AI Agents

An AI agent is a software system that can perceive its environment, make decisions, and take actions to achieve specific goals with some degree of autonomy.

Key characteristics of AI agents include:

• Autonomy
• Goal-oriented behavior
• Environmental awareness
• Decision-making capability
• Learning and adaptation

Human vs agent:

• True cretivity
• Common sense
• Intuition
• Emotional intelligence
• Leadership
• Ethical oversight

Design Patterns

AI Agent components:

• LLM
• Prompt
• Memory
• Tools

AI Agent Design Patterns:

• Reflection (self-evaluation)
• Tool Use (tool integration)
• Planning
• Multi-Agent (agent collaboration)

Benefits of multi-agent:

• Simpler context management
• Shorter prompts
• Freedom to choose a provider for each agent individually
• Fewer tools -> less confusion
• Parallel execution -> less latency (chained -> higher latency)
• Easier to test

Vocabulary

Reasoning - the cognitive process of drawing conclusions, making inferences, and forming judgments based on available information, evidence, or premises. Thinking in a logical way.

Sources

Building AI Agents from Scratch | Full Course by The Neural Maze A simple Python implementation of the ReAct pattern for LLMs by Simon Willison Agentic Artificial Intelligence by Pascal Bornet Goose Architecture AI Engineering by Chip Huyen Agentic AI Design Patterns by Satyam M.