Horizon

Horizon is an open-source news radar from the GitHub user Thysrael, who calls it "Your own AI-powered news radar." You list the sources you follow in its config, such as RSS, Hacker News and GitHub, and it gathers the posts, filters and dedupes them, has the AI model of your choice score and summarize them, and produces a daily briefing in English and Chinese. What stands out is its profiles: each kind of content gets its own scoring rules and writing style, so news gets "why it matters" and engineering deep dives get "what I can use," with built-in profiles for tech news, tech blogs and AI creator material. Briefings can be saved as Markdown, published as a web page, sent by email or pushed to a webhook, and an AI assistant can call its steps over MCP. AI Inspiration World's own source radar runs on Horizon's fetching and dedupe steps.

Who it's for:people who follow lots of sources every day and want a briefing filtered to their own taste. You'll need to be comfortable with the command line and config files; readers who don't write code may find it hard going.

Before you start: it needs Python 3.11 or later and installs with uv sync or pip install -e ., or runs in Docker. Scoring and summaries need an AI model: cloud services such as Claude, GPT, Gemini and DeepSeek, or a local model through Ollama. With a cloud model, the selected news content is sent to that provider and billed at its prices. We only recommend using it with regular sources like RSS and Hacker News. It also has an optional X scraper that asks for your X login cookie and uses anti-detection tricks to mimic a real person browsing; we don't recommend it. The software itself is MIT licensed, free and open source.

Based on the project's public documentation. Not tested by AI Inspiration World.