The original title is: "AI coding tools remove the review layer that catches billion-dollar errors"

Short summary

The article argues that AI coding tools that write, review, test, and ship code without human oversight remove the review mechanisms that catch critical errors. Using historical examples — JPMorgan's $6.2B London Whale loss from a spreadsheet formula error and Reinhart-Rogoff's debt-growth paper with excluded data — it shows that tools appearing finished is not the same as being correct. The core risk is not bad code but the elimination of the human review layer that catches blind spots.

• •AI coding tools increasingly handle writing, review, testing, and shipping — removing the human checkpoint layer
• •Historical precedents (London Whale, Reinhart-Rogoff) show unchecked tool output causes billion-dollar errors
• •The risk is structural: removing review mechanisms, not just generating bad code

Generated with AI, which can make mistakes.