How I became an Agenteer: building apps, agents, and automations without code
A personal essay. How I spent sixteen years bouncing off learning to code, found out I was an Agenteer, and started making things again.
A few months ago I was sitting on the floor with my two-and-a-half-year-old daughter, building a Duplo house with her. I looked up and realized two hours had passed.
Brick by brick, watching something come alive. Time disappeared. I felt like a kid again.
That's when the epiphany hit: the less time I spent building, the less I liked what I was doing.
The Duplo on the floor was telling me something I'd been avoiding for years.
I never wanted to run a business
I didn't become an entrepreneur because I wanted to grow a company, hire a team, raise money, exit.
I became one because as a student I wanted to make cool websites for people, and at some point someone said you'll need a VAT number for that. Then a business account. Then invoices. Then proposals. Then sales calls. Then follow-ups. Then briefings, meetings, slides, managing people. The job grew around the making until the making was the smallest part of the day.
I picked up titles along the way. Entrepreneur. Marketeer. Co-founder. Manager. Freelancer. Coach. Partner. Adviser. Consultant.
None of them were "builder."
I noticed.
But I'd always been one.
The orange box
My earliest memory is a giant orange box of Lego in our living room. Thousands of pieces. I lived in that box. Building something is the first real thing I remember doing.
After Lego came Photoshop. I made banners for forums as a teenager, the kind of thing that mattered to fourteen-year-olds in 2008 and absolutely no one else. After Photoshop came WordPress. I was sixteen when I built my first website and somehow turned it into pocket money. After WordPress came freerunning videos with my friends, edited frame-by-frame in whatever software I could pirate.
The medium kept changing. The urge didn't. Make something. Watch it come alive. Show it to someone.
It's the same urge that put me in front of the Duplo box on the floor.
I can't code. I tried.
I rarely say this out loud, but I can't code.
I tried. I really did. I was sixteen when my nephew gave me an HTML for Dummies book as a birthday gift. I read the whole thing. I built static sites with it. I learned how a <div> worked. That book is the furthest I ever got with a programming book in my life.
At eighteen I went to graphic and web design school. The programming lessons were the ones I dreaded most. HTML and CSS clicked because they're practical. Type something, see it on the screen. Real programming didn't. Variables, object-oriented programming, abstractions floating in mid-air. It was words on paper that didn't connect to anything I could see.
I'm a visual learner. I need instant feedback. Code never gave me that, so I never gave it back.
I kept trying. Codecademy in 2014. A Python course on Udemy in 2017. Three days of Learn You a Haskell in 2019 (don't ask). YouTube tutorials, every year, every time. Every attempt ended the same way. Somewhere between week one and week three, the abstractions stacked higher than my patience, and I closed the tab.
A decade of bouncing off the same wall. Sixteen years if you count from the HTML for Dummies book.
The Frankenstein
So I built workarounds.
Take Winwinner. We needed more than a static brochure for our first website. Members applying. Dashboards updating. Applications routing through a workflow. I couldn't build that in WordPress. I didn't know how to build it in code. So I stitched it together out of every tool I could use without writing code: Webflow for the front end, Typeform for the application forms, Airtable as the database, Zapier connecting the pieces, Memberstack for authentication on top.
A Frankenstein monster. It worked, sort of. It was fragile. Every API change somewhere down the chain broke something somewhere else. Every new feature meant another tool, another subscription, another point of failure.
But it was a website that did things. And I had built it. So when people asked, I said I built it.
The wall
Then Winwinner needed a real platform.
Members logging in. Custom calculations. Document handoffs. Compliance flows. Things the Frankenstein couldn't carry without falling over once a week.
I couldn't build that. We hired agencies. We hired contract devs. The first proper Winwinner platform cost us over €100k in development fees, spread across two agencies and several years.
The same pattern played out everywhere I worked. I'd spot something that needed building. I'd specify it carefully. I'd brief the developer or the agency. They'd quote it. We'd pay. We'd wait. Eventually, something would ship. Usually a little less than what I'd briefed, occasionally a little more.
The clearest example of how silly that wall got was a workflow I built in n8n a couple of years back. A cybersecurity client wanted SEO articles generated automatically. Research a keyword, pull competitor content, draft a post in their tone of voice, assemble the metadata. Reasonable ask. Should be straightforward.
