Horizon: The Free, Open Source AI News Radar That Scores, Summarizes, and Filters Your Daily Tech Feed (2026 Guide)

Checking Hacker News, Reddit, Twitter, RSS feeds, and Telegram channels every morning is a losing battle. You either skim hundreds of headlines and miss the important stories, or you spend an hour reading and lose your whole morning. Standard RSS readers dump raw links into a feed and leave you to do all the filtering manually. Doomscrolling is not a productivity strategy, it is a trap.

Horizon is a free, open source AI-powered news radar created by Thysrael that acts as your personal editor-in-chief. It runs in the background, fetches hundreds of articles from your configured sources, deduplicates them, and uses an LLM to aggressively score each item from 0 to 10 based on your specific interests. It filters out clickbait and noise, summarizes the community discussion threads, and delivers a highly readable daily briefing to your terminal, email, chat app, or a static GitHub Pages site.

In this guide, you'll learn what Horizon is, who it's for, and how to deploy it with Docker or native Python and generate your first AI-scored daily briefing.

What is Horizon?

Horizon is an open source, AI-powered news aggregation pipeline that cures information overload by actually reading and evaluating articles instead of just listing them. Unlike a standard RSS reader that dumps raw links into a chronological feed, Horizon fetches hundreds of articles and posts from your configured sources, deduplicates the stories, and then uses an LLM to score each item from 0 to 10 based on criteria you define.

The scoring system is fully customizable through a prompt. You can tell the AI "I only care about open source AI tools and Python libraries" and it will ruthlessly score generic tech news as a 2 and specific AI releases as a 9, only showing you the top-tier content. This means the briefing you get each morning is already filtered to your exact interests, not a firehose of everything that happened overnight.

Horizon goes beyond just summarizing articles. It fetches the comment threads from Hacker News and Reddit and includes a summary of the community's reaction, adding vital context that the article alone cannot provide. An article about a new pricing model is only half the story. The community debate in the comments tells you whether users are angry, indifferent, or excited, and Horizon captures that for you automatically.

The final briefing can be output in multiple ways: directly to your terminal, sent via email, pushed to Feishu, or automatically published as a GitHub Pages daily website in both English and Chinese. You are not locked into an expensive subscription either, because Horizon supports any LLM including cheap models like DeepSeek, standard OpenAI and Anthropic APIs, or completely free local models via Ollama.

Who is it for?

Information junkies and researchers: People who suffer from FOMO but do not have the time to sift through thousands of daily tech and finance posts to find the actual signal.
Developers: Developers who want to track specific GitHub releases, Hacker News trends, and Reddit programming communities without getting distracted by social media timelines.
Bilingual readers: Readers who want complex English technical news automatically summarized and translated into Chinese or vice versa in a single daily digest.
Homelabbers and automators: People looking to self-host a fully automated news pipeline on a Raspberry Pi or home server using Docker and cron jobs.

What makes Horizon different from standard RSS readers?

Intelligent 0 to 10 scoring: You can customize the scoring prompt to match your exact interests. Tell the AI what you care about and it will score every article accordingly, filtering out everything irrelevant before you ever see it.
Community discussion summaries: An article is often only half the story. Horizon fetches comment threads from Hacker News and Reddit and includes a summary of the community's reaction, adding context like "users are criticizing the new pricing model" to every story.
Bring your own model: You are not locked into an expensive subscription. Use cheap, highly capable models like DeepSeek, standard OpenAI or Anthropic APIs, or run it 100% free and offline using Ollama with a local Llama 3 instance.
Rich delivery options: Horizon does not just spit out a text file. It can compile the news into a beautiful static HTML site, send you a formatted email before you wake up, ping your team's Feishu chat, or print directly to your terminal.
Bilingual output: The briefing can be generated in both English and Chinese simultaneously, making it one of the few aggregators that handles bilingual tech news natively without a separate translation step.
Free and open source: The entire pipeline is free and open source under the MIT license. No subscription, no vendor lock-in, full access to the source code for customization.

What you need before you start

Horizon is a Python-based pipeline that is easy to configure. Make sure you have the following:

Python 3.10+ or Docker: Depending on how you prefer to deploy. The modern uv package manager is highly recommended if installing natively.
An LLM API key: An active API key for OpenAI, Anthropic, DeepSeek, Gemini, Doubao, or a locally running Ollama instance. Ollama is the best option for completely free, offline operation.
A JSON config file: You will need to spend a few minutes defining your favorite RSS feeds, subreddits, and Telegram channels in a simple config.json file.
Git: To clone the repository from GitHub.
A scheduling mechanism: A cron job on Linux or Task Scheduler on Windows to run Horizon automatically each morning.

