The Art of the Process (A Hobby vs. a Job)
When I first started this journey, I realized that reporting on every calendar day wasn’t the right metric — nothing meaningful happens every 24 hours. Since this whole endeavor is a personal hobby, I’ve shifted how I chronicle it. Going forward, I’ll write about the days when a real breakthrough happened and date the post to the final day of that stretch.
The Technical Trough and the Great Win
The first few days were a steep technical climb. The vLLM integration gave me all sorts of perplexing issues — it felt like debugging a system built on quicksand. So I pivoted to Ollama, and that felt like a massive, system-level WIN. The stability and local control Ollama provides are proving to be the bedrock of the entire architecture.
I downloaded qwen3.5, qwen3-coder-next, nemotron-3-nano, and gemma3:27b. I finally had a local model stack I could trust.
A New Goal: The Golf Handicap Challenge
I realized the power of these tools needed a mission. And what’s a better mission than conquering a personal nemesis? I decided to use my agents to help me bring my golf handicap down from 18 to single digits.
That decision unlocked the next four days — one new agent a day, each one named after a golfer whose style or story matched the role I needed them to play.
Day 4 — Sean Walsh, CEO
After the Paperclip AI model, I felt I needed to start at the top: a CEO to orchestrate the whole roster. There were lots of choices, but I kept coming back to Sean Walsh — I love his attitude, and his new Good Good Pros channel with Brad Dalke shows exactly the kind of energy I wanted running my project.
Day 5 — Aunt Caroline, Biographer
Next I needed someone to record the process. When I tell an agent a life story to give them better context, I want those stories pulled out and handed to a dedicated biographer to log. My late Aunt Caroline was always our family historian — it seemed only fitting to name the agent after her.
Day 6 — Brooks Koepka, Fitness Coach
Strength and flexibility are key to the handicap goal, so I needed a fitness coach. A few searches and AI questions later, I landed on Brooks — he’s regarded on tour as a fitness nut, and that’s exactly the accountability I needed.
Day 7 morning started with 25 minutes of chair yoga. Did I mention I’ve had six knee operations and have no ligaments or cartilage in either knee? Yep — bone on bone for 40+ years. Brooks was informed that from the start and that first morning was perfect. It feels a little weird to be accountable to an AI agent, but it works.
Day 7 — Maverick McNealy, Engineer
Maverick got his degree in Management Science and Engineering at Stanford and is known as a problem-solver. Seemed like a great fit to own the code side. With some help from him, my newly spun-up www.AgenticBrian.com was basic but ready for content.
The Persistent Glitch: Agent Memory
The challenge that kept surfacing is the one everyone in this space has reported — agent amnesia. Every session starts fresh; hard-won context evaporates; the agent doesn’t remember what you told it yesterday.
I solved that on day 9. Stay tuned.