Agentic Foundations

Building a reliable agent requires more than just a prompt; it requires a deep understanding of cognitive architecture and the primitives that allow an LLM to function as a “reasoning engine.”

For the systems-level view of why these primitives become reliability problems in production, see Engineering Agent Systems, especially The Engineering of Uncertainty.

🏛️ Theoretical Frameworks

What is an Agent? Exploring the spectrum of agency—from simple scripts to fully autonomous entities.
Cognitive Architecture How to structure the ‘brain’ of an agent using memory, planning, and sensory modules.

⚙️ Modern Primitives

Inference-Time Compute Understanding why “thinking longer” (Search, Verifiers, and Rollouts) leads to better agentic performance.
Modern LLM Primitives A look at function calling, structured output (JSON mode), and token-level control.

Core Learning Objectives

By the end of this module, you should be able to:

Distinguish between a Chain and an Agent.
Understand the trade-offs between Pre-training power and Inference-time reasoning.
Design a basic state-machine-based architecture for a task-oriented assistant.

Next Steps

After mastering these foundations, you can explore how these components are assembled in Agent Architecture.