Ukrainian Fullstack Web Engineer Anton Demchenko talked about his pet project — an interactive map of «all Ukrainian YouTube.» He used the GPT 4o Mini model to create the map.
Ukrainian Fullstack Web Engineer Anton Demchenko talked about his pet project — an interactive map of «all Ukrainian YouTube.» He used the GPT 4o Mini model to create the map.
«This is what 10,000 Ukrainian-language channels on YouTube look like. This is a continent of music. And this is an archipelago of game reviews. Here we have sports. Of course, news and politics… Here we have a cluster with drones. It turns out that we have a lot of channels about Christianity. And this is all Ukrainian IT», Demchenko describes his map.
According to him, the interactive map contains almost all Ukrainian-language channels. Each circle on the map is a separate channel, and its size corresponds to the number of subscribers to this channel. All channels are divided into thematic categories and clusters.
«I did all this, of course, with the help of artificial intelligence and vibecoding. I myself am a developer and I really love all kinds of maps, graphs, diagrams, and recently I have been working with big data and artificial intelligence. I wanted to do my own little pet project and in the process I realized that I could do something interesting and large-scale,» says Anton Demchenko.
Aitivets noted that he had some difficulties collecting data, as YouTube’s API did not allow filtering channels by country and had «strange limits» on the number of requests per day. The freely available resources were not suitable for the amount of data or the price, so Demchenko decided to parse them himself.
For his sample, he took channels with over 1,000 subscribers in Ukrainian. His parser went to each channel’s page and collected the last eight video titles — they were needed to determine the channel’s category.
«To do this, I sent all 10,000 channels to GPT along with the video titles.
«I used the GPT API, the 4o Mini model — it worked very quickly and gave a good result. I spent only $1 on all 10,000 channels,» the map’s author said.
Artificial intelligence also helped group channels into clusters and place them on a map. Among the clusters were military, music, gaming, book, IT, and many others.
You can learn more about creating an interactive map in the video on Anton Demchenko’s YouTube channel.
Recall that in November, YouTube began testing a new feature, Your Custom Feed, which allows users to independently create their home feed based on text queries, rather than relying solely on the algorithm, likes, and the «Not interested» button.
