Six red flags in standard Hashimoto's care plans, vs what real RCTs and meta-analyses show. Levothyroxine in elderly subclinical patients (TRUST), pre-conception TPO+ women (TABLET), the AIP-diet pilots that don't lower antibodies (and one that raised them), the iodine U-shaped curve, plus two under-discussed adjuncts with real antibody-moving evidence: metformin (Jia 2020 meta) and stress management (Markomanolaki 2019 RCT). Every PMID verified via PubMed eUtils.
Read →
Just as Product-Market Fit measures how well a product matches its market, Evidence-to-Person Fit measures how well retrieved medical evidence matches the patient asking. Why it's hard, and what it takes to fix — in four steps: complete questions, visible slots, open schemas, testable parsing.
Read →
The medical AI landscape, organized by audience: tools for everyday people (ChatGPT Health, Consensus, Examine.com, AskClara) and tools for medical pros (OpenEvidence, AMBOSS, ChatGPT for Clinicians, UpToDate AI, Glass Health, and others).
Read →
How “audit your after-visit summary” means three different things depending on who’s doing it: institutional / compliance audits (EHR vendors, Optum), human-driven patient second opinions (2nd.MD, Cleveland Clinic, Mayo, Solace), and the emerging AI-augmented evidence-grounded patient-side category where No B.S. Med fits.
Read →
For developers building clinical-AI tools, sophisticated users evaluating closed ones, and open-source contributors keeping the space honest. Other players (Consensus MCP, OpenEvidence SDK, ClinicalTrials.gov MCPs) and the medical AI benchmarks worth knowing (NOHARM, HealthBench, MedHELM, MultiMedQA, MedQA, PubMedQA).
Read →