Six Months Before the Afternoon

The personal story behind The Individual Operator: how about half a year of learning to work with agents made one afternoon possible.

Before

In The Individual Operator I described an afternoon in which AI agents built me a studio site and its first five projects. That afternoon can look like a sudden jump. It was not. It was the visible end of something much slower.

For the last six to eight months I have worked with agentic AI. Before that, my relationship with AI tools was the one most people still have: ask a question, get an answer, copy what is useful. I treated the tool as a clever search box. I judged its output by whether it sounded right.

That habit is harder to spot than it sounds. A confident answer feels like a correct one. I had not yet learned to look at the shape of an answer, or to ask what a good one would even look like. I knew AI was powerful. I did not yet know how to work with it.

Finding Entropic AI

The change started when I found a sandboxed environment called Entropic AI.

The word that matters there is sandboxed. Working with agents is different from chatting with a model. An agent acts. It writes, runs and changes things, and that is exactly why people hesitate to hand it real work. A sandbox lowers that fear. It gives you a contained place where an agent can try things, break things and be corrected, and where a mistake stays a lesson instead of becoming a problem.

That safety changed how I behaved. I stopped being careful in the wrong way, where you avoid giving the tool anything meaningful, and started being careful in the right way, where you watch closely what it does. I gave agents bigger tasks. I read what came back. I noticed where they were strong and where they quietly went off course.

It was the first time AI felt less like a tool I consulted and more like something I could direct.

Learning next to real developers

Alongside that, I got the opportunity to work with good, real developers, and now with AI engineers, while interning. I will keep the details private, but I want to say what it did for me, because it mattered more than any single tool.

When you learn alone, you only see your own habits. Next to experienced people, you see theirs. I could watch how they read a piece of work: where their eyes went first, what they skipped, and what made them slow down. I could see how they described a task so that the person or the agent on the other end could not misunderstand it. And I could see how little they trusted a first impression. A glance told them where to look, and then they checked.

None of that was a secret technique. It was mostly patience and attention, applied consistently. But seeing it done by people who do it for real is different from reading about it. It made the idea of a trained way of thinking concrete for me.

I am grateful for that. It gave me a standard to measure myself against, and it showed me how far there still is to go.

What changed in me

Over those months, four things shifted. They are the same four I wrote about in The Individual Operator, but here I want to describe them as I lived them, not as an argument.

First, I started reading context before content. I stopped trying to absorb every line of an output and began noticing its structure: what was dense, what was vague, what was too long for its purpose. Those signals told me where to spend my attention.

Second, my taste improved, slowly, through volume. Reviewing a lot of output, and forcing myself to say why one version was better than another, turned a vague feeling into words I could use.

Third, my briefs got better. I used to blame the agent when a result was poor. Gradually I learned that a poor result was usually feedback on my own thinking. If I could not explain what I wanted clearly, I did not know it yet. Writing a precise brief became the moment I actually made my decision.

Fourth, I found a rhythm: judge at a glance, then verify. Trust the quick read enough to know where to look, and never trust it enough to skip the check.

I do not want to overstate this. As I said in the first essay, I am still early in training this way of thinking, and I expect to be for a long time. But early is not the same as absent. The change was real, and I could feel it in how I worked.

The slow part was me.

How it led to the afternoon

So when I sat down and asked agents to build me a studio site, I was not starting from zero. I was bringing months of habits with me.

I knew how to brief precisely, so the agents had something clear to build. I knew how to read an output quickly, so I could tell what was working and what was off. I knew to run and click through the result rather than take it on faith. The site and its first five projects came together in an afternoon, with about 45 minutes of my own time, because the slow part had already happened. The slow part was me.

The agents wrote, tested and shipped. My job was to steer, to judge and to decide. That is only possible once you have built the judgment, and the judgment came from the months before.

Since then the work has kept growing: Lumen Shift, Palette Lab, Meridian, and a narrated showreel made with Higgsfield and Claude. Each one was easier to direct than the last, not because the tools changed overnight, but because I had practised.

This is what I mean when I say the age of the individual operator has already arrived. It is not a trick of a single afternoon. It is what becomes possible when a person spends enough time learning to work with capable agents. I did not need a team to build those projects. I needed to learn how to operate.

Where I am going

I do not have a finished plan, and I would be wary of anyone who claimed to. What I have is a direction.

I want to keep working alongside people who are better than me, because that is where the standard comes from. I want to keep reviewing a lot of output and keep asking why one version beats another. I want to keep writing briefs that force me to decide, and keep checking what I am tempted to trust.

And I want to keep building, in public, so that others can judge the work and not just the claims.

If there is one thing I would pass on, it is that this is learnable. I was not special at the start of those six to eight months. I found a safe place to practise, I learned from people who were good at it, and I kept at it. The habits formed because I repeated them.

If you want to see what that practice produced, and the argument I draw from it, read The Individual Operator at ibrahimhayd.nl/essay/. Then ask yourself the question it ends on, and notice how quickly your answer changes once you have actually tried.