THE NAMED
ONES.
Frameworks referenced across the work, written out properly: what each one is, the actual mechanism, a worked example, and, the part most people leave out, where it stops applying.
A framework with no stated limit is a sales tool. Each of these says where it does not work.
Learning Intelligence Design System
Schools teach what to learn. Almost nobody teaches how to learn. Seven modes, one rule about curiosity, and mastery proved by explanation rather than by grade.
a local-model reference architecture
Everyone is building AI that gets more expensive with every call. This is the opposite. Seven layers where reuse compounds, guarantees live in the components rather than the prompt, and the whole thing runs on a model you host.
a reliable multi-source reporting architecture
Pour five sectors into one model and it drifts, invents numbers, and cannot tell you why it said anything. This is the architecture for a report you can actually stand behind.
a verifiable measurement architecture
Measure hard, surface nothing, and prove that none of it was altered. The architecture behind Project H, generalised so you can build your own.