# Design a Scalable Post, Feed, and Search Service Design a social-post service with three core capabilities: a user can publish a text post, retrieve a personalized feed of recent posts, and search posts by text. Cover the high-level architecture, APIs, data flow, storage choices, event streaming, scaling strategy, and an end-to-end walkthrough. Do not assume one universal consistency model. State what users should observe after publishing or deleting a post and how those choices affect feed and search paths. ### Constraints & Assumptions - The workload is read-heavy, but publish bursts and high-fanout authors exist. - Feed reads need pagination with stable cursors. - Search indexing may lag the source of truth if that behavior is explicit and observable. - Media, recommendations, and advertising are outside the initial scope. ### Clarifying Questions to Ask - Is the feed chronological, ranked, or a hybrid? - What latency, availability, freshness, and retention targets matter? - Which privacy and deletion guarantees are required? - How large can a follower graph and one author's fanout become? ### What a Strong Answer Covers - Clear APIs, identifiers, cursor semantics, and source-of-truth ownership - A justified fanout-on-write, fanout-on-read, or hybrid feed strategy - Durable event propagation to feed materialization and search indexing - Partitioning, caching, backpressure, idempotency, and hot-key handling - Deletion, privacy, observability, and reconciliation of derived views ### Follow-up Questions - How would you prevent a celebrity publish from overwhelming the system? - How does a privacy change propagate to cached feeds and search results? - How would you rebuild the search index without missing concurrent writes? - What does a stable feed cursor contain when new posts arrive between pages?