MedLit.
About MedLit

Patients can read their own records. Most cannot understand them.

Patients have had the right to read their own records since the 21st Century Cures Act rules took effect in 2021, and portals now hand over most of the chart. What they hand over is the hospital's own format: SNOMED codes for conditions, RxNorm codes for drugs, LOINC codes for labs, a value with an H beside it. I wanted the version a nurse would say across a kitchen table, with the page it came from.

So MedLit reads the record as the hospital stores it, looks each code up in the registries the National Library of Medicine and the FDA publish, and asks a language model to explain what it found in plain words at a reading level you choose. The model never decides whether a number is normal; the reference range does that. Every paragraph carries the MedlinePlus or openFDA page it came from, and I score the text itself for reading level before you see it.

What happened next

Epic, OpenAI, Anthropic, Google, Amazon, and Microsoft shipped patient-facing record explanation within months of this build.

I brought full new guy energy to the spring. I legitimately thought I had cracked something nobody else saw, and six companies were landing on the same shape the whole time. Their launches erased the market for a standalone product and validated the architecture in the same quarter. It was graduate research; I'll take the validation.

EpicOpenAIAnthropicGoogleAmazonMicrosoft

Three things you can check.

Sources you can check

Every explanation cites MedlinePlus, openFDA drug labels, and RxNav. The citations are links, and they leave this site.

A reading level you choose, then verified

Simple, Standard, or Detailed. Flesch-Kincaid scores every answer on the way out, so the level you picked is the level you got.

One record, three summariesUS school grade
Simple
5th grade · 5.1
Standard
7th grade · 6.5
Detailed
10th grade · 9.8
Measured on one sample record's summary, on the text itself.
The receipt

The data was lying. The validator caught it.

A physician reviewer noticed a medication card citing education material for ticagrelor, a blood thinner, under a tiotropium label, a COPD inhaler. The record carried RxCUI 1116634; ask RxNav what 1116634 actually names and the registry answers ticagrelor. The fix ships as fhir-code-lint on PyPI, and the first bug is checkable in one click.

RECEIPT №1116634RXNAV · NLM
the record says:Tiotropium 18 MCG Inhalation Capsulethe registry answers:ticagrelorA COPD inhaler label on a blood thinner. Found by a physician reviewer; the root cause was the data itself.
The fix ships as fhir-code-lint on PyPI.pip install fhir-code-lint
Built bywilliam garrowWilliam Garrow

Research prototype. Synthetic patients only. The pipeline, the grounding, and the architecture are written up in full on the technical page.

Read the full technical notes
See it in actionSign in, pick a sample record, and read it at three levels.Sign in to MedLit