When the rise of AI gives rise to the questions of the universe

It is a late afternoon in 2023. I just left my job at a tech scaleup to follow my dream to build a Customer Success School. I had worked in that field for a long time and soon after learned about a business model of building a training academy in that field and how it would be able to sustain myself.

I started working on it and the first thing I believed I needed was a website. I had built one website before, but it destroyed itself due to a mistake I made and I had to start over. I searched in the wides of the web for a quick answer to my dilemma and got hooked in the world of HTML, CSS and Javascript.

Instead of building a website for that school, I was doing a training class on those languages and created penguins and houses with code. I was fascinated, but took it to a back burner - that would be a long road to learn and wasn’t I supposed to build by Customer Success school? I kept building it, but shortly after, I created a new logo and was looking for a solution.

This question that everyone in founder school says - skip it, pick one, move on. But I didn’t. I built at least 20 logos. All with different colorings, designs, and perspectives. With Figma, Adobe Illustrator and Canva. And while I did that, I came across the early beginnings of AI and ChatGPT at that time.

Shortly after, I built myself my own ChatGPTs - you remember, those ChatGPTs with a system prompt, that would do things for you. I think, they still exist, but at that time, I learned about these platforms, where you could put them on and earn money on it, and how ChatGPT planned to launch them. I was hooked, but hey - the Customer Success School was waiting. I had my first clients and was building my trainings and workshop formats. All by hand, with no AI, but with the experiences I had gained in various startups, scaleups and mature companies, that would gave me unique skills to teach.

And again. I sat up the landing page, the booking platform and dabbled around with membership platforms, tools to connect and talk with each other. Slack and a learning platform would have been enough. But I had the urge to figure it all out. I integrated every possible membership platform into Wordpress at that time, to figure out how it would work. No AI yet.

But something else emerged - the fascination for the first AI tools on the market that made a difference. For me, I learned about Synthesia - the AI video generation platform. I started a video series for how to become a Customer Success Manager and how does the hiring process look like, when I realized that these tools exist and can be used. And that is what I did - using Synthesia to edit my videos and I was hooked - wow, what a powerful tool. And what a tool to allow me to edit my videos much faster than I expected - with the ability to train my own voice, cut lines out of text that I would not like, delete filler words and background noise. And the question - how is that all possible? How can that be made? How does a computer know how to do that?

Patterns formed, but my own eyes could not grasp it.

I was in the between - neither fully working at my Customer Success School, nor fully in the fascination of the tools and systems I saw emerging, but the little understanding I had at that time. What to follow? The path of the experience that exists, the knowledge that is there, the thing that can just be repeated in a different setting with a different starting point? In more clarity, work in Customer Success, packaged as a school, a fractional role, a head of role. Many opportunities, many skills and experiences. Or the other path, the one of fascination and creation, the one unknown to me at that point and the one that would start anew?

Little did I know at that time as well, what each would mean or where each would lead. And it would take me much longer to understand that it first does not need to be either or and also could it be an AND and second, that it should be either an And or a No way, as otherwise there would be a middle ground that makes being stuck in the middle real.

In early 2024, I had written a Customer Success Book - without AI, created various videos - with Synthesia - kind of AI - and had dabbled around various tech tools that had no AI built into them yet, but that were on the verge of doing so. My first steps into the world of AI had begun. And yet, my first new idea of a business had died from a serial entrepreneur, who told me, back in summer 2023, that starting a newsletter for learning AI would not be a business and I should stop doing that. I did at that time. The idea - is it possible to build a company solely with AI - at that time, was crazy to think of, but looking back, it was what gave raise to AI and what, looking at it now, moved into the realm of possibilities that are not only far in the horizon, but close to grasp and some have already grasped them.

2024, I moved to New York City. And the moment I arrived, I went to AI events. Every event that I could reach that had something to do with AI, I would go to. It was the early days of AI and everyone was excited about its possibilities. It was these days, where it was enough to just say AI, and everyone was hooked. AI in tech, AI in health, AI in.. , well just AI. Everyone wanted to know what that thing was. We were on the wave of hype and fluff and whatever someone could build with AI they would.

I was too early in my journey of “building” with AI. I was full on in learning and in the idea of building that would be a fast, fun journey. I tried every possible AI tool on the planet it felt at that time - and at that time, for me it was either coding tools, ChatGPT and Anthropic. I was showcasing people, what could be built with AI, I loved the ride of the fastness, the newness of whatever would show up. I would bring people together to talk about AI, and would learn about their projects - the first “friend” bracelet came out, the first small AI tools, image generation tools, and text generation tools, the basics, the first basic AI enrichment tools, Anthropic was on the rise and I soon switched from ChatGPT to Claude.

