Axiomizer

Axiomizer is a curated interactive learning platform for understanding research from first principles — live at axiomizer.sinabolouri.com.

It is explicitly not a PDF summarizer or an LLM-wrapper chat-with-your-paper tool. Academic papers are written for specialists; Axiomizer converts them into structured learning experiences that explain what was discovered, why the problem existed, why the proposed solution makes sense, where the equations come from, and how the ideas connect to what a reader already knows.

Problem

A traditional summary strips out exactly the material a newcomer needs: prerequisite concepts, the reasoning behind a mathematical form, what a chart is actually demonstrating, and how confident the evidence really is. Reading the original paper doesn't fix this either — papers assume background the reader may not have.

Solution

Axiomizer breaks a paper into chapters made of small, single-idea learning blocks, each one required to pass an "explain it back" test: after reading a block once, someone should be able to explain its idea to another person without having memorized anything. Four block types are built and rendering in production today:

Core insight — one headline, one short paragraph, one supporting visual, with an optional link to a prerequisite concept.
Formula — a setup sentence, the equation itself typeset with real math (KaTeX, not an image), a short variable glossary, and one plain-language takeaway.
Predict-first quiz — the reader answers before anything is revealed; selecting an option locks it and shows a dedicated correct/incorrect color (never the brand accent colors) plus an explanation shown regardless of whether the answer was right.
Takeaway — the single sentence someone would repeat if explaining the chapter back.

Concept pages sit alongside papers: a one-sentence definition, an explicit "Requires" section (with a mandatory empty state for foundational concepts — never a silently blank section), and relationships to neighboring concepts.

Current Status

Axiomizer is a deployed, working MVP of the content platform's foundation — not yet a content library. It currently ships with one complete, real, end-to-end example (Fitts's law, the HCI pointing-time paper) to prove the reader pipeline works, not dozens of published papers.

Built and live:

The full data model: papers, authors, topics, concepts with typed relationships, references, and the semantic learning-block architecture described above
The reader UI for all four block types, plus a concept page, explore/catalog page, and home page
A brand design system (self-hosted Fraunces + Inter variable fonts, light/dark color tokens, an animated brand-mark loader) implemented as CSS custom properties, not hardcoded hex
Django Admin as the interim editorial tool, with an idempotent ensure_superuser bootstrap so an operator can get admin access without needing Render's (paid-tier-gated) Shell feature
Docker + Render deployment with a Neon Postgres database, self-hosted KaTeX, and CI (pytest + ruff)

Deliberately not built yet (see STATUS.md for the full list): user accounts/auth flows, the AI Research Coach, simulations, a production concept-relationship graph visualization (explicitly deferred — see ADR-006), and any real editorial content pipeline beyond one seeded example.

Tech Stack

Django 5.1, PostgreSQL (hosted on Neon), Tailwind CSS (CLI build, no Node framework), self-hosted KaTeX for math typesetting, self-hosted variable fonts, Docker, Gunicorn + WhiteNoise, deployed on Render.

Axiomizer

Axiomizer

Problem

Solution

Current Status

Tech Stack

Links