Your AI Cannot Use the Posts You Saved, So I Built Social Memory

Likes and bookmarks on social media platforms like X and Threads are simple to accumulate, but difficult to utilize effectively. Users often save interesting details, thoughtful comparisons, or ideas for future projects, only to later struggle finding these saved posts. Meanwhile, AI assistants like Codex or Claude Code cannot readily use these saved posts as context unless the user manually locates and pastes them into the AI system.

To bridge this gap between collecting useful posts and actually working with an AI assistant, a user created Social Memory. This is a local-first evidence library for X/Twitter and Threads, focusing on providing context for AI interactions rather than acting as just another feed to keep up with. When saving posts, it's important to remember that a "like" doesn't mean the same thing to everyone.

It could signify a "useful reference" or simply "thanks." Social Memory allows users to independently decide which likes, saves/bookmarks, and reposts should be included in their collection signals. The storage model keeps posts separate from the reasons they were collected, storing each post once by platform and external post ID.

This way, a post can be captured multiple times based on different signals, without creating duplicate copies. The workflow begins in the user's existing Chrome profile. An unpacked Manifest V3 extension connects this profile to a local Native Messaging host, allowing for a local library to be created without exporting passwords or manually entering profile IDs.

Multiple Chrome profiles can use the same extension folder and connect with a single click. Each profile gets its own installation identity, while the local library deduplicates posts across profiles to avoid duplicate search results. The process then involves choosing collection signals, collecting posts into the local library, searching for relevant evidence, and letting a connected assistant work with the results.

The architecture is deliberately described, not claiming the extension is fully tested. The current limitations include limited platform support (macOS/Linux), an unpacked Chrome extension, and unverified daily collection on Chrome. The source is under PolyForm Perimeter 1.0.1, and developers are encouraged to review the license before building upon or distributing any derivatives.