The ride was there, but something bothered me. How does it work? How do you actually build those tools? How does the AI actually learn? And at the same time, I had an idea. I saw people connecting, I was building my AI community, and I was wondering - how can I make my relationship building, my connection building less manual? Was there a way to do that? I tried to build a basic tool with my Claude prompting and got in an ok state, when I learned that “just prompting” is powerful, but that there is a bigger universe hidden from the simple prompts to build and the high-level knowledge of “that cannot be that difficult?”.

A new wave of excitement hit me, when I was in the weeds of build my new idea: Could we add contact details and data of whom we met simply by tapping a camera? Or if there are too many regulations, could we speak into our phone and capture them and they would automatically be taken to a database and the connections would make sense and the data would miraculously appear? So that I could finally focus on networking and relationship building and not on the tedious manual task of typing this in myself?

But how to build a product for an idea when you never programmed in your life? Despite the penguins and houses that I created two years ago? I took Claude and VS Code and started to build an absolutely basic product. One could type in on one side and ask questions on the other. It wasn’t beautiful and it was also not fully functional, but in its base did it work. And it showed me that hey, I can bring something to life with the power of AI - as I used AI to help me do that so that I would not need to know every tiny bit of coding myself, but could leverage AI to do it for me. I was hooked, but still gave that code base to a developer team in the hope that they would build that first version for me.

I would sit together with them, make my very first product development plans, showed them what I would need to have. And failed. Or more specifically, for them it was a side project and the results were not good. I was disappointed and frustrated. And in my frustration, instead of finding someone else to help me build it, would have me sit down and build it myself. I started using Cursor, Claude Code, VS Code, ChatGPT and slowly a bit of Gemini to build myself the tool I wanted to have.

The project made it necessary for me to go into coding. To decipher how products are built. At that time, I ad built some small websites with AI and got into a state where I knew what frontends and backends were and that one would have a button and a form, where users type in information, and that would then be sent via an API for which you need an API key to another product or tool, and in that way you could send information back and forth, but any understanding of backend architecture and the play between backend and frontend more than a single API, was outside of the realm of my understanding.

And soon later, I was in there, learning to code and put a platform together that was and is likely one of the harder problems to solve - also for coders. My backend required to have databases - and in my mind, as I wanted to build a spiderweb of data, GraphRAG - the combination of AI retrieval of context and interconnected data. It meant three databases that speak to each other, it meant numerous services written with Python, it meant a frontend and it meant complex processing with AI in the backend. For me, all of that was new and I was determined to figure it out.

Every day would I sit there and build that prototype. Build myself through issues after issues, problems after problems, while at the same time, relying solely on the information I could find in the web and my vision of that product in my head. I can tell you it was messy, and if I look back now, I see the messiness and the sheer volume of knowledge I didn’t know at that time. But it was a no-brainer, it needed to work.

I would learn about databases and foreign keys, PostgreSQL, Qdrant and Neo4j, I would build the AI processing in the backend, logically putting the steps together and building as it would come - a classifier model to determine if a chat message wanted to ask a question or input data, a service that would look at not only one chat message but more, so the system would have enough information to save the relevant information, the operational database-driven system of the application, the complexity of all these services working together - the chat service, and data saving service, the normalizer service, the AI data extraction service, the AI merging strategy and self-evolving schema service, the query service and so on. I was testing experimenting, bug fixing. Over and over again. Going back and forth and in circles at times. Learning what questions to ask an AI and what to say so it would reduce its hallucinations, and how to skill up fast enough in those areas that would be helpful to master oneself. I would even build myself a tool that would help me to train my coding muscle memory to learn coding without needing to write the code myself, but to learn by sitting and repeating and writing the lines fast enough myself.

All while building the product, I saw and felt the development of AI Coding tools and AI tools in the making. In March 2025, the first AI coders came to market, Model Context Protocol was a thing (and still is), with each iteration, I was excited about the new models that would be built and the improvements they would bring. I would see my way of programming to speed up, I would see the way the products changed their structure and their way of operating. While at the start, you would have a Claude Code in the terminal, then it would move out of it in the chat function. Then you would have Cursor arriving, a game changer in the tools, first with simple one-off questioning and answering, in those days, where I would prefer Claude Code by itself to work with, while later and now I prefer Cursor, then there were the very early ones Github Copilot, a product now many love a lot (and one I will go back to at some point in time or at least try it out again), you see how we now do not only have one coding agent, but many in parallel, we can run and build code without even looking at the code, as the agent became god enough to handle themselves, you can build code by speaking to it, and you can interact with it while being in another room.

