AI Research

Alignment research and experimental tooling.

Exploring resilient, interpretable intelligence that augments human capability. This space gathers experiments, models, and field notes from the Computer Wizard lab.

Loss Benchmarks Research Hub Physics Lab
Now in progress: alignment prototypes, evaluation harnesses, and model diagnostics tied to real-world tasks.

Latest Work

Read the current working paper outlining our loss-informed training regime and evaluation harness.

AI Alignment

Our research formalises alignment at the objective level via the LP-PM-ERT architecture, focusing on cooperative stability and calibrated truth-seeking.

Working Paper

LP-PM-ERT Alignment Architecture and the full PDF are available on this site.

Direct PDF

LP-PM-ERT Alignment Architecture (PDF)