Reset ticksPrint / save PDF

1. Large language models (LLMs)

• Explain what a transformer is
• Explain how an LLM works by predicting the next token
• Explain what weights are in an LLM
• Cover model types:
  • Frontier models
  • Everyday models
  • Open-source models
• Show how people can run open-source models on their own machine, with no cost for external providers

2. Model basics

• Explain what a token is
  • Give an example with a tokenizer
• Explain what a context window is
  • Short context window
  • Long context window
• Give scenarios that need a long context window
• Explain the advantage of a long context window
• Explain how cost differs between short and long context windows
• Explain why input tokens and output tokens have different prices

3. AI agents

• Explain how an agent works together with the model and its tools
• Explain why an agent needs tools
• Explain how an agent uses tools based on instructions from the model
• Explain what a tool call is
• Walk through a small example: creating a website
  • Agent sends a request to the model
  • Model tells the agent which tool to use
  • Agent makes the tool call
  • Agent sends the result back to the model

4. Agent harness

• Show that LLM + model + agent are grouped together inside a harness
• Explain what a harness is
• Explain how a harness is different from an agent
• List the extra capabilities a harness gives an agent:
  • Tool calling
  • Skills
  • Memory
  • Execution
• Explain how the harness handles memory
• Explain how the harness handles skills
• Explain what a skill is
• Show how to create a skill

5. Terminal-based UI and Hermes

• Explain what a terminal-based user interface (TUI) is for a harness
• Compare a TUI with a graphical user interface (GUI)
• Introduce Hermes as one example of a TUI-based harness
• Start the Hermes section from here