Selected technical writing
Writing
A small selection of essays relevant to AI systems, optimisation, authority, closure and verification. Each piece is published on the author's archive and linked here; Andaluz does not republish it.
Selected
Relevant to the current programme
- AI Safety & Systems Theory7 min read
The Coherence Wall: Why Optimisation Without Binding Constraints Fails
Powerful optimising systems — from markets to algorithms to AI — can reach a point where soft principles no longer hold. In my formalism, this is the Coherence Wall: a structural failure mode that appears when binding constraints are missing.
Read at sergiumargan.com - AI Safety & Philosophy8 min read
Can AI Have Values? The Moral Kernel Framework
AI alignment is usually treated as an engineering problem. My research suggests it is first a philosophical one — and that getting the philosophy wrong makes the engineering unsolvable.
Read at sergiumargan.com
Broader essays — philosophy, theology, personal and cultural writing — are outside the Andaluz programme and stay on the author's archive.
More writing at sergiumargan.com