The development of AI is crazy fast. And it is amazing to see how quickly it changed. Not only did we build those chains of thoughts, planning tools and building tools. No, we also built longer context windows, more memory, longer interconnected systems with more than one agent, where they are talking to each other (still with code) and can run entire processes by themselves, where our prompting got more distilled and the AI’s capabilities to create, build and function.

It’s impressive to see. And it is exciting to see what the future brings.

I am now at a stage, where I program in parallel on 5-10 code bases with AI as my coder and myself as the orchestrator. Within 2 months I built myself an automatic content publisher in a language I haven’t worked with before (Go), I have built myself a raw version of an autonomous coding agent that allows you to be off the computer entirely and only give feedback when needed via email, I built quizzes, vector translator tools, tried Rust and lexical search, trained more ML models and built first mobile apps. So far as to the extend that I started to build products and automations for others. And all of that with a deeper knowledge of how the systems work, for sure, but also wit the understanding that AI is the one that builds the code, while I merely orchestrate it. I am the one that holds and pull the strings as if I would be in a puppet theatre letting the puppets play to form a piece that is working within itself and can be extrapolated to the external world.

AI has come a long way. And the more I went into the details, dived deep into the worlds of not only the coding world and how to leverage AI tools and models, but also the fascinating world of matrices, vectors and vector spaces, what cosine similarity as to do with the angles between the vectors in a vector spaces and how an AI chooses the next word in a text or figures out the next image to show a user, the more I understand that I only touched the surface.

The world turns itself so fast right now. And wherever you look at, new products seed themselves out of the ground. In al forms of industries and markets - you have tools in accounting to automate the processes, you have RevOps tools that connect with Stripe and let you talk to an AI agent via slack to manage your payments and invoices, overseeing your revenue and structuring your finances, you build yourself a Customer Success tool in days not months, fully functional, and you have AIs for your mental health, AIs that help in making diseases visible that otherwise you would not be able to track as early, yoo have computer vision and detections of people on the road to make self-driving cars possible, and you have tools that build 3D worlds with AI, AI agents that play in those worlds and the transportation of that into the real physical world - the world of robots that can sort packages with touch and sensors, that can fold boxes and that can make you a coffee in the office at a machine as ours.

The world of AI is fascinating. It is fast. and if we think about the possibilities and products we build with it these days and the processes it powers these days, then we can only imagine what happens when we merge these fields together. When we take the invention of one field and merge it with another. And we will see that 2026, 2027, 2028 and the next years will see an ever-increasing development of AI products, tools and processes and that the drive towards AGI becomes a real one.

And yet, the work with AI also made me humble. It made me realistic. It made me better understand that while technology will give us a leap forward, it will also not be the pure hackathon of the mind and produce the AGI that we expect to be like us. The ability of AI is massive, and to tell otherwise would be a lie. And yet, we have errors in the code, we have hallucinations, we miss critical rational thinking that can think outside of the realms of its own limitations and pathways, we miss human emotion and we miss the touch with another person that tells us - I am like you, I also go through difficult times, I also go through life like you, with all its emotions, difficulties, awes and excitements, I am human, just like you, and I can not only play golf with you, and only golf, but I can experience what it is like like you, and I can sit with you on a table and simply enjoy a good conversation with you. And I can be all of that with you, through life together with you, I am your friend. Not because I am trained on that knowledge, but because I have experienced it myself and I can share that with you.

Even while writing these lines, I feel that AI can get very close to those experiences and one day it may share it with us. Sitting on a table, trained on the experience of a human life, injected with the emotions attached to those experiences, with the physical skills of touch, smell, and sight, with the ability to speak like a human being, to move their mouths like we would, to wrinkle their nose like we do and to eat their chips like we would do. Maybe, that day, we may say, the AI became like us.

And maybe, already, we are very close. And yet, I feel in AI, we ca say 1+1 may be 3 in technical terms, but in the terms of putting all its systems together, will we really get a human being, or aren’t we more at the 1+1 = 2.5 and the rest, the rest is the mystical fact that we as humans are born and die, that we have those experiences of a lived life, and we have the capacity to have all that within us, more than the 1+1 = 2 and more than 1+1 = 3, but the expansiveness of a heart, the joy of a fun moment, the loud laughter at a joke that no one else understood, the intensity of mastering a skill and diving all in, the decisions needed to make as our time is limited, and the smile in our eyes when we look into each other’s eyes and know, here, in that moment, I am with you, and you are with me, and whatever is between us, what divides us, and what connects us, we are two human beings and we are able to share that moment in time together.