(Probably) a retrospective for Secateur
summary
A small Django web app grows to 70,000 users and makes four billion Twitter API calls before free API access dies, costing about $6,000 over its life. Backend work uses Django and Celery with Redis, Postgres tuning, OpenTelemetry/Honeycomb for observability and per‑user rate limits to survive load. The product blocks an account and its followers (temporary or permanent), scales by experiment, and ultimately stops when Twitter revokes developer access.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 0:50Content warning
- 3:35Twitter Protection
- 4:27How it came about
- 7:16Lockdown
- 12:14Cost
- 13:00Blockers
- 15:04Rate Limits
- 16:58Buckets
- 17:53Big Database
- 19:09Open Telemetry
- 20:49Measuring what people want
- 21:34Asking for money
- 22:35Twitter
- 23:35Abuse protection
- 24:52Russian propagandists
- 25:34They started looking into me
- 26:09They messaged Anonymous
- 26:49They messaged Russia
- 27:23How much data is online
- 28:11The end of weirdness
- 28:51Wellington Security
- 29:12Facepalm Emoji
- 29:48Its worth it
- 30:04Emotional and physical limits
related talks
28mOn a shoe-string and a t2.small: scaling on a [zero] budget.
31m"Notes on Over-Engineering: A Project Post-Mortem"
22mThe Open Source Technology Behind a Tweet
26mDjango at
47mHigh Performance Django at Ten: Old Tricks & New Picks
52mChallenges in developing a large Django site
42mA Nice Problem to Have: Django Under Heavy Load
31mDjango for life (sciences)