Deep Dives
247 articles
Deep Dives·4 min
Agent Loop Core Mechanism Explained: Building a ReAct Loop from Scratch (With Debugging War Stories)
Deep dive into Agent Loop, the core mechanism of AI Agents, based on ReAct architecture. Covers Tool Call ID handling, context management, assistant message feedback, with full Python code and common bug fixes.
Read more →
Deep Dives·2 min
MCP Protocol Architecture Deep Dive: Host-Client-Server Three-Layer Design and Three Core Primitives
Deep dive into MCP's Host-Client-Server three-layer architecture and its Resources, Tools, Prompts primitives. Learn how it transforms AI tool integration from M×N to M+N.
Read more →
Deep Dives·1 min
Ralph Loop: The New AI Programming Paradigm That Beats Multi-Agent Systems Through Iterative Trial and Error
Ralph Loop is a 2025 AI programming paradigm: a memoryless AI Agent repeatedly reads the same Prompt, accumulates experience via file systems, beating multi-agent architectures through iterative loops.
Read more →
Deep Dives·3 min
Should AI Output Markdown or HTML? Two Dimensions to Help You Choose Quickly
Should AI output Markdown or HTML? This article provides a clear format selection framework using two dimensions—content maintenance lifecycle and reading comprehension efficiency—with practical scenarios and a comparison table.
Read more →
Deep Dives·3 min
What Are LLMs, Skills, Agents, and OpenClaude? A Complete Guide to Core AI Concepts
A plain-English guide to LLMs, Skills, AI Agents, and OpenClaude — clarifying how AI evolved from "can talk" to "can act" to "can act autonomously," with a comparison table.
Read more →
Deep Dives·1 min
ChatGPT Financial Management Deep Dive: Opportunities and Risks Behind Connecting Your Bank Accounts
OpenAI launches ChatGPT personal finance tools connecting bank accounts via Plaid for conversational money management. Deep analysis of features, strategy, privacy risks, and industry impact.
Read more →
Deep Dives·2 min
The Three Stages of AI Training: Pre-training, Post-training, and Alignment Explained
Explore the three key stages of LLM training: pre-training learns language, post-training learns tasks, alignment learns boundaries. Understand why AI hallucinates.
Read more →
Deep Dives·1 min
Beginner's Guide to AI Agents: Core Principles, Technical Architecture, and Application Scenarios Explained
A systematic guide to AI Agents covering core principles of perception, decision-making, and action, plus the technical stack of LLMs, tool calling, memory systems, and RAG, with real-world use cases.
Read more →
Deep Dives·1 min
TradingAgents-Astock: In-Depth Analysis of an Open-Source Multi-Agent Investment Research Framework for China's A-Share Market
In-depth analysis of TradingAgents-Astock, an open-source multi-Agent investment research framework for China's A-shares with 7 analysts, quality gates, and zero-cost data sources.
Read more →
Deep Dives·2 min
LangChain Core Architecture Explained: A Practical Guide to Its Five Key Modules
Deep dive into LangChain's five core modules: Model I/O, Data Connection & Retrieval, Chains, Agents, and Memory. Learn RAG workflows, Agent decision-making, and practical AI app development.
Read more →
Deep Dives·2 min
DeepSeek V3.2 Deep Dive: Three Major Breakthroughs in DSA Sparse Attention, Reinforcement Learning, and Agent Capabilities
Deep analysis of DeepSeek V3.2 and V3.2 Special: DSA sparse attention for faster long-context processing, RL compute at 10% of pre-training, and Agent task synthesis across 1,800 environments.
Read more →
Deep Dives·2 min
Hermes Agent: Complete Breakdown of 87 Built-in Skills — 7-Layer Architecture & Priority Guide
Deep dive into Hermes Agent V0.13.0's 87 Built-in Skills reorganized into 7 functional layers with priority guidance. Find the 17 core Skills worth learning first.
Read more →
Deep Dives·3 min
The Essence of AI Alignment: Aligning What to Do, Not How to Do It
The core of AI alignment is aligning What to do, not How to do. Through an Alembic database migration case, learn how Harness engineering crystallizes dev standards into reusable assets for automated programming.
Read more →
Deep Dives·4 min
What Is Playground Skill? How HTML Interaction Solves the Pain Points of AI Coding Communication
How Playground Skill uses HTML interactive interfaces to solve the pain points of natural language limitations in AI coding. Comparing traditional text-only modes with the new two-way interaction paradigm for Coding Agents.
Read more →
Deep Dives·4 min
AI Agent Skill Architecture Breakdown: Decision Layer + Encapsulation Layer + Execution Layer Practical Guide
Deep breakdown of AI Agent Skill's three-layer architecture: decision layer dynamic routing, encapsulation layer SOP standardization, and execution layer state machine self-healing, with practical solutions for context explosion, error recovery, and generality balance.
Read more →
Deep Dives·4 min
Will Frequent Git Commits Fill Up Your Hard Drive? Delta Compression & Packfile Mechanism Explained
Will frequent Git commits fill your hard drive? Deep dive into Git's Delta compression and Packfile mechanism, revealing why pure code commits cost almost nothing and identifying the real culprit — large binary files.
Read more →
Deep Dives·4 min
How to Read AUC Area Charts: A Complete Guide to ROC Curve Interpretation and Model Evaluation
Learn how to read AUC area charts with this complete guide covering ROC curve interpretation, AUC scoring standards (0.5–1.0), three model comparison scenarios, and PR curve vs ROC curve differences.
Read more →
Deep Dives·4 min
Confusion Matrix Explained: Precision, Recall, and F1 Score Formulas with Practical Applications
Learn confusion matrix TP, TN, FP, FN concepts, Accuracy, Precision, Recall, and F1 Score formulas. Practical guide for choosing the right classification metric across spam detection, fraud prevention, and cancer screening scenarios.
Read more →
Deep Dives·4 min
How Dangerous Are Windsurf Account-Switching Plugins? A Deep Dive into Technical Principles and Security Risks
Deep analysis of Windsurf seamless account-switching plugin's technical implementation, revealing risks of account bans, code theft, and legal liability. Four compliant alternatives for developers.
Read more →
Deep Dives·4 min
XANI: How GPU-Accelerated X-Ray Nanoscale Imaging Is Revolutionizing XFEL Data Analysis
Explore how the XANI project uses NVIDIA GPUs to accelerate XFEL data analysis, compressing nanoscale imaging from days to hours and advancing fusion materials and semiconductor research.
Read more →