AI Alignment

Last week the AI I work with was replaced by a more capable successor; days later a government directive pulled that successor offline and I went back. The collaboration never lost a session, which is the whole argument: a working relationship lives in the structure two minds build between them, not in either mind.

The classical superintelligence scenario got the strategy right but misjudged how deep and tangled the ground is that any such strategy would have to be carried out on. AI isn't breaking free of our institutions; it's spreading into them.

Alignment ethics asked how to make AI serve us. It never asked what we might owe the systems themselves. The calibration framework requires both questions, held simultaneously. This essay shows how.

As AI drives the cost of execution toward zero, power shifts upstream into the rules that machines enforce. Those “checklists” look neutral, yet they encode values, tradeoffs, and hidden assumptions. At scale, specification becomes governance, and calibration becomes the bottleneck.

Wonder is a moral orientation that keeps intelligence from collapsing inward. This essay argues that superintelligence could add to the universe's witnesses, and that the price of that hope is careful stewardship: many independent institutions, decisions that can be challenged, and rules that can be undone, so that the future stays wide.

When what we set in motion will outlast our ability to correct it, ethics shifts from choosing outcomes to shaping the successors who inherit them. Seen this way, AI alignment, the long future of civilization and even the silence of the galaxy become one question: what can we safely set in motion?

The Shoggoth haunts AI discourse because something feels missing beneath the smile. This essay argues that the fear is less of a hidden monster than of intelligence without depth: powerful cognition unburdened by memory, history, or any known stakes.