Back to blog
11 September 2026

Efficient != robust

Yet another entry to the long list called "what software engineering can learn from other fields", this time about why automating everything is solving one set of problems while simultaneously creating whole new ones.

tl;dr: as more part of a system get automated, the competence of the human operators declines. If your role is just monitoring and pushing buttons occasionally, when things go wrong you’re on the back foot - trying to figure out what the system does, while simultaneously trying to figure out why it isn’t doing it.

The solution‘s counter-intuitive: DON‘T automate everything all the time, and instead make sure that the humans do need to interact with the full process often enough to maintain familiarity.

https://spectrum.ieee.org/ai-engineer-skills

Photo — 10 September 2026 British electricity pylons – Telcontar.net