Most AI news tools try to solve information overload by summarizing more content, faster.
That was not the product I wanted to build.
I wanted something closer to a personal news radar: a system that could watch Hacker News, Reddit, RSS, GitHub, Telegram, and other sources for me, reduce the noise, connect the context, and still leave room for human judgment.
So I built Horizon.
Horizon is an open source AI-powered news radar that generates daily briefings in English and Chinese. It fetches from your chosen sources, deduplicates overlapping stories, scores and filters them, adds background context, pulls in community discussion, and delivers the final briefing to GitHub Pages, email, webhooks, MCP, or local files.
But the real idea behind it is simpler:
I do not want to outsource my taste to an algorithm.
Good news is scattered, bad news is endless
The links worth reading rarely come from one place.
You might catch an interesting release on GitHub, a thoughtful thread on Reddit, a sharp comment on Hacker News, or a niche RSS post before the topic reaches the mainstream. At the same time, there is an endless amount of repetitive, low-signal content everywhere.
I did not want a tool that just compresses all of that into bland summaries. I wanted a better first pass: something that helps me spend my attention on the right stories without pretending AI should make every editorial decision for me.
Horizon is not just another summarizer
What Horizon does is practical:
- watch your own mix of sources
- merge repeated stories across platforms
- score and filter items with AI
- enrich unfamiliar topics with background context
- surface the original discussion around a story
- publish a bilingual daily briefing
That part is useful, but it is not the whole point.
The point is that news still needs human taste.
The best sources are often personal, weird, niche, and constantly evolving. A good information diet is not only a ranking problem. It is a judgment problem. It depends on who you trust, what communities you follow, which comments actually change how you interpret a story, and which small sources deserve more attention.
That is the part I care about most. AI should help reduce repetition and noise, but the final signal should still feel human.
I also built a website for sharing and exploring sources
This is why Horizon is not only a summarization pipeline.
I also built a companion website for sharing and exploring news sources:
- to make good sources easier to discover
- to let people share the feeds, communities, and links that actually help them stay sharp
- to make source discovery more communal and more human
That site matters because better inputs create better briefings.
If all AI sees is a generic pile of links, the output will usually feel generic too. But if the source graph reflects real curiosity and strong taste, the result becomes much more useful.
To me, that source-sharing layer is a core part of the product, not a side feature.
Live site: https://horizon1123.top
The product idea I care about most
I think there are already enough products that say, "give everything to AI and let it decide what matters."
I want the opposite balance.
My ideal workflow is:
- humans choose and refine the source graph
- AI handles collection, cleanup, filtering, and enrichment
- humans keep the final layer of taste
That is the version of AI I find most interesting: not replacing taste, but supporting it.
Why I open sourced it
I wanted Horizon to be inspectable and hackable.
People should be able to:
- run it themselves
- customize sources, thresholds, models, and output channels
- adapt it to their own reading habits
- contribute new sources back to the wider community
That felt much more honest and much more useful than building a closed black-box feed product.
Links
- GitHub: https://github.com/Thysrael/Horizon
- Live demo: https://thysrael.github.io/Horizon/
- Source-sharing site: https://horizon1123.top
If you care about building a better information diet without losing the human part of the process, I would love to hear how you think about it.
Top comments (1)
Some comments may only be visible to logged-in visitors. Sign in to view all comments.