Analysis updated 2026-05-18
Follow a breaking news story as it evolves across multiple updates.
Aggregate Hacker News, RSS feeds, and Reddit into one importance-scored feed.
Ask an AI chat panel follow-up questions about a specific story.
| wislonl/newsline | 0xhassaan/nn-from-scratch | 3ks/embedoc | |
|---|---|---|---|
| Stars | 0 | 0 | — |
| Language | Python | Python | Python |
| Last pushed | — | — | 2023-06-08 |
| Maintenance | — | — | Dormant |
| Setup difficulty | moderate | moderate | hard |
| Complexity | 3/5 | 4/5 | 1/5 |
| Audience | general | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires macOS 14 or later and your own AI provider API key.
Newsline is a personal news tool that runs entirely on your Mac, requiring macOS 14 or later. Unlike typical news apps that show a flat daily list, newsline organizes items into storylines, which are ongoing threads that grow as new articles about the same event arrive. So if a story breaks Monday and updates come Wednesday and Friday, all three pieces are grouped together rather than appearing as unrelated rows. The tool pulls from multiple sources: Hacker News, RSS and Atom feeds with eight pre-configured, and Reddit subreddits via public JSON. Each incoming item is rated on a 0 to 10 importance scale by an AI model of your choice. When multiple items are determined to be about the same event, through entity extraction and AI verification, they merge into one story, and the summary is rewritten to fold in the new information. Everything stays on your machine. A Python daemon fetches content, scores it, and stores it in a SQLite database. A native Mac app built with SwiftUI reads that database and displays the storylines in a sidebar, letting you filter by score, search full text, and tag or dismiss items. There is also a chat panel where you can ask follow-up questions about a specific story, with responses streamed back as they are generated. Newsline supports both English and Chinese, and one configuration flag switches all AI-generated output to Simplified Chinese. The AI provider is pluggable, working with Anthropic, OpenAI, MiniMax, or any OpenAI-compatible API endpoint, so you bring your own key and choose your own model. The full README is longer than what was shown.
A private Mac app that groups news from multiple sources into AI-scored, evolving storylines instead of a flat feed.
Mainly Python. The stack also includes Python, SwiftUI, SQLite.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.