Each step pairs a heading and description with a decorative 3D preview of the Equivox UI. Those previews are skipped in screen reader navigation and are not interactive for assistive technology—the text beside each preview is authoritative.
Posts follow common social media patterns:
Equivox threads are inspired by Reddit, replies nest recursively and can grow arbitrarily deep. This seemed more fitting than flat comments.
Each thread supports likes, and authors can edit or soft delete their own comments. Threads can be collapsed and expanded for readability. The page first renders skeleton loading states while post and thread data loads, then hydrates into the full discussion tree.
A top-level composer starts new threads, and inline reply composers appear throughout the tree. Discussions and replies both support rich text. Engagement actions are auth-gated, so logged-out users are prompted to log in before participating.
Discussions can include optional image and video attachments, processed asynchronously through a moderation pipeline. A worker hosted on Railway integrates AWS Rekognition to classify and gate media before it is broadly visible.
Each media item is tracked with a status: pending, processing, ready, failed, blocked, or review. Public viewers only see media after it reaches ready. Post owners can still view their own media regardless of status.
Multi-attachment posts use an Instagram-style grid, and selecting media opens a custom fullscreen modal. Delivery uses signed URLs for access control.
Discussion posts can optionally include polls. On create, options are normalized and validated (trimmed, deduplicated, minimum two options), and votes are switchable.
After voting, the UI enters results mode with percentages and total vote count, while preserving the user's selected option. Voting is optimistic, so the interface updates immediately and server state reconciles in the background.