Past AI News

All

1. Abridge launches Assistant Prompt Library across 300+ health systems with real clinician prompts

   Prompt assets and footprint for hospital AI assistants.

   @AbridgeHQ18 likes · 0 reposts · 2 replies · 1,362 viewsView post

   ExpandCollapseAbridge launches Assistant Prompt Library across 300+ health systems with real clinician prompts

   Abridge published an Assistant Prompt Library of real clinician prompts for chart prep, chart query, clinical decision support, note editing and new artifacts. It says Assistant is its fastest-growing product and is available at 300+ health systems. Company and trade coverage note it runs on the same enterprise-approved security and EHR stack. Vendor-reported; sites should verify outcomes and governance locally.

2. Annals of Internal Medicine: Topol et al. question whether clinical AI "human-in-the-loop" oversight is an illusion

   A procurement-side critique of HITL frameworks.

   @EricTopol38 likes · 7 reposts · 8 replies · 9,050 viewsView post

   ExpandCollapseAnnals of Internal Medicine: Topol et al. question whether clinical AI "human-in-the-loop" oversight is an illusion

   Eric Topol and co-authors published in ACP's Annals of Internal Medicine on health systems' reliance on post-deployment human-in-the-loop monitoring of AI for patient safety, with a title framing an illusion of oversight. The tweet has no quantitative results; see the full paper. For IT and compliance, it is a supervision-design question that procurement must answer.

3. Nature Health: five LLMs converge on drug product names and often favor higher-cost alternatives

   Cost and formulary risk for drug assistants.

   @EricTopol75 likes · 30 reposts · 18 replies · 16,486 viewsView post

   ExpandCollapseNature Health: five LLMs converge on drug product names and often favor higher-cost alternatives

   A Nature Health study shared by Eric Topol spans 123 disease–drug pairs and 1,440 sociodemographic profiles; five LLMs often converge on the same product name for a given active ingredient and overall diverge from usual practice toward higher-cost alternatives. For pharmacy and payer teams, drug-assistant rollouts need separate cost and formulary checks.

4. Insilico launches inClinico Agent for clinical-trial intelligence and outcome predictionUnconfirmed

   Agent productization for pharma trial decisions.

   @InSilicoMeds21 likes · 6 reposts · 1 replies · 1,482 viewsView post

   ExpandCollapseInsilico launches inClinico Agent for clinical-trial intelligence and outcome prediction

   Insilico Medicine released inClinico Agent, saying it converts natural-language queries into structured end-to-end analytical workflows that combine LLM reasoning with deterministic tools for clinical-trial intelligence and outcome prediction. The public post gives no accuracy or external-validation numbers; read the release and later papers before treating it as procurement-ready.

5. JAMIA quasi-experiment: 220 primary-care clinicians using ambient AI saw 16% less note time and 19% less pajama time by day 150

   Verifiable longitudinal ROI for ambient scribes.

   @JAMIAno post-level likes, reposts, replies, or viewsView topic

   ExpandCollapseJAMIA quasi-experiment: 220 primary-care clinicians using ambient AI saw 16% less note time and 19% less pajama time by day 150

   A JAMIA quasi-experiment published 18 Oct 2026 followed 220 primary-care clinicians across 36 clinics and 314,845 encounters. Ambient AI scribe use cut note-writing time 7% at day 0 and 16% by day 150; pajama time and time outside scheduled hours were unchanged at day 0 but down 19% and 13% by day 150; wRVU rose 2% by day 150. Not randomized; single-system; generalize carefully. Figures from the paper abstract, not an X original.