There is a tool sitting on a friends browser in Spain that transforms apartment booking data into the exact format required by the Spanish Registradores — the property registry system. It works. It is fast. It is useful.
I built it.
I used OpenCode — an AI coding agent that can build applications from scratch based on natural language prompts. I have zero professional experience with JavaScript. I wrote my first line of it in 2026, prompted OpenCode to help, and got something working in one shot. A couple of refinements later, it was done.
This matters — and not for the reason you might think.
The Experiment
A friend runs a small real estate business in Spain. He was struggling with a painful manual process: extracting booking data from various sources, transforming it into the specific CSV format the Registradores require, and doing this repeatedly as bookings came in.
He mentioned it casually. I said, Let me see if I can help.
I used OpenCode — an AI coding agent that can build applications from scratch based on natural language prompts. I described what he needed. It generated the code. I tested it. It worked. Almost.
A couple of refinements — a field that was not quite right, a CSV separator that needed adjusting — and it was ready. Total cost: zero (OpenCode provides free access to AI models through their Zen provider). Total time: a few hours, most of which was me waiting for my friend to test rather than the AI being slow.
My friends tool cost nothing because I did not charge him, and the tool to build it cost nothing because OpenCode is free and open source, and they provide free access to AI models. The entire application — data transformation, CSV generation, error handling — has a total cost of ownership of zero euros.
That is not a demo. That is not a proof of concept. That is a real thing solving a real problem for a real business.
What This Actually Means
I want to be careful about what I am claiming and what I am not.
I am not saying I am now a software engineer. I am not saying AI will replace developers. What I am saying is something more interesting: the barrier between can build and cannot build has collapsed.
I do not know JavaScript. I cannot read the code fluently. I cannot debug it from memory. But I can describe what I want and get something that works. For someone who understands the problem domain — in this case, hotel booking data and Spanish registry requirements — that is 80% of what matters.
The other 20% (syntax, structure, debugging) is now optional if you have the right AI tool and are willing to iterate.
Why This Changes Everything
Consider what this enables:
- Small businesses solving their own problems without hiring developers
- Domain experts building tools in their own field without learning to code
- Prototypes that become production systems without the traditional build process
- Cost — zero. Not low cost. Zero.
My friends tool cost nothing because I did not charge him, and the tool to build it cost nothing because OpenCode is free and open source, and they provide free access to AI models. The entire application — data transformation, CSV generation, error handling — has a total cost of ownership of zero euros.
That is not a demo. That is not a proof of concept. That is a real thing solving a real problem for a real business.
The Implications for Leaders
If you are leading an organisation — any organisation — here is what this should tell you:
The question is no longer Can we build it? The question is Can we describe what we need?
This shifts the bottleneck from technical capability to domain expertise. The people who understand the problem are now the bottleneck, not the people who can write code.
At Wall Street English, where I lead product, engineering, and design, this changes how we think about tooling. It changes how we think about training. It changes who we empower to solve problems.
Not everyone needs to learn to code. Everyone needs to get better at describing what they want.
Actionable Takeaways
- Empower domain experts — The people who understand the problem can now solve it, even without coding skills
- Redefine technical debt — The real debt is the gap between what you can describe and what gets built
- Start small — Find one painful manual process in your organisation and see if AI can solve it in hours, not months
Leave a Reply