The dataset is a large-scale Bluesky social media research corpus collected for a preregistered field experiment during the 2024 US presidential election. It contains streamed Bluesky firehose records, curated trending and most-liked feed data, preprocessed post records, ML- and API-enriched content labels, multimodal and transformer embeddings, recommendation candidates and ranked feeds, participant-facing feed session logs, and aggregated telemetry/activity tables. Its purpose is to support controlled evaluation of feed-ranking algorithms, measure real user exposure to political content, study social norm misperceptions and algorithmic amplification, and enable analysis of engagement and survey outcomes in a live social media environment.