Building FetchMark: A clutter-Free Bookmark Manager (No AI, Pure DOM parsing)

Hey DEV community! 👋

I recently deployed FetchMark, a clutter-free bookmark manager designed to solve two common web reading problems: dead links and ad-choked pages.

The Problem

Most bookmarking tools simply save a URL. But web pages change, site structures break, and reading an archived article often means dodging auto-play video ads, cookie popups, and tracker clutter.

How FetchMark Works

When you save a link, FetchMark automatically extracts and archives the clean, readable text content while filtering out dead links, homepages, and non-article elements.

Why I Decided Against Using AI/LLMs

The trending approach for text extraction is wrapping an OpenAI or Gemini API call. However, I deliberately built FetchMark using deterministic DOM parsing heuristics instead:

1. Instant Latency: Structural parsing (evaluating paragraph density, <article>/<main> tags, and text-to-link ratios) completes in milliseconds.
2. Zero Hallucinations: Verbatim source content remains 100% accurate—no AI rewriting or summaries.
3. Ultra-Low Cost: Keeps infrastructure lean without paying per-token API costs.

Tech Stack

• Frontend: React + Tailwind CSS
• Backend & DB: Supabase (Auth, Database, Edge Functions)
• Email: Resend

Looking for Feedback

I’d love for the community to give it a test run at fetchmarkapp.com!

Specifically looking for feedback on:

• Extraction accuracy: How does it parse your favorite tech blogs or news sites?
• UI/UX: Speed and readability of saved articles.
• Edge cases: Any URLs that fail to extract properly.

Drop any thoughts, feedback, or questions in the comments below!