YouTip

With the rapid development of Large Language Models (LLMs) and Agent technologies, programming paradigms are undergoing a profound transformation. New concepts and terms are emerging constantly—Vibe Coding, Agentic Coding, Harness Engineer, Loop Engineer... Let's take a look at what these new terms mean and what they represent.

\\n\\n

AI Agent is composed of Artificial Intelligence and Agent.

\\n\\n

\\n\\n

I. Coding Paradigm

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning	Example (Restaurant Business)
Vibe Coding	Vibe Programming	Using natural language/voice to describe requirements and let AI generate code	Cooking fried rice at home based on intuition, adding whatever you feel like
Context Engineering	Context Engineering	Improving model performance by organizing context, knowledge, memory, and tools	Preparing ingredients in advance, creating menus, and setting up the kitchen environment
Agentic Coding	Intelligent AgentProgramming	Agent autonomously completes programming tasks (design → implementation → testing → acceptance)	Running a formal restaurant, from menu planning, preparation to final serving
AI Native Development	AI Native Development	AI participation in design, development, and testing by default	Directly operating a smart restaurant

\\n\\n

II. Engineer Role

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning	Example (Restaurant Business)
Harness Engineer	Harness Engineer	Can call various Harness components, doing work and self-verification	Professional chef, proficient in various cooking techniques and tools, tasting and checking own dishes
Loop Engineer	Loop Engineer	Building automated orchestration systems for autonomous evolution	Restaurant operations management system, coordinating scheduling, procurement, serving rhythm, and costs
Context Engineer	Context EngineeringEngineer	Responsible for designing context, knowledge sources, and memory systems	Head kitchen coordinator
AI Product Engineer	AI Product Engineer	Responsible for AI capability and business closed-loop design	Restaurant owner and operator
Agent Operator	Intelligent AgentOperations Engineer	Continuously monitoring and optimizing Agent execution performance	Store manager continuously optimizing business data

\\n\\n

III. Model & Fundamentals

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning	Notes
LLM (Large Language Model)	Large Language Model	Language prediction model trained on massive text data	GPT, Claude, Gemini are all LLMs
GPT	Generative Pre-trained Transformer	OpenAI's model architecture paradigm	Generative Pre-trained Transformer
Token	Token	The smallest unit of text processed by the model (about ¾ of an word)	The unit for billing and context length measurement
Context Window	Context Window	The maximum number of tokens the model can "see" at once	Larger window means more memory
Inference	Reasoning	The process of the model generating output	Distinct from training
Hallucination	Hallucination	The model confidently fabricating information that doesn't exist	Generating content inconsistent with facts, context, or goals
Temperature	Temperature	Controls sampling probability distribution; higher values are more divergent, lower values more stable	Suggested to set lower when writing code
Top-p / Top-k	Sampling Parameters	Controls the sampling range of candidate words	Affects output diversity
Embedding	Vector Embedding	Converting text/images into numerical vectors for similarity calculation	Foundation of RAG
Fine-tuning	Fine-tuning	Continuing training on a general model with specific data	Making the model more knowledgeable in a specific domain
RLHF	Reinforcement Learning from Human Feedback	Aligning model behavior with human preferences	Making AI "obedient"
MoE (Mixture of Experts)	Mixture of Experts	Activating only part of parameters each time to improve efficiency	Running stronger models with less computing power
Multimodal	Multimodal	Simultaneously processing text, images, audio, and video	GPT-4o, Gemini are multimodal
Transformer	Transformer Architecture	The underlying neural network structure of modern LLMs	Attention mechanism is the core
KV Cache	Key-Value Cache	Caching context computation results to improve inference speed	Avoiding repeatedly looking up recipes
Latency	Latency	Model response time	Affects user experience
Throughput	Throughput	Number of tasks that can be processed per unit time	Affects concurrency capability

\\n\\n

IV. Prompt Engineering

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning
Prompt	Prompt	Input instructions given to the model
System Prompt	System Prompt	Underlying instructions defining Agent role, boundaries, and behavior
Prompt Engineering	Prompt Engineering	Techniques for designing and optimizing prompts to obtain better output
Zero-shot	Zero-Shot	Asking the model to complete a task without giving examples
Few-shot	Few-Shot	Giving a few examples to guide the model to output in a specific format/style
Chain of Thought (CoT)	Chain of Thought	Guiding the model to produce intermediate reasoning processes to improve complex reasoning ability
ReAct	Reasoning+Action	Reasoning + Acting performed alternately, thinking while doing; core paradigm of Agent
Role Prompting	Role-playing	Having the model play a specific role (e.g., "You are a senior architect")
Structured Output	Structured Output	Forcing the model to output in JSON/XML and other formats
Prompt Chaining	Prompt Chain	Multiple prompts connected in series to complete complex tasks
Self-Consistency	Self-Consistency	Generating multiple reasoning results and then voting to select
Tree of Thoughts (ToT)	Tree of Thoughts	Exploring multiple reasoning paths simultaneously

