February 1, 20255 min

Agentic AI: Building Systems That Think in Steps

Moving beyond chatbots — how I'm exploring autonomous AI agents that plan, execute, and self-correct.

AILLMArchitecture

The AI landscape is shifting from "ask a question, get an answer" to "give a goal, watch it execute." This is the agentic paradigm, and it's where I'm focusing my exploration.

What Makes an Agent?

An AI agent isn't just a chatbot with tools. It's a system that can:

Plan — Break a complex goal into steps

Execute — Use tools and APIs to carry out each step

Observe — Evaluate the results

Adapt — Adjust the plan based on what happened

This loop — Plan → Execute → Observe → Adapt — is remarkably similar to the OODA loop (Observe, Orient, Decide, Act) that military strategists use. My naval background makes this feel natural.

RAG: The Agent's Memory

Retrieval-Augmented Generation is the backbone of any useful agent. Without it, you're limited to what the model was trained on.

I've been experimenting with:

Vector databases for semantic search over documentation

Hybrid search combining keyword and semantic approaches

Chunking strategies that preserve context boundaries

The key insight: RAG quality depends more on your chunking and embedding strategy than on the LLM itself.

Where This Is Going

I see agentic AI transforming backend engineering:

Automated incident response — Agents that can diagnose and fix common production issues

Code review agents — Beyond linting, actually understanding architectural implications

Test generation — Agents that understand your domain and generate meaningful test cases

The engineers who understand both the AI capabilities and the systems they're being applied to will be the ones who build the most impactful solutions.

That's the intersection I'm positioning myself at.

Back to Captain's Log