The Troubles of Automating “all the Things”
summary
Shows concrete failures: an invoice/reporting/payment stack accrues undocumented scripts and monitoring, auto‑scaling hides faults, and an HFT deployment loses $170k a second for 45 minutes totaling $460 million and folds the firm. Compares the 737 MAX long drift into failure. Recommends concrete fixes: treat automation as a product, put scripts and Jenkins pipelines in version control, use simulation and chaos engineering, adopt a checklist with verifier, and pragmatic frameworks like quickrelease/rerun to make automation resumable and auditable.
AI-written summary · about
chapters (11)
- 0:00Start
- 0:44A Quick Survey
- 22:31Other Problematic Automation Aspects
- 24:57Joint Cognitive System
- 29:16The "Animacy" Paradox
- 32:07The Rasmussen Triangle
- 37:04A Manifesto
- 38:26The Checklist
- 40:54The Real Value: A Requirements Spec
- 41:16Version Control: Mattering. Still.
- 49:59Improving Coordination
related talks
41mWhen /bin/sh Attacks: Revisiting "Automate All the Things"
42mWhen /bin/sh Attacks: Revisiting "Automate All the Things"
31mExploring the Unintended Consequences of Automation in Software
38mWhy Automating Everything Adds to Your Toil
38mAutomation Is Hard & We Are Doing It Wrong
32mManual work is a Bug!
41mLessons from Mission-Critical Systems in Fast-Moving Industries
32mWhat Not to Automate