Herman Study Engine

Turn useful documentation into a local study loop.

Upload notes, manuals, markdown, or structured JSON concept packs. Herman extracts study concepts, builds a diagnostic, exposes blind spots, and recycles weak ideas through an adaptive review flow.

Engine snapshot

0 Sources
0 Concepts
0% Readiness
0 Mastered

Static, local-first, and ready for GitHub Pages.

Step 1

Load study material

Markdown and JSON work best today

For this first Herman version, the cleanest inputs are `.md`, `.txt`, `.html`, `.csv`, or `.json`. If you have a PDF, paste the useful text below for now.

Core loop

What Herman is doing

1. Ingest

Turn uploaded documentation into a local source pack with normalized concepts and evidence snippets.

2. Diagnose

Generate multiple-choice checks from those concepts to find weak or unfamiliar ideas quickly.

3. Adapt

Traverse the weakest branch depth-first, keep misses local to that branch, and weight low-confidence answers almost like misses.

4. Cluster

Group source material into topic clusters so the engine can stay inside a weak neighborhood instead of sampling unrelated questions.

The implementation is intentionally static-first right now so this can live comfortably in Herman while the future upload pipeline gets smarter.

Step 2

Run the study engine

Step 3

Gap map

Weakest concepts

Next step

Cluster coverage

Recent sessions

Concept map

Search the current knowledge base

Library

Loaded source packs