This Open-Source Framework Builds a Self-Writing Industry News Site — No Coding Required
What this project does
AIHOT is a website framework that finds trending topics on its own and writes a daily digest for you. In plain terms, you feed it a set of sources — RSS feeds, news sites, or MCP endpoints all work — and every day it quietly goes out, collects the latest material, uses a large language model to strip away the noise, then independently scores the survivors twice over, picks what is genuinely worth reading, and turns it into Chinese headlines and summaries. Better still, it automatically clusters stories from different outlets that describe the same event into a single event card, ranks them by how many sources are talking about them, and every morning spits out a nicely formatted daily report. Swap the sources and the curation rules for your own industry and it becomes your own industry hotspot site — law, HR, finance, precious metals, anything you care about.
Why it blew up recently
The repository was only created on 2026-09-28, yet it crossed 3,100 stars within roughly 48 hours, which is a genuinely fast climb. It did not go viral because the tech is revolutionary; it went viral because it hits a long-standing pain point. People working in law, HR, and finance often know far better than anyone else what counts as hot in their field, but they cannot write crawlers and have never touched prompt engineering, so they watch other people's AI newsletters update daily and feel locked out. The author is himself a designer by trade who, by his own admission, barely read code half a year ago; he rebuilt the whole framework together with AI and open-sourced it, laying bare the curation pipeline, the clustering algorithm, and the exact prompt text used for scoring. That posture — handing over the methodology instead of a black box — is devastatingly attractive to anyone who wants to replicate it.
Technical highlights
The most compelling part is that AIHOT makes how we decide what is trending completely transparent. The repo is not just a website and a backend; it ships the actual files that define which sources enter the first filter, the literal prompt text the model uses when scoring, and the threshold values that decide whether an event makes the cut. None of that is buried in a server-side hardcoded rule. The daily workflow runs like this: pull sources, then the LLM does a first rough pass, then the same content gets scored independently a second time — a cross-check that reduces single-model bias — then a clustering algorithm merges stories pointing at the same event into one card, then it ranks by how many people are talking, generates Chinese headlines and summaries, and publishes the morning digest. The stack is TypeScript on Node.js 24 with PostgreSQL 17, and Docker Compose wraps it all up so you can get it running in one pull. Thoughtfully, it ships with 18 public overseas AI news sources as a demo, so you see real output the moment you clone, then swap in your own industry feeds at leisure. The prompts are not minified or obfuscated; you can read them, question them, and rewrite them to match the taste of your own readers, which is the part most other tools refuse to expose.
Who it is for
The sharpest fit is the domain expert who cannot code content entrepreneur: a legal-tech blogger, an HR-weekly creator, a precious-metals analyst. They already hold the sources and the judgment; they just lack a shell that runs automatically. Next, it suits developers who want to build a vertical news site — AIHOT hands you a battle-tested skeleton and saves weeks of designing curation logic from scratch. For small teams it is also a ready-made low-cost internal intelligence board, the kind of thing that usually takes a quarter to build and is abandoned the moment the intern leaves. Because the whole thing is editable, a non-technical founder can hand the repo to a freelancer and say build my industry newsletter, and the freelancer actually has a clear spec to follow instead of guessing.
Quick try
Requirements are Node.js 24, PostgreSQL 17, and Docker Compose. Grab it in one line:
~~~
git clone https://github.com/KKKKhazix/AIHOT && cd AIHOT && cp .env.example .env && docker compose up -d
~~~
Once it is up, open the local port to see the daily report with the 18 demo sources, then follow docs/customize.md to swap in your own sources and thresholds. The demo sources are real and public, so you will see a plausible daily digest within minutes of starting, which makes the abstract idea of self-hosted curation concrete fast.
How it compares
Against other AI news aggregator projects, AIHOT's difference is editable curation logic rather than more sources. Many aggregators, various RSS radars and newsletter generators, do the fetching and summarizing but still let the author decide what counts as hot; AIHOT opens that layer completely, handing you the editorial judgment standard. The trade-off is that it is not a finished product you switch on; you must invest effort tuning your own sources and thresholds. Unlike HN or Reddit-style community voting, it follows a model scoring plus same-source clustering route, which fits vertical, niche, community-less industries far better. It ships under the MIT license, so you can rebrand and go commercial. If your goal is a generic tech-news page, the bigger aggregators win on breadth; if your goal is a focused feed that reflects your own expertise, AIHOT is the rare tool that treats your judgment as the product instead of a bug to be automated away.