Back to browse
GitHub Repository

📡 Your own AI-powered news radar. Generates daily briefings in English & Chinese. | 用 AI 构建你专属的新闻雷达

5,939 starsPython

Horizon – My AI-powered personal news aggregator and summarizer

by gg_equal_G·Feb 22, 2026·1 point·0 comments

AI Analysis

MidShip It

AI news filter for your feeds; 10+ similar tools exist (Feedly, Nuzzel, Curated).

Strengths
  • Multi-source aggregation (GitHub, HN, Reddit, RSS) with deduplication across platforms
  • Supports Claude, GPT-4, Gemini, DeepSeek—flexible API abstraction
  • Genuine use case: information overload is real, and framing as 'personal intelligence agent' resonates
Weaknesses
  • Core idea is solved: Feedly, Nuzzel, Curated, and countless AI-powered newsletters already filter feeds by relevance
  • No evidence of novel scoring methodology; 'LLM scores based on interests' is table-stakes for 2024-2025
Category
Target Audience

Researchers, academics, information-heavy professionals

Similar To

Feedly · Nuzzel · Curated

Post Description

Hi HN,

I built Horizon because I was drowning in information overflow. I follow dozens of RSS feeds, subreddits, GitHub users, and, of course, HN itself. I found myself spends hours "doomscrolling" just to find the 3-4 things that actually mattered to my work and interests.

Horizon is a personal intelligence agent that automates this. It doesn't just aggregate; it filters and enriches.

How it works:

Scrape: It pulls from GitHub (releases/events), Hacker News, Reddit, and any RSS feed. Score: It uses LLMs (Claude, GPT-4, Gemini, or DeepSeek via OpenAI-compatible APIs) to score content based on your specific interests/criteria. Enrich: For high-scoring items, it performs a web search to add context that may not be in the original snippet. Summarize: It generates a clean, daily Markdown report and optionally deploys it to a static site (GitHub Pages).

Demo: https://thysrael.github.io/Horizon/

Similar Projects

Developer Tools●●Solid

Snapbyte – personalized email digests from HN/Reddit/Lobsters

Pick sources, select from 135+ topics, and get a short, Gemini-powered summary delivered on your schedule — the workflow is exactly what you'd expect from a focused digest. The implementation choices (trafilatura scraping, small services on k8s, gemini-3 for summaries) are pragmatic and show real engineering; what I'd like to see next is clearer scoring/ranking methodology and more source coverage before it unseats Mailbrew or daily.dev.

Solve My ProblemSlick
onatm
203mo ago