How Megadose handles server requests, local preferences, analytics, and the public ranked feed it builds from other publishers.
Megadose is a reader of public AI news and benchmark snapshots, not a primary source. The ranked home feed is assembled from about twenty public sources, classified, deduped, and time-decayed with a two-day half-life so recent field-shifting items surface first. Megadose does not claim those articles, papers, or leaderboards as its own reporting. Opening an item sends you to the publisher; their privacy practices apply after you leave.
The ranking math uses an importance score, a per-source trust weight, and exponential decay. Dedupe uses canonical URL hashes and title signatures. Audience filtering drops policy, politics, surveys, and trivia rather than profiling individual readers. Live SWE-bench Verified, LMArena, Terminal-Bench, and METR Horizon tables are mirrors of upstream boards, not original evaluations.
The public feed does not require an account. Megadose servers receive normal request information needed to deliver the service, such as IP address, request time, route, and user agent. Preferences such as theme and dismissed stories are stored on your device.
The current Megadose iOS App Store release does not offer account creation or sign-in. It does not collect an email address or account profile through the app.
The current Megadose iOS App Store release does not initialize PostHog or another third-party analytics service. The Megadose website may use PostHog to understand page views and interactions when web analytics is configured. Megadose does not sell personal information or use analytics data for cross-app advertising.
You can clear local preferences by deleting the app. Megadose may retain operational server records where reasonably necessary to deliver and secure the service, prevent abuse, and satisfy legal obligations. For a privacy or deletion request, use the support page on this site. Do not include verification codes, access tokens, or other secrets.