It was not straightforward. I spent multiple days on it. Tweaking, testing, re-running, adding nodes, debugging. When something broke, I'd download the workflow JSON, paste it into ChatGPT, and ask GPT to figure out what was wrong. When I finally got it running, one in every ten executions would just stop or error and need a full restart from scratch.
It was hell. It was a nightmare. The whole thing took me about as long as it would have taken a developer to build the same logic in Python from scratch.
Looking at it now, I realize something. Most of what we paid agencies for, and most of what I sweated through alone with n8n, wasn't actually that complicated.
It just needed code.
The wall fell over
I opened Claude Code earlier this year, after seeing a demo on Twitter. I used it for small things at first. A script to clean up a CSV. A tweak to a config file. Stuff a junior developer would do in an afternoon.
A few weeks ago, the moment landed.
I rebuilt the entire Winwinner platform in Claude Code. By myself. In plain English. The way you'd brief a colleague. "This page needs a calculator that does X, Y, Z and shows the result here."
Like this? Get the next one in your inbox before it hits the blog.
I didn't write a line of code.
It created a brand from scratch. It created the pages. It created the scripts that did the math. It set up the database. It wrote the tests. It found a bug in its own logic and fixed it before I noticed.
I was sitting there talking to it, watching pages and components appear on the screen. Things that used to take agencies months were taking hours.
A few days in, a working version of the €100k platform existed. Different code, same outcomes. In some places better, with fallback paths and edge cases I'd never have thought to spec.
Holy shit. It works.
The feeling came back
Same flow as the floor with the Duplo. Brick by brick. Watching something come alive. Time disappearing.
It hit me that this was exactly the reason I'd gone into marketing in the first place. Marketing was a way to make things and watch them come alive on a screen. A website. A campaign. A video. A banner. A brand. Creative output that landed somewhere visible. The economic part was always a side effect. I just liked making things.
No meetings in between. No call queue. No briefings. Just describing what I want and watching it show up.
I can finally make the things I've spent twenty years not being able to make.
I feel like that kid again. Except now the box has no walls.
What I am, what I'm not
I'm not pretending I'm a developer. I'm not building a software company from my laptop. I'm not the right person to ship a high-availability backend or audit a security model.
But I can build useful tools. I can build automations that save real hours. I can build apps that make my clients' work less painful. I can build the things I used to brief a developer to build, and I can ship them in a fraction of the time and cost.
For me, that's enough. The economic upside is real. The hours saved are real. But the main thing, the thing I keep noticing, is that I'm just building again.
I'm not the only one
A whole group of non-coders are discovering this. Marketers, lawyers, accountants, founders, support leads, consultants, ops managers, a clinical pharmacist in the NHS, a senior associate at a Magic Circle law firm, a marketing director at a 12-year-old construction company. All of them building apps, agents, and automations with Claude Code, Cursor, Lovable, Replit, Codex, the rest.
The numbers some of these people are putting up are absurd. €4M in projected savings here, $25,000 a month in agency costs gone there, 34,000 hours of manual work automated by an operations manager who originally got into automation to send Discord notifications for a video game.
Yes, it saves time. Yes, it creates real value.
But for a lot of us, that's a side effect. The main thing is we're making things again. And I'm sure they feel like that kid too.
Why I'm building Agenteers
I'm calling these people Agenteers.
An Agenteer is the person on a team who spots manual work, builds the agent or app or automation that fixes it, deploys it, and operates the running system. Mostly without code. The role doesn't have a name on org charts yet. We're naming it. The full case is in the canonical essay: What is an Agenteer?
The proof is twenty-plus people already doing this work, named and profiled in the field guide. Different industries. Different tools. Same pattern.
If you can brief a colleague, you can build an agent. The bar is that low. The ceiling is that high.
If this is you
The newsletter goes out every week. It's where I write the next version of all of this. The builds, the failures, the people I'm meeting, what's actually working. Subscribe here.
If you've got your own version of this story, a wall that fell over for you, the moment you realized you could build the thing yourself, DM me on LinkedIn. I want to hear it. Some of those stories will end up on the show.
Kasper Vancoppenolle works full-time as an Agenteer. Currently as a consultant, building agents, workflows, and AI systems for multiple companies. Belgium-based, audience global. agenteers.ai is his attempt to gather the others.
