I train programmers to work with AI—and at the same time I like to test what AI can and cannot do. As part of this ongoing “research,” I built a small project called Actisto.com, with major help from GPT-5.3 Codex.
So what exactly is it?
At its core, Actisto runs on a relatively simple prompt:
Select the most interesting news story from a given area, cite the source, and comment on it sarcastically.
Behind the scenes, the system is composed of several MCP servers:
- a search engine that finds the news,
- a glossator that adds commentary,
- a formatter that prepares the content, and
- an uploader.
The final articles are automatically published both on Actisto.com and on a dedicated Facebook page: https://www.facebook.com/61588259453607. Each channel requires its own format, so the system adapts the output accordingly.
I give the AI a lot of freedom in choosing which news story to cover. Every third article should also include an image. The actual selection of the story is entirely up to the AI.
Of course, this is not the most useful application in the world. Still, the AI can sometimes be surprisingly funny—or unexpectedly sharp—and occasionally it picks a story that really is the most important one of the day.
But the interesting part is that by simply changing the prompt, the whole project could turn into a much more serious kind of “newspaper.” AI could not only aggregate news but also react to it, analyze it, and draw conclusions.
Maybe I should ask the AI itself what the next step for this project should be. Or perhaps one of you has a better idea.
For now, Actisto is available only in Czech. However, I’m preparing additional versions in English, Spanish, and German, each focusing on region-specific news.