Your First Real Agent
Built with an SDK, understood without one: tools, memory, and the loop wired into one research agent over the awesome lists.
Welcome to all things agentic.
I started this blog to keep notes on what I learn while building with LLMs. Agents, tools, orchestration, evals, and the rough edges that only show up once something is running.
Expect working notes more than polished essays. The goal is to get useful ideas out while they're still fresh. If a post saves you an afternoon of head-scratching, pass it along.
Get the free newsletter 👇
Drop your email if you'd like new posts on agents, RAG, and evals in your inbox. No fixed schedule, just when I have something concrete to share.
Featured posts
Recent posts
Why Your Agent Goes In Circles?
The three loop topologies (ReAct, plan-and-execute, orchestrator/subagent), why your agent goes in circles, and what Runner.run is actually doing for you.
Memory: Why Your Agent Forgets You?
Why agents forget, the three layers of memory you actually need, and how to wire a vector store into the loop without the framework hiding what's happening.
Tools: How Agents Perform Actions?
What an agent tool actually is, how OpenAI Agents SDK, Anthropic, and Google ADK each declare one, and why every tool is a blast-radius decision.
How agents actually reason?
Inside the reason phase. How LLMs turn a goal into the next action, why chain-of-thought is just more tokens, and why your prompt is the agent's architecture.
What even is an AI agent?
Why an AI agent is more than a chatbot with extra steps, and the perceive, reason, act loop every framework eventually maps back to.