How It Works

Smart Bookmarks

Browser bookmarks are where good content goes to die. Postiller extracts, indexes, and connects your saved content so you can actually use it.

8 min read

The Bookmark Graveyard

You've done this a thousand times: find an interesting article, bookmark it, never see it again.

Browser bookmarks are flat lists of URLs. No content preview. No search beyond titles. No way to surface relevant bookmarks when you need them. They're digital hoarding with a clean conscience.

The average person has hundreds of bookmarks. Studies suggest fewer than 10% ever get revisited. The rest sit in folders (or worse, unsorted) until a browser migration loses them entirely.

This is a solved problem. We just solved it wrong.


What Makes a Bookmark "Smart"

Postiller bookmarks capture more than URLs. When you save something:

Full content extraction — The article text, not just the link. Headlines, paragraphs, key points—all stored locally on your device.

AI summarization — A concise summary generated on-device. See what a bookmark contains without opening it.

Semantic embedding — Mathematical representation of the content's meaning. Enables search by concept, not just keywords.

Topic tagging — Automatic extraction of 3-5 topic tags per bookmark. Powers discovery and connection.

Your annotations — Notes, learnings, and highlights you add. The context that makes content useful later.

A Postiller bookmark isn't a pointer to content—it's the content itself, indexed and ready to use.


The Safari Share Extension

Saving bookmarks happens through Safari's share sheet:

  1. Find an article you want to save
  2. Tap the share button
  3. Select Postiller
  4. Optionally add notes or tags
  5. Done

The extension captures the URL and passes it to Postiller. Background processing handles the rest—you can close Safari immediately.

What Happens Next

Content extraction — We fetch the URL and extract the article content. Not the full HTML with ads and navigation—just the actual content you wanted to save.

Metadata capture — Title, author, publication date, featured image. The context that helps you remember what you saved.

Processing queue — The bookmark enters a queue for AI processing. This happens on-device when your phone has capacity.

AI enhancement — Summarization, embedding generation, topic extraction. All computed locally, no cloud involved.

Ready to use — The bookmark appears in your library, fully indexed and searchable.

Total time: seconds to save, minutes for full processing (depending on device load).


Content Extraction Deep Dive

Getting clean content from web pages is harder than it sounds. Modern websites are cluttered with:

  • Navigation menus
  • Advertising
  • Related article links
  • Newsletter popups
  • Social sharing widgets
  • Comment sections
  • Footer content

Postiller uses intelligent extraction to identify and capture just the article content:

DOM analysis — We analyze the page structure to find the main content container. Article tags, content divs, and semantic HTML provide hints.

Noise filtering — Known ad patterns, navigation elements, and boilerplate content get stripped.

Text normalization — Clean paragraphs, proper heading hierarchy, normalized whitespace.

Image handling — Featured images and inline article images are captured. Decorative elements are ignored.

The result is clean, readable content that's ready for summarization and indexing.

What We Can't Extract

Some content doesn't extract well:

  • Paywalled articles — We can't bypass paywalls. You'll get whatever content is available to Safari.
  • Heavy JavaScript sites — Content loaded dynamically may not extract completely.
  • PDFs and documents — Currently focused on web content. Document support is planned.
  • Video/audio content — We extract any available transcript or description, not the media itself.

For these cases, you can still save the bookmark and add your own notes. The URL and your annotations remain searchable.


Chunking: The Secret to Good Search

Long articles can be 2,000+ words. Searching for "leadership" in a 2,000-word article about management might return a match—but the relevant section could be anywhere.

Postiller chunks content into smaller segments:

Chunk size — Approximately 400 characters each, breaking at sentence boundaries when possible.

Individual embeddings — Each chunk gets its own semantic embedding. Search matches specific sections, not entire articles.

Topic extraction per chunk — Topics are extracted from chunks, not full articles. A single article about "remote work" might have chunks about "async communication," "timezone management," and "team bonding."

When you search, results point to the specific chunk that matches—not just the article. When AI generates content, it pulls the relevant chunks, not everything.

Chunking is invisible to you. You just see better search results and more focused AI output.


