Model vs Agent: What Ollama Does, What ADK Does, and What My AI Tool Was Doing the Whole TimePart 6 of the series where I build an AI agent to learn how agents really work. For days I used the word "agent" for everything. This is the post where I pulled it apart. Somewhere in the middle of t7 min read·theagenticworld.hashnode.devNov 2
Tokens, Context Windows, and Why My Model Ran With 4,096 of 32,768Part 5 of the series where I build an AI agent to learn how agents really work. The last post left me with questions I couldn't answer. Here are the answers, using my own numbers. At the end of the l8 min read·theagenticworld.hashnode.devNov 2
I Printed Every Request My AI Agent Sends. Here's What I Got Wrong.Part 4 of the series where I build an AI agent to learn how agents really work. I'd been calling it an agent for days without ever reading what it actually sends to the model. By this point in the pr7 min read·theagenticworld.hashnode.devNov 2
Rules First, Then the AI: Guardrails, Evals, and Every Way My Agent BrokePart 3 of the series where I build an AI agent to learn how agents really work. This is the long one: the whole middle of the project, from writing the rules to measuring three different models. Here14 min read·theagenticworld.hashnode.dev11SNov 2
Talk to the Model Before You Build the Agent (and the 260-Second Mystery)Part 2 of the series where I build an AI agent to learn how agents really work. One request took 260 seconds to answer. I almost "fixed" the wrong thing. Before I wrote any agent code, I did the leas6 min read·theagenticworld.hashnode.devNov 2