AI Agents
AI Agents are intelligent systems capable of autonomously perceiving environments, planning decisions, and executing actions. From classical BDI architectures to modern LLM-driven agents, this field is undergoing a fundamental shift from "tools" to "autonomous collaborators."
Lilian Weng's canonical formula: AI Agent = LLM + Planning + Memory + Tools
This section comprehensively covers the theoretical foundations, core technologies, development frameworks, and industry ecosystem of AI agents:
Contents:
- Agent Overview — Definitions, taxonomy, history, milestones
- Cognitive Architectures — BDI, ACT-R/SOAR, LLM cognitive architectures, design patterns
- Planning & Reasoning — CoT, ReAct, ToT, Reflexion, reasoning frontiers
- Memory Systems — Context management, vector databases, RAG, memory architecture
- Tool Use — Function Calling, MCP, Computer Use, sandboxing
- Multi-Agent Systems — Communication protocols, collaboration patterns, game theory, emergence
- Virtual Embodied Agents — Generative agents, memory streams, NPC evolution, social emergence
- Agent Frameworks — LangChain, Claude SDK, OpenAI SDK, low-code platforms
- Agent Applications — Code generation, Web, data analysis, research agents
- Evaluation & Benchmarks — AgentBench, SWE-bench, cost-benefit analysis
- Deployment & Operations — Security sandbox, human-agent collaboration, monitoring, alignment
- Industry Ecosystem — Company landscape, open-source, market trends, challenges