The Peltzman Effect - by Jeff - JoT

When teams add safety nets like heavy CI, feature flags, or mandatory code reviews, they often feel protected and shift verification downstream. This feeling of safety reduces individual vigilance, so developers rely on the net to catch mistakes instead of catching them early. The result is risk compensation: faster shipping, more technical debt, and incidents that consume the safety margin the guardrails were meant to provide.

The article shows concrete examples: CI pipelines become a weak signal when they turn red frequently, near-misses are ignored, and rollback incidents are treated as routine. It argues that safety systems must surface real risk - for instance, annotating CI failures with concrete impact, tracking "saved by CI" rates, or making test failures visible in team channels. Practices like error budgets keep risk visible, and progressive trust adjusts constraints based on demonstrated reliability, tightening after incidents.

By turning safety mechanisms into feedback loops that expose danger rather than hide it, teams keep their internal risk model calibrated. This prevents the Peltzman effect from undermining quality, ensuring that safety nets reinforce disciplined behavior instead of encouraging careless shipping.