Measured, Not Read: The First Numbers

A toy sports feed built over one tournament's results gets benchmarked on real data. Running against a hosted database, the measurement reveals actual performance numbers and an unexpected source of slowness. Walk through the results, the design decisions that shaped them, and why testing failure on purpose beats meeting it in production. 00:02:01 - Scoring mechanics: relevance, importance, freshness 00:03:31 - Measured performance: ranking 66 cards in 0.21 milliseconds median 00:06:46 - Feed architecture and component design 00:08:21 - Infrastructure lessons from Amazon Builders' Library 00:11:39 - Testing failure deliberately vs. in production 00:14:46 - Data contracts and shape validation --- Sources & further reading: • AWS, Schedule your containers on Amazon ECS — — services vs standalone vs scheduled tasks, EventBridge Scheduler, AZ spreading — checked 2026-10-21.: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/scheduling_tasks.html • AWS, Use a target metric to scale Amazon ECS services — — target tracking, fast scale-out and gradual scale-in, ALBRequestCountPerTarget — checked 2026-10-21.: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/service-autoscaling-targettracking.html • Dean and Barroso, The Tail at Scale, CACM 56, 2013 — — "within 100 milliseconds" — checked 2026-10-15 for episode 9.: https://research.google/pubs/the-tail-at-scale/ • Jas Chhabra, Caching challenges and strategies, Amazon Builders' Library, 2019-12-05 — — cache addiction, soft/hard TTL — checked 2026-10-15 for episode 9.: https://builder.aws.com/content/3K9TeXp6strMiJ4vhakpRD7BQWO/caching-challenges-and-strategies • M. Nottingham, RFC 5861, May 2010 — — stale-if-error — checked 2026-10-15 for episode 9.: https://www.rfc-editor.org/rfc/rfc5861 This podcast episode was fully generated by AI — research, script, voices, and production. Built with Claude, Piper TTS, and automated pipeline tooling.