Zero to Agent in 30 Minutes: Build Your Own Content Engine

Max Johnson, the founder of the AI studio briix, spends a lot of time sharing bite‑sized AI tips for founders. But ideas are cheap unless they actually make his own work easier. So he set out to replace the long‑winded, handwritten process of hunting topics, scoring them, writing hooks and then drafting full scripts with a single, automated assistant built on Claude Code.

The first thing Max did was give the model a little “brain”. He gathered a modest knowledge base – a few markdown files that described his business, his audience, sample scripts he liked, and some notes on his preferred voice. Those snippets act like a personal style guide, letting Claude Code judge relevance and mimic the tone he normally uses.

Next, he kept the instructions deliberately human. "Vibe coding," he calls it – basically spelling out in plain English what you want the AI to do, then stepping back and letting the model construct the rest. Of course, Max never walked away completely; he set the permissions, answered clarification prompts, and gave the final thumbs‑up on every output.

He then broke the whole operation into three bite‑size stages. Stage 1 asks Claude Code to research a list of possible topics. Stage 2 scores each topic and picks the most promising. Stage 3 spins three hook ideas for the winner and expands the strongest one into a full script, all while pulling from his knowledge base. Running those prompts manually gave him a three‑step pipeline that worked, but required a terminal command for each step.

To make it feel like a real “agent” rather than a collection of prompts, Max asked Claude Code to stitch the stages together. The result? A single command that churns out a structured JSON file containing topics, scores, hooks and scripts. That, he says, is the moment the system stops being a set of disjointed prompts and becomes an autonomous content engine.

He didn’t stop there. Max built a tiny web dashboard that reads the JSON and displays everything in a clean, browser‑based UI. Now he can glance at the top‑rated topics, compare hook variations, and launch a brand‑new research run without ever opening a terminal again. His tip? "Make the interface look purposeful, not like a default template – you’ll be using it a lot, so give it some personality."

The first iteration had two quirks: it only produced a script for the single highest‑scoring topic, and starting a fresh run still required the command line. By asking Claude Code to generate hooks and scripts for the top five topics and adding a "Run Again" button to the dashboard, Max got fifteen script options per cycle. That’s a lot more flexibility for testing headlines, A/B‑testing copy, or just having backup ideas.

If you’re thinking about replicating this, Max’s advice is simple: start with a repetitive task you can describe in a sentence or two. Build a working prototype with the coding tool you prefer, use it, watch where it breaks, then iterate. The learning curve is surprisingly gentle – you’ll have a functional agent in under half an hour.

Want to see the whole process in action? Check out the latest episode of “Zero to Agent in 30 Minutes” on O’Reilly Radar, or tune in on YouTube, Spotify, or Apple Podcasts. O’Reilly members can even attend live sessions. Grab a seat and start automating your own content flow.