Ground Truth
The ground beneath the model.
First-principles explainers on the data layer beneath modern medical AI and drug development. What high-fidelity clinical data is, how it is coded and governed, and why it sets the ceiling for the models and therapies built on top of it.
Model Development & AI
EHR Data for LLM Pre-Training: What It Is and Why Frontier Labs Want It
A first-principles guide to EHR data for LLM pre-training: how it differs from fine-tuning and RAG, why structured clinical data beats scraped text, and what labs want.
Synthetic vs. Real-World Clinical Data: What Actually Trains Better Medical AI
Synthetic vs real-world clinical data for training medical AI: where synthetic data helps, where it breaks on comorbidity and temporal coherence, and what wins.
What Is "Ground Truth" in AI — and Why Data, Not Compute, Sets the Ceiling
Ground truth is the verified reference reality AI models are measured against. Here's why model capability is now gated by data fidelity, not compute.
Why Longitudinal Patient Data Is the Key to Clinical AI World-Models
Longitudinal patient data captures how health changes over time, giving clinical AI the trajectory signal that snapshots can't. Here's why it matters.
Drug Development & Pharma
External Control Arms: How Real-World Data Is Replacing the Placebo Group
External control arms use real-world data instead of a placebo group. Learn how they work, where they fit, regulatory considerations, and the data they require.
Real-World Evidence (RWE), Explained: How Pharma Turns Data Into Decisions
A clear, first-principles guide to real-world evidence (RWE): RWD vs RWE, the FDA regulatory shift, core use cases, and the trade-offs of each data source.
Foundations & Explainers
Drug Identity in Medical AI: How RxNorm and NDC Stop Hallucinations
RxNorm and NDC give drugs a stable identity. Learn how coded drug data grounds medical AI against drug-class confusion and pharma hallucination.
HIPAA De-Identification, Explained: Safe Harbor vs. Expert Determination
A first-principles guide to HIPAA de-identification: PHI, the 18 Safe Harbor identifiers, Expert Determination under 164.514, BAAs, and why de-identified data unlocks broader use.
Medical Ontologies Explained: SNOMED CT, ICD-10, RxNorm & LOINC
A clear, first-principles guide to SNOMED CT, ICD-10, RxNorm, and LOINC — what each codes, who maintains them, and how they cross-walk to one clinical reality.
Multimodal Clinical Data: The Five "Modes" That Make Health Data Hard
Multimodal clinical data is more than images. Learn the five modes—codes, narrative, numeric, temporal, and financial—and why fusing them is the hardest, highest-value signal.
What Is HL7 C-CDA? The Format Behind Every U.S. Health Record
HL7 C-CDA is the XML standard behind U.S. clinical documents. Learn what C-CDA is, why it exists, how its header, sections, and codes work, and why it matters for AI.