AI & Agents — Reading Order

A structured path through my AI & Agents posts. The progression moves from foundations through LLMs and retrieval into agentic systems and production concerns. Each section builds on the previous one.

1. Foundations — What AI Actually Is

Start here if you’re coming from a pure software engineering background. These posts build the mental model for everything that follows.

1. AI Hierarchy — the nested capability stack: AI ⊃ ML ⊃ DL ⊃ GenAI
2. Machine Learning Models & Learning Paradigms — supervised, unsupervised, reinforcement — when to use which
3. Top 20 ML Algorithms — the algorithms that power production systems
4. Evaluation Metrics — precision, recall, F1, AUC — choosing the right metric
5. Deep Learning Architectures — CNNs, RNNs, transformers — the neural network zoo

2. GenAI & LLMs — How Language Models Work

The transformer revolution. Understanding this layer is prerequisite to building anything useful with LLMs.

1. LLM Architecture & Training — from transformer internals to training at scale
2. Controlling Randomness — Temperature, Top-k, Top-p — the creativity-coherence tradeoff
3. Fine-tuning vs RAG vs Prompting — three ways to customize LLMs, with tradeoffs
4. LLM Evaluation — measuring quality across accuracy, latency, cost, safety

3. RAG & Knowledge Retrieval

Grounding LLMs in your own data. This is where most enterprise AI projects start.

1. RAG Architecture — retrieval-augmented generation end to end
2. Vector Databases — semantic search at scale with embeddings

4. Agent Frameworks — The Toolbox

Comparative deep-dives into the major frameworks. Read the first two to understand the landscape, then the ones relevant to your stack.

1. Agent Orchestration & Handoffs — patterns for wiring agents together
2. Agent Memory & State — turning stateless LLMs into systems that learn
3. LangChain & LangGraph — composable primitives and stateful graph-based workflows
4. CrewAI — role-based multi-agent systems with minimal boilerplate
5. Microsoft AutoGen — multi-agent conversations with code execution
6. OpenAI Agents SDK — lightweight agents with handoffs and guardrails
7. Google Agent Development Kit — hierarchical orchestration on Vertex AI
8. Claude API & Agent SDK — Anthropic’s production agent framework

5. Protocols & Interoperability

How agents connect to the outside world and to each other.

1. Model Context Protocol (MCP) — the open standard for agent-tool integration
2. Agentic Protocol Landscape — the full protocol stack from tool calls to payments

6. Production — Shipping AI Responsibly

The hard part. Everything above is table stakes — this section is what separates demos from products.

1. Harness Engineering — the operating system around the model
2. Evals & Guardrails — catching regressions and preventing harm
3. Cost Control & Optimization — keeping agentic workloads economically viable
4. Async Agent Workflows — long-running agents that checkpoint and resume