Step-by-step installation

Step 1: Clone the repository and copy config files

Open your terminal and clone the repository, then copy the example configuration files:

git clone https://github.com/Thysrael/Horizon.git
cd Horizon
cp .env.example .env
cp data/config.example.json data/config.json

Step 2: Configure your API keys

Open the .env file in a text editor and paste your API keys. For example, if you are using DeepSeek for cheap scoring:

DEEPSEEK_API_KEY="sk-your-deepseek-key"

If you prefer to run completely free and offline, set up an Ollama instance instead and point Horizon to your local endpoint. No API key needed.

Step 3: Define your news sources

Open data/config.json and define which RSS feeds, subreddits, Hacker News settings, and Telegram channels you want to track. You can also customize the scoring prompt here to tell the AI exactly what topics you care about and what to filter out.

Step 4: Run Horizon via Docker

The easiest and cleanest way to run Horizon is via Docker. Once configured, run the container to fetch, score, and generate your first briefing:

docker compose run --rm horizon

Alternatively, if you prefer native installation using the uv package manager:

uv sync
uv run python -m horizon

If you want to use the OpenBB financial news scrapers, install the extra dependencies with uv sync --extra openbb.

Step 5: Schedule it to run daily

To get your briefing automatically every morning, set up a cron job on Linux:

0 7 * * * cd /path/to/Horizon && docker compose run --rm horizon

This runs Horizon at 7 AM every day. Adjust the time to match when you start your day. The resulting Markdown briefings will be saved locally in your output folder, or dispatched via your configured delivery method.

Common errors and how to fix them

Error	What it means	How to fix it
Reddit or Twitter scraping returns empty results or errors	These platforms have strict anti-bot measures. Fetching data anonymously often results in IP blocks or HTTP 403 Forbidden errors.	For Reddit, ensure you are not polling too aggressively. For X/Twitter, Horizon relies on RSS bridges or specific API configurations. You may need to provide auth tokens or use a service like RSSHub to reliably bypass Twitter's login walls.
High LLM API costs	Asking an LLM to read, evaluate, and score 300 different articles every day consumes a massive amount of input tokens.	Do not use GPT-4o for the bulk scoring process. Switch your primary LLM endpoint to a highly affordable, fast model like DeepSeek-Chat, Claude 3.5 Haiku, or a local Ollama model like Llama 3 8B to keep daily costs down to pennies.
ModuleNotFoundError: No module named 'openbb'	You added OpenBB finance sources to your `config.json` but did not install the optional OpenBB Python dependencies.	Run `uv sync --extra openbb` to install the necessary financial SDKs. If you encounter wheel build errors, force binary installation with `uv pip install --only-binary=:all: openbb`.
Docker container exits immediately	The `.env` file is missing or the LLM API key is invalid, causing the pipeline to fail before fetching any articles.	Verify your `.env` file exists and contains a valid API key. Run `docker compose run --rm horizon` without the `-d` flag to see the full error output in your terminal.

Horizon vs commercial news aggregation services

Feature	Horizon (open source)	Commercial services (Feedly AI, Matter)
Software cost	$0, free MIT license	$8 to $12+ per month
AI model choice	Choose any API: OpenAI, Claude, DeepSeek, or local Ollama	Locked into proprietary vendor AI
Comment thread summarization	Yes, invaluable for Reddit and Hacker News context	Rarely supported natively
Bilingual output	English and Chinese natively	Single language only
Delivery options	Terminal, email, Feishu, GitHub Pages static site	Web and mobile app only
Best for	Developers who want full control over scoring and delivery	Users who want instant 1-click setup with zero configuration

Bottom line: Horizon is the ultimate antidote to information overload. While standard RSS readers force you to skim through hundreds of headlines, Horizon actually reads the articles and judges them based on your strict criteria. The inclusion of community comment summaries elevates it from a simple aggregator to a true research assistant. While Feedly AI charges $8 to $12 per month and locks you into their proprietary AI, Horizon is free, MIT licensed, and lets you use any LLM including completely free local models via Ollama. If you want a daily, high-signal briefing without the doomscrolling, setting up Horizon on a cron job is the best open source option available.

3 alternatives worth checking out

Omnivore (omnivore.app): One of the most beloved open source read-it-later apps, featuring excellent text extraction and organizational tools. While less focused on autonomous AI scoring than Horizon, it is fantastic for users who want a beautiful, mobile-friendly interface to manually save, highlight, and read articles later.
Huginn (github.com/huginn/huginn): The granddaddy of open source IFTTT alternatives. If you want to build incredibly complex, multi-step automation logic for your news feeds, Huginn lets you scrape websites, check for specific keywords, and trigger emails. It is more powerful than Horizon for general web scraping but lacks the out-of-the-box LLM scoring pipelines.
Feedly AI (feedly.com): If you do not want to mess with terminal commands, Docker, or API keys, Feedly's premium tier is the commercial standard for enterprise news aggregation. Its Leo AI lets you mute specific topics, track industry keywords, and summarize articles inside a polished web and mobile app. You trade control and cost for convenience.

Found this guide useful? Check out more open source projects and free tools for developers on Sudo Scout.