What is deepkin?

DeepKin is a personal content curation and publishing platform. It lets you collect articles, blog posts, videos, and research papers in one place, and generates short AI reading summaries for each item so you can judge how deep to go before committing time to the full piece. The site itself is built with React, TypeScript, Tailwind CSS, and Vite, with content authored as Quarto files, and it deploys to Vercel. The frontend has a video landing page, a browsable grid of content that can be filtered by type or tag, and individual pages for each piece. Content can be added two ways: through command-line tools that fetch a YouTube video or scrape an external article and automatically generate summaries, or by writing a file directly using a template. Beyond the curated CLI flow, DeepKin also supports signed-in user accounts, each with their own public shelf of articles at a URL like /u/username. Authentication and storage run through Supabase, with row-level security so anyone can read a shelf but only its owner can add or edit it. When a signed-in user adds an article by URL, it gets scraped through the Anakin.io scraping API into clean text, then summarized by a large language model, Gemini by default though Claude or Anakin's own AI app can be configured instead, before being saved to the database. The AI reading gists come in three lengths, roughly 150, 600, and 1500 words, generated with a single call to Claude Haiku, and require an Anthropic API key to run the command-line tools locally. This key is not needed to deploy the site itself. The project is licensed under MIT, so it can be freely reused, modified, and redistributed, including commercially, as long as the copyright notice is kept.