Show HN: MEDF – Mutable Expression Description Format (github.com)
Hi HN,
I’m sharing MEDF (Mutable Expression Description Format), a document format and CLI focused on one narrow problem: making text content verifiable without freezing how it is presented or reused.
MEDF treats text blocks as immutable and verifiable (using RFC 8785 canonicalization and hashes), while explicitly allowing rendering, indexing, and layout to change over time.
It does not attempt to solve trust, identity, correctness, or authority. It only answers: “Is this exactly the same text that was published?”
The reference CLI supports block-level hashing, optional signatures, and offline verification.
Repo: https://github.com/maskin/medf
Feedback and criticism welcome — especially on real-world use cases
1 comment
[ 2.8 ms ] story [ 14.7 ms ] threadWe’re releasing MEDF (Mutable Expression Description Format) — an open document format designed for the AI era.
The core idea is simple:
AI-generated documents are not the problem. Unverifiable documents are.
Instead of trying to detect or ban AI-generated text, MEDF focuses on fixing the canonical meaning of a document, while allowing its expressions (PDF, Markdown, indexes, summaries) to remain mutable.
What MEDF does
Documents are split into semantic blocks
Each block and the document as a whole are hashed and signed
Offline verification is possible (no external services required)
Indexes and views are explicitly non-canonical and regenerable
AI usage is optional metadata — not a requirement for validity
What MEDF does NOT do
No AI detection
No watermarking
No blockchain dependency
No centralized trust service
We believe this approach works better for:
government and administrative documents
academic papers and citations
long-term archives in the AI era
Repository (spec + early foundation): https://github.com/maskin/medf
This is not a final standard — it’s a starting point. Critical feedback is very welcome.