Every week, I have a new patient tell me I should already know their history because their PCP sent over their medical records. Usually we don’t receive anything. If we receive anything, it’s about 100 pages of the same information repeated over and over. I explain to patients that the EMR gathers a lot of data, but that data isn’t always useful. Even if I comb through the data and determine what is useful, the most useful information always comes from the patients themselves.

AI is no different. With nearly everyone using LLMs these days, its easy to get lost in the details. The core information is still useful, but it’s so watered down the message gets lost. LinkedIn, Facebook, Twitter, and even Instagram is completely overrun by AI slop. AI slop includes low-quality content created by generative AI. It is being repeated in high volumes, typically for clickbait, marketing, or to flood social media platforms. Just like the EMR, the real message gets lost.

What’s happening this week in AI:

Plus: Tool of the week

Let’s dive in.

💊 Novo Nordisk Uses AI to Accelerate Drug Launches: Novo Nordisk shared that it is now using AI across regulatory drafting, safety analysis, and commercialization workflows to compress the timeline between late stage trials and regulatory filing. What caught my attention was not just the automation itself, but how deeply the company has integrated AI into operational infrastructure across global teams, particularly in Bengaluru.
Why this Matters: We are starting to see healthcare AI move beyond documentation support into enterprise level clinical and operational acceleration.

🧬 Zuckerberg Biohub Releases Protein “World Model”: The Chan Zuckerberg Biohub unveiled a massive AI system trained across billions of proteins, with early evidence that it can model biologic interactions and design functional protein interfaces in silico. It feels increasingly likely that the next major leap in therapeutics will come from computational biology platforms rather than traditional discovery pipelines alone.
Why this Matters: Clinicians may soon see dramatically shorter timelines between molecular discovery and therapeutic development.

⌚ Oura Expands Into AI Guided Care Pathways: Oura’s upcoming AI enabled ring goes beyond passive tracking and now connects physiologic signals directly into AI triage workflows and clinician escalation pathways. The interesting part is not the wearable itself, it is the emerging closed loop between consumer health data, AI interpretation, and real clinical follow up.
Why this Matters: Remote monitoring is slowly evolving from wellness tech into longitudinal clinical infrastructure.

📚 What the AI Era Doctor Should Know: A new scoping review in npj Digital Medicine synthesized proposed AI competencies for medical education and argued that medical schools still lack coherent standards for teaching AI literacy. The paper emphasized practical competencies around bias, interpretation, workflow integration, and clinical oversight.
Key Takeaway: AI literacy is rapidly becoming a core clinical competency rather than an optional technical skill.

⚙️ Framework for AI Implementation Research in Healthcare: Researchers proposed a consolidated implementation framework designed to help health systems understand why so many promising AI tools fail during real world deployment. The framework focuses heavily on workflow integration, governance, trust, and sociotechnical barriers.
Key Takeaway: AI has the potential to rapidly improve our research. however, it has a long way to go before it can take over everything. For now, we can use it to improve our data collection and efficiency.

⛪ Even the Vatican Is Warning About AI Noise: Pope Leo XIV released his first major AI manifesto this week, calling for stronger oversight of artificial intelligence and warning that unchecked systems could amplify misinformation, distort truth, and weaken human decision making. What stood out was the focus on information integrity rather than just technical safety.
Why this Matters: Concerns about AI generated noise, manipulation, and erosion of trust are moving far beyond tech circles and becoming mainstream societal issues.

⚖️ Healthcare AI Governance Moves Center Stage: A new Nature Digital Medicine paper outlined tiered governance frameworks for healthcare AI deployment, with growing emphasis on auditability, monitoring, and institutional accountability. The tone feels very different from even a year ago, much more operational and less theoretical.
Why this Matters: Health systems deploying AI will increasingly need governance structures that look more like quality and safety programs than innovation labs.

What it does: Searches and summarizes findings directly from peer reviewed research, while showing the underlying papers and level of evidence.

I have started using Consensus when I want a quick reality check on a claim before diving into a literature review. What I like is that it points me back to actual studies instead of giving me a polished answer without context. In a world where AI can generate convincing sounding medical content in seconds, having a tool that surfaces the underlying evidence feels increasingly valuable.

Prompt to try:

Healthcare AI is becoming bogged down by AI slop, and its up to us to filter it out. Patient’s won’t be able to understand what is true and what is misinformation. The more AI content we generate, the more valuable human judgment becomes.

The bottleneck in healthcare is no longer information. We are entering an era where AI can generate summaries, studies, notes, recommendations, and educational materials faster than any clinician could ever read them. The challenge is deciding what deserves our attention.

One practical takeaway: start treating information quality as a clinical skill. The clinicians who thrive in the next few years may not be the ones using the most AI, but the ones who get best at separating signal from slop.

If this sparked a reaction, send it to a colleague or reply with the most useful, or most frustrating, AI generated thing you've seen this month.