Pocket is shutting down. I built a replacement that doesn't use AI

Pocket is shutting down, and the user lost over 400 bookmarks. Instead of complaining, the user decided to create an alternative that respects user data, stays offline-first, and avoids cloud bloat. Most read-later apps trap data behind closed cloud APIs, require monthly subscriptions, or slow down the browser. The user examined how newer tools handle article text extraction and found them using OpenAI or Gemini API calls, which adds latency, per-token costs, and hallucinations that rewrite or mangle the original text.

The user wanted a fast, private, and deterministic solution, so they built a DOM parsing and heuristic extraction engine using JavaScript. The system evaluates paragraph density, tag trees, and text-to-link ratios to isolate core content in milliseconds. The text saved is 100% accurate, with no AI hallucinations. The app uses a local-first architecture, keeping bookmarks private and under user control.

The tech stack includes React, Tailwind CSS, custom DOM heuristics, a Chrome extension, and Supabase & Edge Functions for optional features. The user invites the community to try the alternative at fetchmarkapp.com and share feedback on parsing favorite tech blogs or heavy documentation sites.