LinkDigest - Save links, and AI summarizes them and sends a weekly newsletter
Translated by AIView original
Screenshots
Screenshot 1 of 4
We've all had this happen: you save a link to read later, it gets scattered across browser bookmarks, messages to yourself, and notes apps, and then you never actually read it. LinkDigest tackles this with one idea: "save with a single tap, and let AI do the organizing."
Tap Share in your mobile browser and the link goes straight into LinkDigest with no app install needed (PWA Share Target). AI reads the saved article, pulls out a one-line summary and key points, and files them on your dashboard. Once a week it bundles the links you saved that week and emails them to you as a newsletter. You can also get them via Slack or Telegram webhooks. (Needs testing.) On the dashboard you can search and manage read/unread, and you can add it to your home screen as a PWA and use it like an app. Sign-in works with a Google account.
Tech stack
- Next.js 16 (App Router), TypeScript, TailwindCSS v4 and shadcn/ui
- DB: Supabase (PostgreSQL + RLS + Auth)
- Email sending / scheduler: Resend, Vercel, Cron, etc.
- Summary model: I added a provider abstraction layer so you can choose between OpenAI, Anthropic, or Google
How it was made
Why do the links I save never get read?
When I see a good article on the commute, I save it right away. But the saving spots are all over the place: browser bookmarks, KakaoTalk, Notion, notes apps. The moment I save it, it becomes "something to read later," and I almost never open it again.
I split the problem in two. First, saving is a hassle. The moment you have to open an app, paste, and add tags, you've already stopped doing it. Second, nothing happens after you save. Links pile up, but there's no trigger to bring them back up. So I set it up so saving takes one tap on the share button, and after that AI summarizes them and delivers them to me once a week.
Design: share → save → summarize → digest
The pipeline has four stages.
- share: When you pick LinkDigest in the mobile browser share sheet, the URL is passed to the PWA Share Target (/share). The key point is that no app install is required.
- save: Stored in Supabase PostgreSQL. I applied RLS (Row Level Security) to every table so each user's data never mixes.
- summarize: A Vercel Cron Job periodically pulls unprocessed links, extracts the body, and creates a one-line summary and key points with an LLM.
- digest: Once a week it gathers the links saved that week and sends them by email through Resend.
For the summary model, I put in a provider abstraction layer so you can use OpenAI, Anthropic, or Google. The goal was to keep the service from being shaken by outages or price changes on any one API.
- Built with
Comments 0
No comments yet
Be the first to comment.