kkkkhazix/aihot is a site framework that gathers multiple information sources, curates industry-specific hot news, and produces a daily report every day. It is written in TypeScript and follows the MIT license. No information on the latest release was given in the public repository.
The engine first collects information and makes a preliminary selection, then edits the daily report through two independent rounds of scoring and event clustering. It groups together reports that cover the same event, and per-event heat reflects each independent source only once within48 hours. Reports older than 24 hours decay by half. It compares the titles and summaries of the past two weeks as vectors to find candidates for clustering, and falls back to character overlap when no vector service is available. The prompts and selection criteria live in markdown files under industry/prompts/ instead of in code, so the criteria can be changed to apply the framework to other industries. Eighteen public overseas AI news sources for the demo are included.
That said, the center of this design is a collection and curation flow built from chained large language model calls, and the novelty of the algorithm itself is limited. The repository does not include the operating service's model leaderboard, Codex reset monitoring, the per-topic event timeline, or the actual source list and operating data. The author has stated that they are not a professional developer, so the published design can be read as a reference case for a collection pipeline, but it does not demonstrate the completeness of the operating service.