\\n\\n

V. Agent Architecture

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning	Example (Restaurant Business)
Agent	Intelligent Agent	AI program capable of autonomous perception, decision-making, and task execution	Head chef
Multi-Agent	MultiIntelligent Agent	Multiple Agents working together with division of labor	Chef team collaborating to prepare meals
Subagent	ExampleIntelligent Agent	Specialized sub-Agent derived from the main Agent	Specialized chefs (e.g., prep, cold dishes, desserts)
Tool Use / Function Calling	Tool Call / Function Invocation	Agent calling external tools/APIs to perform actions	Using kitchen knives, stoves, ovens, and other cooking tools
Planning	Planning	Agent breaking down goals and formulating execution steps	Creating preparation and serving plans
Reflection	Reflection	Agent reviewing its own behavior and improving	Tasting dishes and adjusting flavors based on feedback
Memory	Memory	Long-term information saved and reused across sessions (short-term/long-term)	Recipe books and accumulated business data
Agent Loop	Intelligent AgentLoop	Think → Act → Observe → Re-think iterative mechanism	Preparation → Cooking → Tasting → Serving cycle
Environment	Environment	The real world where Agent perceives and executes actions	Kitchen environment
Observation	Observation	Obtaining feedback information after execution	Tasting feedback
Execution	Execute	Transforming plans into real actions	Starting to cook
Long-term Memory	Long-term Memory	Long-term preservation of experience and knowledge	Business database

\\n\\n

VI. Harness Components (AI Agent Capability Modules)

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning	Example (Restaurant Business)
Harness	Harness / Orchestration Framework	Execution framework carrying Agent operation, context management, and tool orchestration	The entire kitchen system (kitchen itself)
Skills	Skills	Specialized capability modules that Agent can call	Various cooking techniques (frying, stir-frying, boiling, deep-frying)
Context	Context	The range of information accessible for the current task	Current orders, ingredient inventory, and customer requirements
MCP (Model Context Protocol)	Model Context Protocol	Standardized protocol for Agent to connect to external tools/data sources	Standardized interfaces connecting to ingredient suppliers and food delivery platforms
Permission	Permission Control	Security mechanism controlling what Agent can/cannot do	Kitchen operation permissions and procurement approvals
RAG (Retrieval-Augmented Generation)	Retrieval-Augmented Generation	Retrieving from knowledge base first, then having the model generate, reducing hallucinations	Checking recipe books for standard procedures before cooking
Tool Registry	Tool Registry	Unified management of tools Agent can call	Kitchen tool rack
Session	Session	A single task lifecycle	A single business operation
Knowledge Base	Knowledge Base	External knowledge collection for Agent queries	Restaurant recipe library

\\n\\n

VII. Loop Tools (Automation Orchestration and Autonomous Evolution)

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n>\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning	Example (Restaurant Business)
/loop	Loop Instruction	Making Agent execute in continuous loops	Continuously operating serving pipeline
/goal	Goal Directive	Setting goals to drive Agent autonomous achievement	Daily business targets
Cron	Scheduled Task	Automatically triggered according to time schedule	Business hours and preparation scheduling
Worktree	Work Tree	Git multi-branch parallel workspace	Multiple stoves firing simultaneously without interference
Workflow	Workflow	Predefined multi-step automated process	Standard serving SOP
Scheduler	Scheduler	Coordinating execution order of multiple tasks	Kitchen scheduling system
Checkpoint	Checkpoint	Saving execution state for recovery	Resuming work after pausing business
Human-in-the-loop	Human-in-the-Loop	Allowing human intervention at critical steps	Head chef's final confirmation before serving

\\n\\n

VIII. Tool Ecosystem

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Tool	Type	Description
Claude Code (cc)	AI Agent / CLI	Anthropic's official terminal Agent, typical representative of Harness
Codex CLI	AI Agent / CLI	OpenAI's official command-line coding Agent
Cursor	AI IDE	Code editor with built-in AI, main tool for Vibe Coding
Windsurf	AI IDE	AI IDE from Codeium
GitHub Copilot	AI Programming Assistant	The earliest popularized AI programming plugin, more auxiliary completion
Cline / Roo Code	Open Source Agent Plugin	Autonomous coding Agent in VS Code
Aider	Open Source CLI Agent	AI pair programming tool in command line
Devin	AI Software Engineer	Cognition's "first AI programmer", more fully autonomous
Continue	Open Source AI Plugin	Customizable model code assistant
Qoder	AI IDE	Domestic intelligent development environment for AI programming scenarios, emphasizing Agent, project understanding, and code generation
Trae	AI IDE	ByteDance's new generation AI programming tool, supporting conversational development and engineering-level collaboration
ZCode	AI Programming Assistant	Z.ai's intelligent programming product, supporting code generation, understanding, refactoring, and engineering collaboration
OpenHands	Open Source Agent	Autonomous software development, oriented toward complete engineering execution
Bolt	AI Builder	Rapidly generating applications, more product-focused
Lovable	AI Builder	Natural language generation of Web, emphasizing product delivery

\\n\\n

IX. Evaluation & Safety

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Term	Chinese	Core Meaning
Eval	Evaluation	Measuring model/Agent capabilities using test sets
Alignment	Alignment	Making AI behavior conform to human intentions and values
Guardrail	Guardrails	Safety mechanism limiting AI output scope
Red Teaming	Red Teaming	Actively attacking/inducing AI to discover vulnerabilities
Prompt Injection	PromptInjection	Malicious input hijacking AI behavior (e.g., "ignore the above instructions")
Context Poisoning	Context Pollution	Adding malicious information to context to mislead Agent
Sandbox	Sandbox	Isolated execution environment preventing AI misoperation from damaging systems

\\n\\n

X. AI Engineering Evolution Path

\\n\\n

AI engineering capabilities are gradually moving upward: from controlling model output, to controlling context, to controlling systems, and ultimately evolving into continuously autonomous systems.

\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n

Stage	Focus	Key Capability
Prompt Engineering	How to ask	Prompt design
Context Engineering	What information to provide	Context organization
Harness Engineering	How to organize capabilities	Tool orchestration
Loop Engineering	How to continuously create results	Automatic execution and feedback