Summarization

Every bookmark gets an AI-generated summary. These are created on-device using Apple's foundation models.

What summaries capture:

  • Main thesis or argument
  • Key supporting points
  • Actionable takeaways (if applicable)
  • Notable data or quotes

What summaries don't capture:

  • Every detail (that's what the full content is for)
  • Author's exact phrasing (these are generated, not excerpted)
  • Context you might find important (that's why you can add notes)

Summaries appear in the bookmark list view, letting you scan your library without opening each item. They're also searchable—sometimes the AI summary captures concepts that the original text expressed differently.


Your Annotations Layer

Extracted content and AI summaries are useful. Your annotations make bookmarks valuable.

Notes — Why did you save this? What resonated? Notes capture context that the content itself doesn't contain.

Learnings — Key takeaways in your own words. When AI generates content later, your learnings are weighted heavily.

Highlights — Mark specific passages (coming soon). Surface the exact parts that matter to you.

Tags — Your organizational system. Hashtags connect bookmarks to your existing topics and ideas.

Annotations are first-class citizens in Postiller. They're searchable, weighted heavily in AI context, and visible throughout the app.

The combination of extracted content + AI summary + your annotations creates a knowledge artifact that's actually useful—not just a URL you'll never click again.


Organization Without Folders

Browser bookmarks use folders. Folders seem logical but fail in practice:

  • Articles about multiple topics get filed in one folder
  • You forget your folder structure
  • Hierarchy creates friction when saving
  • Search doesn't help you browse

Postiller uses tags and semantic connections instead:

Hashtags — Flat, flexible, multi-assignable. A bookmark about "remote team leadership" can have #remote, #leadership, and #management without choosing one folder.

Topic clustering — AI-extracted topics connect related content automatically. You don't have to organize—connections emerge from the content itself.

Semantic search — Find by meaning, not memory. Don't remember if you filed it under "management" or "leadership"? Search for "how to lead remote teams" and find it anyway.

Smart views — Filter by tag, date, read status, or combinations. Create the organization you need in the moment.

The goal is zero-friction capture with intelligent retrieval. Save fast, find later.


Background Processing

Bookmark processing doesn't require the app to be open:

Background app refresh — iOS periodically wakes Postiller to process queued bookmarks. Content extraction, summarization, and embedding happen in the background.

Scheduled processing — Heavy processing (large articles, batch embeddings) is scheduled for when your device is charging and on WiFi.

On-demand processing — If you open a recently saved bookmark before background processing completes, it processes immediately.

You save a bookmark and move on. Hours later—maybe the next morning—everything is indexed and ready. The system works around your usage, not the other way around.


What This Enables

Smart bookmarks aren't an end in themselves. They enable:

Semantic search — Find conceptually related content even when keywords don't match. Search "negotiation tactics" and find that article about "getting to yes in difficult conversations."

AI content generation — Generate posts informed by your saved content. The AI references your bookmarks, not generic training data.

Idea connections — Discover links between bookmarks you didn't know existed. Two articles from different sources about related topics, surfaced automatically.

Knowledge building — Over time, your bookmark library becomes a genuine knowledge base. Searchable, connected, actually useful.

Browser bookmarks are storage. Smart bookmarks are infrastructure for thinking.


Getting Started

The shift from browser bookmarks to smart bookmarks takes a few steps:

  1. Install the Share Extension — Postiller adds itself to Safari's share sheet during setup.

  2. Save something — Next time you find an interesting article, share to Postiller instead of bookmarking in Safari.

  3. Add a note — Even one sentence about why you saved it makes future retrieval better.

  4. Let it process — Give the app time to extract, summarize, and index. Check back in an hour.

  5. Try searching — Search for a concept you saved recently. See how semantic search differs from keyword matching.

The more you save, the more valuable the system becomes. Each bookmark strengthens the semantic index and gives AI more context to work with.

Start with new content. Migrate old bookmarks if you want. Either way, you'll never lose a good article again.

Experience private AI for yourself

Your data stays on your device. Your ideas stay yours. See what privacy-first content creation feels like.