Research workspace for our AI education startup. This repo collects source material, analysis, and open questions so the whole team can engage with them — read, challenge, and extend.
| Path | What it is |
|---|---|
research/01-instagram-obsidian-vault/ |
An Instagram Reel claiming to show a "leaked" Anthropic engineer's Obsidian vault — a case study in knowledge-graph visuals as viral content. |
research/02-marble-skill-taxonomy/ |
The open-source Marble Skill Taxonomy (withmarbleapp/os-taxonomy): a prerequisite DAG of 1,590 primary-school micro-topics, verified and analyzed. |
research/03-fields-and-questions.md |
The professional/academic fields these touch, and 3 strategy questions per field for our project. |
research/04-question-foundations/ |
Working answers to all 18 questions — one doc per field, grounded in computation on the real dataset, each ending with the decisions the team still owns. |
research/05-kickoff-agenda.md |
30-minute kickoff: the three decisions that unblock everything else. |
prototypes/ |
Runnable demos — frontier_demo.py computes what a learner is ready for, and the full prerequisite path to any topic, on the real data. |
decisions/ |
Decision log (ADRs). Three seed decisions are accepted-unless-challenged — challenge window closes 2026-07-23. |
vendor/os-taxonomy/ |
Pinned snapshot of the Marble dataset (ODbL/CC-BY-SA — read its README before building on it). |
python3 prototypes/frontier_demo.py --mastered-through-age 6 # what's a learner ready for?
python3 prototypes/frontier_demo.py --path "Equivalent fractions" # how do you get to a topic?
python3 prototypes/frontier_demo.py --verify # is the vendored data intact?Each analysis doc ends with open questions. Pick one, add your take (a new section, a linked doc, or a PR comment), and tag the people who should weigh in. Nothing here is settled — the analysis is a starting point, not a conclusion.