How EdTech Founder Stack is put together.
A knowledge base for edtech founders, stored as plain markdown. No app, no API, no backend, no dependencies — just files that an AI tool (Claude Code, Cursor, ChatGPT) or a person reads. Edit a file and the next read is current. Every claim traces to a source you can check.
All of it lives in data/, in three kinds of files.
Structured domain knowledge, grounded in real sources rather than model training data:
- Regulatory — FERPA, COPPA, and state privacy law (K-12); accreditation and accessibility (higher ed)
- Market — competitive landscape by segment, buyer personas, buyer demand signals, funding landscape by stage, procurement, pilot benchmarks
- Frameworks — ESSA evidence tiers, AI-native vs. bolted-on, defensibility moats, the higher-ed jobs atlas, founder traps, and the demand-validation toolkit (the 5-question diagnostic plus the JTBD Switch interview method)
Each regulatory and market file carries a "last updated" date. Update cadence is roughly quarterly; regulatory data when laws change; the competitive landscape goes stale fastest.
Hundreds of peer-reviewed papers across the major learning-science topics, each stored in a numbered markdown table: #, Title, Takeaway, Type, Year, Citations, DOI. The index lives in data/research/README.md. This is the evidence base — claims about what works in learning cite specific papers with author, year, finding, and DOI.
Dozens of field lessons from operators and investors, distilled and attributed from the public archive of Lenny's Podcast and Lenny's Newsletter, then mapped to selling into schools, universities, and L&D. These are practitioner experience, not peer-reviewed evidence — the research corpus is the evidence layer, and the file says so. The same practitioner-not-peer-reviewed labeling applies to the summit-sourced files (data/buyer-demand-signals.md, data/ai-risk-and-trust.md), which each carry their source and an evidence-tier note.
Point an AI tool at the repo or a single file and ask; it reads the relevant knowledge and answers in context. Or read the markdown directly. Because the knowledge is files you can diff and audit, it stays current and checkable in a way baked-in model knowledge isn't.
The repo ships a small instructions file for each major agent. AGENTS.md is the canonical one, read at the repo root by Codex and Cursor; CLAUDE.md and GEMINI.md import it, and .github/copilot-instructions.md points GitHub Copilot to it. They all tell the agent the same thing: read the relevant data/ file and cite the source rather than lean on training data. AGENTS.md is the single source; edit it, not the pointers.
Anything that can read markdown works: Claude Code, Cursor, ChatGPT, Claude, or a plain text editor. There's nothing to install beyond cloning the repo.