Think deep. Know more. Heal well.
Analysis of medical artificial intelligence, imaging informatics, and clinical workflow automation.
Read the ArticlesClinical reasoning, research ethics, and decision-making under uncertainty in contemporary practice.
Machine learning, data science, and emerging medical technology, and their measured effect on diagnosis and treatment.
Care delivery in which automation is evaluated against its effect on the clinician-patient relationship.
My two most recent posts addressed separate questions. The first examined whether artificial intelligence (AI), or in the term now more commonly used, superintelligence (SI), poses an existential risk to humanity. That probability is oft…
A dual-process account of AI architecture, and a first evaluation of Jev on note-level PHI detection across 1000 synthetic clinical notes.
On September 9, 2026, Jacob Coxon, a pretraining researcher who had worked at both OpenAI and Anthropic, resigned and posted a thread on X stating that the two labs are "racing straight to self-improving superintelligence and gambling wi…
A small transformer trained from scratch at test time scored 44% on ARC-AGI-1 for 67 cents. What its ablations suggest about sample efficiency, priors, and annotation cost in medical AI.
The word "AI" is now applied to nearly all parts of life and science, including nucleic-acid classifiers, chatbots, robotic surgical arms. The challenge is that ‘AI’ is applied as though a single set of strengths, weaknesses, and risks a…
A review of the published evidence for AI-associated efficiency gains in radiology, covering report drafting, worklist triage, and workforce claims.
LLMs are probabilistic components with a distinct failure signature. We argue that a deterministic automation backbone and an explicit exception-handling loop are prerequisites for their safe use.
Validation performance is weakly informative regarding clinical utility. We review the modes of deployment failure, the human factors that determine adoption, and what characterizes successful implementation.
Deployed models degrade as operational data diverge from their training distribution. We review the mechanisms of drift, its clinical consequences, and the monitoring required to detect it.
Aggregate accuracy conceals disparate subgroup performance. We review how demographic bias propagates through the development pipeline and what should be required before deployment.
Binary alerts convey no information about model certainty. We review why uncertainty quantification is required for actionable clinical AI.
Models described as foundational in medical imaging are generally restricted to one modality and one anatomical region. We review the source of that constraint and its consequences.
Error compounds multiplicatively across sequential agent steps. We propose a hybrid architecture allocating decisions to deterministic rules or agent reasoning by risk and reversibility.
Machine learning is being applied to protein design, gene editing, and molecular data storage. We review the reported evidence and its current limitations.
From autonomous acquisition systems to robotic intervention, we review how embodied AI may alter medical practice and what evidence currently supports it.
Clinical practice is changing, and computational methods are changing with it.
The obligation of the physician is unchanged: to reason carefully, to evaluate evidence critically, and to act in the interest of the patient.
MDSynapse.org publishes analysis of medical artificial intelligence for clinicians, researchers, and others working at the intersection of medicine and computation.
The objective is accurate characterization of what these methods have been shown to do, under what conditions, and with what confidence. Claims are traced to their primary sources, quantitative results are reported as published, and limitations are stated explicitly rather than deferred.
The material is intended for readers who evaluate these systems in practice: clinicians deciding whether to rely on a tool, researchers designing validation studies, and institutions making procurement decisions.