By Michalis Strouthos
Co-Founder & CEO, Cocoon Creations
Last week, we were on a family holiday. Thessaloniki and then Crete, together with our two children. Before we left, based on an idea I got from a good friend of mine, Paris Thomas, I asked ChatGPT and Claude to help us organise the trip. I gave them a rough idea of what we were interested in, how many days we would be staying in each city, our flights and location of our stays, and a few constraints because of the children. They both created an itinerary with suggestions for each day, then they created a personal website with the itinerary, links to google maps for each location, weather forecasts for each day along with transportation suggestions and estimated costs for them.
So far, nothing particularly remarkable. AI is now pretty good at creating these kinds of plans.
Then I thought it would be nice if we had something more than just the photos in iCloud: a small digital diary of the trip. Something the children could open years from now and remember what we had done. So I asked Claude to help me build it.
The idea was simple: every day Elena and I would upload our photos and videos to a specific folder in Google Drive and keep a few notes about what we had done on that day, funny encounters, etc. At the end, the AI would use all that material, analyse the photos and videos, take the locations of the content and combine it with the daily notes to create a website documenting our trip with a different brief for the parents (more detailed) and a different one for the kids.
And it did. It even filled out the gaps from our notes based on what is was seeing in the photos, location of them and the times.
Today we have a proper website documenting our holiday in Thessaloniki and Crete, already installed on both children’s iPads. I didn’t have to write any code, design a website from scratch, or learn a programming language. I simply described what I wanted and AI did 99% of the work.
There is something I should explain here: I am not a programmer. I have technical understanding, but I don’t write code. I can, however, understand roughly what a technical tool is suggesting and, more importantly, I can stop and ask: Why?
That turned out to be quite important.
At one point, the AI asked for access to Google Drive. I could very easily have clicked “Allow” and carried on. I didn’t. I created a separate Google Drive with no other personal content in it and gave access only to the specific folder needed for the project.
Later, the tools suggested making the folders public so they would be easily accessible. Again, I said no. We created separate Gmail accounts for the children, connected through Family Link, and gave them view-only access.
It wasn’t a complicated process. The important thing was understanding that “public” means public — accessible to anyone.
The same thing happened with the way the website was built. The first solution proposed included Node.js, a database and quite a bit of infrastructure for a normal web application. Except we didn’t need a normal web application. We needed a digital album that we would create once and never change again. So I asked for it to be built as static HTML instead. The application became a collection of simple files, hosting became significantly easier, and a whole lot of infrastructure that was never needed in the first place disappeared.
Then came publishing. The AI could handle that part too, and the easiest option would have been to publish the website on the open Internet. There was no reason, however, for a website containing photos and information about our family holiday to be publicly accessible. So I took care of the hosting myself and the website is protected by a password. Most of the photos are also not stored directly on the website; they remain in Google Drive, with links to them from the site.
Looking back, the interesting thing is that none of this was actually difficult if you understood what all these things meant. If you didn’t, you would simply accept what the AI suggested: full access to your personal Google Drive, public folders, Node.js, a database just because these are things you don’t understand and therefore have to trust the tool to decide for you.
This, in my opinion, is one of the most important changes AI is bringing to software development. For the first time, we can give almost anyone the ability to create software without necessarily knowing how to program.
And that’s remarkable.
At the same time, however, it means that decisions that until recently would have been made by a developer, a system administrator or a security engineer can now be made by someone who is simply having a conversation with a chatbot.
The problem is that AI doesn’t necessarily know our goals or the context in which we are asking it to do something.
It doesn’t see a Google Drive the way we do. To the AI, it is a storage location that needs to be accessed in order to complete a task. To us, it might contain photos of our children, tax documents, personal files and years of our lives. It doesn’t see a public folder as a potential privacy risk for a family. It sees an easy way to share a file. And it has little reason to think that a simple website might not need a database, a server and an entire application architecture.
It can suggest something that works perfectly and still not be the right solution for you.
This is perhaps where technical knowledge doesn’t disappear because of AI. It changes form.
You no longer necessarily need to know how to write the code. But you do need to know enough to understand what the AI is suggesting and ask the right questions.
Why does it need access to this? Who will be able to see the data? Why does this need to be public? Do I really need a database? Where will this be hosted? What happens if something goes wrong?
And perhaps this will be one of the most important skills we need in the coming years: not necessarily the ability to write software, but the ability to exercise technical judgment.
This is also something we are increasingly seeing in our work at Cocoon Creations. More and more people and businesses can now create software themselves with the help of AI. The question that follows is no longer simply “Can we build it?” It is “Is this what we actually need?” and “Is this the right way to build it?”
Perhaps AI doesn’t make people less useful in technology after all. It simply changes the point at which human expertise is needed.
Our website is now finished. The children can look back at our holiday from their iPads and, most likely, years from now they’ll laugh at the photos and the stories we put together.
AI did almost all the work of creating it.
But the most important part of the project wasn’t the code the AI wrote.
It was all the times I had to tell it:
“No. We don’t need that.”
Because AI has made software much easier to build.
It hasn’t made good technical judgment any less important.
Cocoon Creations helps organizations understand how emerging technologies can create practical business value, from identifying opportunities and redesigning processes to deciding what should be built, bought or integrated. If you’re wondering what your organization would look like if you designed it today around the technology that is now available, that’s a conversation we’d be happy to have.
