/eye-dem-POH-tent/
Producing the same result no matter how many times the operation is applied. In computing, an idempotent request can be safely retried without side effects.
Idempotent means an operation produces the same outcome whether you run it once or a hundred times. This concept is foundational in API design, distributed systems, and database operations. When you tell an AI agent to 'make this idempotent,' you're asking it to build something that won't break if a network hiccup causes a retry. HTTP PUT and DELETE are idempotent by design — calling them twice doesn't create duplicates or double-delete. POST is not idempotent, which is why payment systems need careful deduplication.
For AI operators, idempotent thinking prevents the most expensive class of bugs: the ones that silently corrupt data through repetition. When you're building automations that trigger on webhooks or cron jobs, idempotency isn't optional — it's the difference between a system that self-heals and one that multiplies errors.
When designing any API endpoint, webhook handler, or automated workflow that might be retried — payments, form submissions, data pipelines.
Idempotent systems are self-healing. Non-idempotent systems multiply errors. Every production-grade API you admire is idempotent by design.
Think of an elevator button — pressing it 10 times doesn't call 10 elevators. That's idempotent.
A Mac app that coaches your AI vocabulary daily