Project active

Lumin

A sanctuary for mastery in an age of AI

Knowledge workers consume far more than they synthesize. Lumin inverts this: AI handles first-draft production so you can focus on the craft of understanding.

Approach

Lumin treats learning as curation rather than consumption. The system generates structured knowledge artefacts from source material, then hands you the editorial role: reorganising, challenging, connecting, and refining until the knowledge is genuinely yours. A polymath architecture supports parallel exploration across unrelated domains without forcing premature integration.

What's novel

Paradigm inversion

Instead of you reading and AI summarising, AI produces and you curate. The cognitive work shifts from intake to editorial judgement.

ii.

Polymath architecture

Parallel knowledge tracks that don't force premature connections. Cross-domain insight emerges naturally when you're ready, not when an algorithm decides.

iii.

Craft over gamification

No streaks, no points, no leaderboards. Mastery is its own reward. The interface respects your attention rather than hijacking it.

iv.

LLM as pedagogue

Language models used not for answers but for Socratic challenge. The system asks better questions, not easier ones.

Stack

Next.js 14FastAPIPostgreSQLpgvectorRedisClaude SonnetGPT-4o-mini

Dual-LLM architecture: Claude for deep synthesis, GPT-4o-mini for fast retrieval and classification.

Scale

236 tests

60+ database models

14 versions

Status

Active development. v14 focused on the curation interface and spaced-repetition integration. Core polymath architecture stable.

GitHub