real Story of Doing a Month's Work in 2 Weeks
A complete retrospective of using Claude Code to refactor 10,000 lines of legacy Vue code. From inheriting a 3-year-old codebase nobody dared touch, to shipping in 2 weeks with zero incidents. Includes complete Prompt templates for test generation, code diagnosis, refactoring execution, and pitfall avoidance.
Multimodal AI Application Development Guide: From Model Selection to Production Deployment
A comprehensive guide to multimodal AI application development, covering mainstream model comparisons (GPT-4V, Claude Vision, Gemini), practical code for image/video/document processing, and best practices for cost optimization and deployment
Multimodal AI Application Development: A Complete Guide to Three-Modal Fusion
Compare GPT-4V, Gemini, and Claude platforms with complete code examples for text, image, and audio fusion. Learn system architecture design principles and cost control techniques to master multimodal development core skills.
LangChain LCEL in Practice: From Legacy Chains to Streaming Responses - A Modern Paradigm
LCEL restructures LangChain development with the | operator, reducing code by 70% and adding automatic streaming support. Deep dive into Pipe mechanics, Runnable interface, and streaming with a practical migration guide.
LLM Evaluation Framework Comparison: LangSmith vs W&B vs MLflow
An in-depth comparison of three major LLM evaluation frameworks—LangSmith, Weights & Biases, and MLflow—covering tracing capabilities, evaluation methods, production deployment, and real costs to help you make the best choice.
Self-Evolving AI: Key Technical Paths for Continuous Model Learning
Deep dive into the four technical paths of self-evolving AI: model-level evolution, context evolution, meta-learning, and architecture evolution, exploring AI's journey from static knowledge to dynamic growth
Self-Evolving AI: 4 Methods for Continual Learning in 2026
A deep dive into 2026 continual learning trends—from SDFT self-distillation to MiniMax M2.7's self-evolution pipeline. Exploring 4 methods for models that learn while they use, with practical insights from the LangChain three-layer evolution framework.
LangGraph vs AutoGen State Tracking Comparison: Checkpoints, Timeout Recovery, and Framework Selection
In-depth comparison of LangGraph vs AutoGen state tracking: 12-dimension quantitative comparison covering checkpoint mechanisms, timeout recovery, distributed support, with real-world pitfalls, decision trees, and runnable code to help you choose the right framework.
Hyper Company Brain: Engineering an AI Agent Knowledge Base
Using Hyper's public architecture details, this guide breaks down how an AI agent company knowledge base should handle fact validity, permissions, decision trails, hybrid retrieval, hooks/MCP injection, and human correction.
