The Indie Algorithm

What data is stored in your browser

Your browser stores a small preference profile using localStorage. This data never leaves your device except when used transiently to calculate recommendations.

• Which posts you have read
• Which topics appear frequently in what you read
• Which authors or sites you tend to read

Why this matters: This allows personalization without accounts, logins, or central user databases.

What data is stored on the server

The server stores only anonymous, aggregated statistics that cannot be linked to individuals.

• Total read counts per post
• Co-occurrence counts (posts read together)
• Derived weights used for recommendation quality

No IP addresses, user agents, fingerprints, or personal identifiers are stored.

Why this matters: Global learning improves recommendations for everyone while avoiding personal tracking.

Session identifiers

A short-lived, random session ID is stored in a cookie to connect actions within a single visit.

• Not tied to an account
• Not reused across browsers
• Not used for advertising or tracking

Session data is discarded after learning is merged into anonymous aggregates.

How recommendations are generated

Each recommendation is scored using multiple signals:

• Recency (newer posts are slightly favored)
• Posts related to what you read recently
• Anonymous popularity across readers
• Exploration to avoid filter bubbles

You can always inspect why a specific post is recommended using the “Why am I seeing this?” explanation on each card.

Your control

You remain in control of your data at all times:

• You can clear preferences at any time
• You can export your preference data
• You can disable personalization by clearing storage

Why this matters: Personalization should be transparent, reversible, and optional.

What this project is not

• No ads
• No behavioral profiling
• No cross-site tracking
• No selling of data
• No dark patterns

In short

Personalization happens primarily in your browser.
The server learns only anonymous patterns.
You can see, inspect, export, or delete everything.

Technical Appendix (for the curious)

This section provides a precise description of how data flows through the system. It is optional reading and intended for technically inclined users.

Client-side (your browser)

• Preferences are stored using localStorage
• Stored data includes post IDs, topic keywords, and author counts
• This data is sent to the server only as input for scoring and is not persisted there

Session layer

• A random session ID is stored in a cookie (INDIE_SESSION)
• Session data exists only in server memory
• It is used to connect actions during a single visit

Global learning

• The server stores only aggregate statistics (e.g. read counts, co-occurrences)
• These aggregates contain no user identifiers or timelines
• Data is periodically written to disk to survive restarts

Recommendation scoring

• Each post receives a score composed of multiple weighted signals
• Signals include recency, related reads, and global popularity
• Every contribution is exposed via “Why am I seeing this?”

Design goal: The system favors transparency and resilience over opaque optimization.