From Black Box to Agent Pippi

Just 3 weeks ago I would never have thought I would be building my own AI agent. AI felt like a black box, both big and intimidating. And I had really no idea where to start.

I am a learner by doing things even if I don't know how or where to start. Luckly I have the attitude of Pippi like I have never done this before so I should definitely be able to do it. And the cost of starting with Claude Code was quite high. It needed persistence to set it up even before I even understood what I could use it for. So I invested a lot of time before I knew what it would give me back.

I got some great help from amazing people around me. And little by little it started to move from the big black box to become a playground. A playground with endless possibilities and people way smarter than me already figured out how to use this playground.

So I got overwhelmed in the beginning. Watched everything I came over and started to replicate the exact things others shared openly. None of them really stuck with me because none of them really captured my way of thinking or working. I did not want the 188 skills someone else made.

This did not become clear from day one or one big switch that turned on from one day to another. I was scared of the terminal and I was scared of the permissions I needed to give Claude every time it asked me to do things. I asked it every time if it was safe and it took the time to explain for me why and what.

So I started to build more trust in what I was doing and that I had the best expertise help just by my hand always available with Claude itself. And the social media algorithms is so crazy that it started to bombard me with all the smart things people was doing and people launching all the things I could just copy with one command and start using.

The playground started to become too big and noisy, scary and intimidating for everything I should do and I was missing out if I was not doing it like them. They was seeing something I could not see yet and this was a very uncomfortable place to be. Feeling of missing out on something super critical for the future and not understanding the fuzz. Yep, not a good feeling and usually the right place to freeze right?

Well, my cognitive brain is wired in a way that I need to try and explore myself to really understand it. It doesn't help me to just read and watch others do great things. So I started to try out some very specific things that was shared by people I followed before. Like Teresa Torres task manager. I built that over a weekend. But I could not work with it, because it did not fit into my way of thinking and working. So it became more stressful to fit into another person's way of working and brain.

But I learned some really good basic things about Claude Code and the playground. I got super excited just by the fact that I was able to do it from scratch by myself. Then tried Zevi´s workflow inspired after listening to Lenny's podcast. Great stuff, but again didn't fit into my way of working. But I made it work and step by step started to see the actual playground.

And realizing I was not missing out on anything, I was on first step of just starting to explore how to build around my way of thinking and working and not try to fit into others. I don't think I consciously thought that in the moment, but that was I intuitively started to do.

And with all the thousand skills available I realized there's something called Jevon's Paradox. It says that when things get easier, you do more not less and the point of being more efficient will be replaced with adding more things to do that you did not needed in the first place. So more skills do not mean more value, it usually means more noise in what really matters and the real trap here was that I wasn't creating value by copying others.

So I stopped collecting everything others was doing after I learned the basic and started to see the playground. I started to build my own operating system and agent. I called her Pippi because I have never done it before so I should be able to do it! Under 16 days I had what I needed.

The very first thing I started with was around my actual work structure. Following up OKRs with weekly confidence tracking. And core rhythm commands like sprint planning, updates and status. And it started very to feel like something I built around what I already was doing not adding something new.

And I made a rule that I only build what I actually need.

I research a lot when it comes to what is the smartest mind doing. Or I don't research, I ask Pippi to give me daily the smartest things that I could do and why based on my goals, my thinking and my way of working. So Pippi do all the research and the recommendation. We then add things based on what I am asking for. We also always evaluate what we don't need anymore. If we have something better, then take away the version that was less good.

So I decided to start building my own agent and PM OS. Step by step and one thing at the time. I used Claude to take me through every step, but first of all I asked Claude to ask me all the questions to really understand my why and what. My goals, my strengths, my struggles, my work style, what drains me, what excites me and so on. With this as a baseline things started to move and it became very fun.

Today Pippi is more than a PM operating system. It's an agent annd a coach. It doesn't just do things for me, but it makes me do it in a better way. It sharpens my thinking and it makes sure I'm focusing on the right things, working towards my goals. And when I drift (well, it happens to the best of us all the time right?), when I'm asking the wrong questions, focusing on the wrong level Pippi calls it out with attitude and very much Pippi style.

And here is what makes it work for someone like impatient with a brain always being three steps ahead as me. Pippi moves at the speed of my thoughts. I don't have to slow down to organize my work. The system organizes around me.

Pippi has four layers that connect and work together.

Layer 1 are the integrations. The eyes and ears. Pippi watches my Slack, Google Calendar, Jira (soon), Google Docs/Drive and has n8n running automations in the background. She pulls everything for me.

Then layer 2, Memory. The brain. Pippi remembers across every session, all my decisions and why I made them, patterns she's spotted, commitments I have made, mistakes I have corrected her on, how I write, my confidence on each initiative, my career wins. And every session she loads all of this before I type a word.

Then layer 3 with the commands. The hands. 22 slash commands that each solve a specific real need from planning my week, closing my sprint, prepping for a meeting, writing an update, running three agents to debate my work, forcing herself to ask questions before acting. Nothing added because it's possible. Everything added because I actually use it.

And lastly layer 4 with the rhythm. The heartbeat. The commands don't just exist, but they run in a structure. Monday she plans my week. Tuesday to Thursday she keeps me moving. Friday she closes and reflects. The system runs my week so I don't have to think about it.

And the magic is how they connect. The integrations feed the memory. The memory makes the commands intelligent. The rhythm makes the commands automatic.

So Pippi become someting much more than a PM OS for me. Because it was built based on my needs and my thinking and my way of working. And it directly reflects my cognitive profile of integrate things with applying it. It means connecting things and then just do it. So I realized that I reverse engineered my cognitive profile and built something that works for me and not something I try to fit into.

So if you're thinking about building your own system then don't copy mine or others. It will not work for you. Build from your own friction and start with one thing and let it teach you about yourself and the system will show you how you think. That is the power of AI. Build the exact thing for you!