πŸ“• subnode [[@jakeisnt/sieve]] in πŸ“š node [[sieve]]

Sieve was an idea for a multimedia inspiration engine, largely inspired by AI-driven solutions and some need to decrease the insane amount of information that I was consuming at this point in time of my life. It's definitely still viable, but to get off the ground is going to require a huge amount of data plumbing, and it might get you banned from Instagram/Twitter/Social terms of service… APIs are being deprecated as companies are less and less willing to allow you to access their stuff.

Inspiration

To give others a brief summary of the idea and its goals, Sieve is similar to existing social media and content aggregation services (I see it as being most similar to Tumblr) but it:

  • Supports only open standards (RSS!)
  • Prioritizes content, has few or no social features whatsoever
  • Views are content-specific, because the best view for images is not the best for video, audio, or articles
  • The recommendation engine doesn't track users or collect any user data; it functions only by examining content you've saved.

Prior Work

Tumblr Isolate image browser

Value Proposition

  • Aggregates content from the user's other social media and content feeds with little or no effort.
  • Filters incoming content and sorts it according to media type.
  • Views are media-first and are designed to prioritize focusing on content rather than social interaction.

Framework

Sieve aspires to serve as a feed for content aggregation and exploration, removing unnecessary distractions from other platforms while collecting and displaying their content.

It allows for the viewing of other social media and content platforms in one place from a variety of mediums – video, audio, pictoral, and written.

It provides multiple views conducive to each type of content to enhance the exploratory experience, and allows for saving types of content for later as well as publishing such content to your own.

At first, this engine that will be self-hosted, and the self-hosted instance will independently generate content best for you provided links to follow.

To improve usability, this may extend to a service in which users can create their own workspaces and subscribe to the feeds of others ('meta-feeds'), though the focus of this project is to provide a variety of different ways to customize and display content.

UX

Operationally, Sieve is an RSS feed reader – but one that puts multimedia content first. It can be provided feeds and can crawl the internet for topics similar to those already in feeds via a web crawler and content aggregation system.

The innovation here is the multiple views of content it provides – while attribution is important, other aspects of social content publication can be distracting or addicting. Sieve doesn't have 'likes', 'follows', or 'comments' – it seeks only to provide different ways of viewing the content we already interact with on other platforms.

On the back end, Sieve scans each of the feeds a user follows to find a cohesive 'theme' or 'style' from the content it displays, and prioritizes displaying more of the content a user saves. It does not track you or learn from you in any other way.

Design

Sieve is minimal first. No clutter.

It puts the content first and hides all excess information, allowing for 'focused' inspiration without potential distractions.

Previous efforts with seamless, content-first design include Archillect. This project seeks to avoid creating a new Tumblr, Instagram or Facebook – those platforms have become too bloated and interaction-driven.

This platform seeks to minimize interaction with other users and instead prioritize interaction with their content.

These distractions are deliberately obfuscated, hidden behind menus and obscured by a minimal interface – the content itself.

There are no notifications, no numbers, no flashing visual cues to indicate additional information, as these are flaws that all make social media content more addictive. It's as simple as visiting a page and viewing content catered to the user.

Image View

The image view presents itself as a multipanel scrolling feed of image-based content. This scrolling can be infinite or paginated. I'll likely implement the latter first. Images can be clicked to open a larger view of the image and to view attribution if available.

Video View

This is probably the trickiest to get right; it may end up being an image view with GIFs instead of just images, but that would disparage any audio content that the videos may contain.

The information density of audio in videos is much worse than audio in isolation, so playing audio may not be relevant to the content, but providing a distraction free way to consume this content feels necessary.

Sound View

This view will be similar to the 'theatre mode' that many services already provide, showing an oversized album cover while playing a specific song.

It'll focus on playing previews of songs rather than their entire contents so that the media can be set aside and continue to be consumed on another platform.

Article View

RSS is overwhelming and often doesn't present the information you're interested in – even good writers are bad at summarizing their own content and prioritizing the information that readers want, and often blogs address a variety of topics rather than focusing on a subset of articles that a specific reader may be interested in reading.

As such, the article view aims to do a few things that other services currently do not accomplish:

  • Generate a summary of the content provided on the page
  • Prioritize articles based on content of previously saved content
  • Always use user-customizable feeds for such content

These articles will be viewable through a 'reader view' that the article view provides.

Advanced features for the article view include saving individual passages as opposed to entire articles.

Categories

In addition to sorting based on type of media, users should be able to identify categories that they're interested in. These categories are used by the recommendation algorithm under the hood and content is tagged with one or more of these categories when it enters Sieve.

Viewing Saved Content

This operates identically to the inspiration views, but it only shows content you've already saved.

Future Considerations

I have several adjacent ideas that, while outside the scope of this project, could be added in the future.

  • A service for viewing your own, internal content in the same way that users of Sieve consume external content.
  • A way to 'sieve' through integration with content on other platforms, including a bookmarking system that automatically determines the type of content that was bookmarks and places it in the appropriate place on the user's feed.

Productization

I believe this project to be a viable product.

The open-source version will be provided entirely for free, and will provide the Sieve engine. No user accounts, following, liking, etc. will be involved with this draft (federation might be cool in the far future but it's by no means necessary). However, people will be able to try out and use the service on an individual basis by hosting it themselves.

The commercial features:

  • Views should be plugins. They can be associated with a specific type of content and will display that content in a manner best fit for it.
  • Users should be able to make their own accounts and start their own inspiration engines for a fee or with advertisements interspersing their content.
  • Users should be able to follow other users. Following other users adds posts those users have saved to their own feeds; these posts then show that they came through those users. This is as far as the social interaction on this platform will go.
  • External platforms should be added and supported; i.e. Instagram RSS feeds through Bibliogram. The aggregation of external platforms is huge for external users to hop on and start focusing on this platform.

Tools

https://github.com/RSS-Bridge/rss-bridge bibliogram for subscribing to instagram invidio for youtube https://github.com/avencera/fast_rss for parsing rss from elixir backend https://github.com/miniflux/miniflux feed reader with Go https://github.com/zserge/headline ascetic RSS reader without server. 4kb and beautiful, works offline

Search

Sieve's search shouldn't help users find whatever they'd like; rather, it plays a curatorial role in the process.

someone tweeted about rss rewind, where you're able to replay feeds day by day to trace the news, articles, etc chronologically what's in your rss feed interviews!

seenaburns/isolate is a lightweight tool for viewing art inspiration.

memory metadata – When I take a photograph of my sister and niece on my iPhon… is an exploration of the metadata of a photo, reflecting on what it means to capture a moment in time.

Scattered recommendation engine ideas

  • Add inputs you like to different 'channels'
  • Submit collections of data and get to see the generated channels from that data
  • good design aesthetic
πŸ“– stoas
β₯± context