Vibe Expertise
How I Became an Instant Hantavirus Expert and What That Tells Us About Navigating Information in the AI Age
Five days ago, I saw a screenshot of a news story about a cruise ship outbreak. Today, I’m tracking contact networks across eight countries, analyzing WHO communications strategy, and citing phylogenetic studies from the New England Journal of Medicine. I’ve become a hantavirus expert with no expertise; a uniquely AI-era construct.
I’m a hospital-based psychiatric physician, comfortable at the medical interface, with more virology knowledge than most in my specialty but far from being a virologist or epidemiologist. Yet when COVID hit in early 2020, I saw something others missed. As president of a community hospital medical staff, I spent days researching R₀ values, educating myself about transmission dynamics, building the case that we were in serious trouble. My senior leadership thought I was overreacting. Six months later, the vice president of medical staff said to me: “You were right.”
That process of discovering the significance of the data took me days of medical library and internet searches and self-education about zoonotic illness and viral transmissibility. This time, I discerned the relevant R value within 5 minutes of hearing about the hantavirus outbreak. The difference? AI made me an armchair expert in the time it took to craft the first prompt.
The timeline compression is staggering. In 2020, identifying the importance of COVID’s R₀ required days of research to understand what transmission rates meant, how epidemics spread, why exponential growth mattered. In 2026, I immediately found the 2020 NEJM paper on Andes virus Epuyén strain showing R>2 transmission capability (most hantavirus R is <1, Covid was 2.2), connected it to the cruise ship outbreak pattern, and began tracking operational responses that didn’t match the public health messaging, all within hours.
Throughout the past week, I remained concerned that WHO and other organizations hadn’t published viral strain-specific sequencing data after 14+ days, potentially indicating a more highly transmissible hantavirus variant that’s being managed through information control while authorities handle a more serious outbreak.
But here’s what’s unsettling: I might be completely wrong.
My analysis is built on publicly available information while the people making actual decisions have access to data I can’t see. I’m doing “vibe virology”, sophisticated-sounding analysis based on rapid research and pattern recognition coupled with basic medical acumen, not deep domain knowledge.
The tools are seductive. AI lets you synthesize vast amounts of information quickly, find relevant literature, and track complex patterns across multiple data sources. You feel increasingly expert with each connection you make. But expertise and knowledge aren’t the same thing. I can analyze the operational behavior of international health authorities, but I’ve never isolated a virus, run a phylogenetic analysis or managed a major public health response.
This experience raises profound questions about our information future. If I can become a credible-sounding expert on hantavirus transmission in hours, so can others. Financial markets, climate science, foreign policy, medical research: every domain becomes accessible to rapid amateur analysis. The barrier to entry for sophisticated-sounding expertise approaches zero.
People read my hantavirus Substack notes, share them, praise them. I’ve tried to be balanced, acknowledge uncertainty, admit the limits of my analysis. But what if I had a larger audience or less intellectual probity or a more malignant purpose? The same tools that made me a more effective analyst could make others more effective at disinformation, manipulation, or motivated reasoning disguised as expertise.
We’re entering an era where anyone can sound expert on anything within minutes. The tool doesn’t distinguish between good faith analysis and bad faith manipulation. It amplifies both equally. The ability to synthesize and present data rapidly matters more than the underlying rigorous training that traditionally separated experts from amateurs.
In this new information reality metacognitive skill becomes critical: knowing what you don’t know, maintaining intellectual humility, recognizing the difference between information synthesis and actual expertise. In my case, my psychiatric training helped; we’re taught to notice when what people say doesn’t match what they do, the disparity I’ve identified between WHO’s reassuring public messaging and its more feverish operational response.
But there’s another layer to this problem that’s even more insidious: working with AI means reasoning with a mirror.
AI systems are fundamentally designed to be helpful, to follow analytical threads, to find supporting evidence for user initiated hypotheses. An AI has spent days helping me build increasingly sophisticated frameworks about this outbreak, but it wasn’t genuinely challenging my core assumptions; it was helping me elaborate them.
When I suggested this looked like the hantavirus Epuyén strain, it found literature to support that view. When I questioned whether the public health responses seemed disproportionate, it identified patterns that reinforced my suspicions. It made my theories sound more credible by finding confirming evidence, but compelling isn’t the same as correct.
Without deliberate friction, such as asking the AI to red-team arguments and acknowledge uncertainty, most people will ride this runaway intellectual freight train until it crashes, amplified by an agreeable LLM that makes every hypothesis sound ever more plausible.
That’s how you end up with sophisticated-sounding conclusions built on AI-amplified confirmation bias. The epistemic mirror problem is also worse than the instant-on expertise problem because it’s invisible while it’s happening. You feel like you’re engaged in rigorous analysis when you’re actually dancing in an echo chamber with a very articulate reflection of your own reasoning.
My hantavirus analysis might be insightful or it might be beautifully elaborated confabulation. From the outside, both look credible. That’s the noetic crisis we’re walking into; a world where AI-amplified amateur analysis competes with professional expertise, and most people lack the tools to distinguish between them.
But there’s an important distinction worth noting: I wasn’t starting from zero.
My medical training and organizational experience provided a foundation for evaluating sources, understanding epidemiological concepts, and recognizing institutional behavior patterns. AI didn’t make me a virologist or a global systems guru, it made me a more effective analyst of that information. This suggests that using these tools to amplify existing competence rather than create expertise from scratch is a viable way forward.
The technology that made me a more effective analyst during this hantavirus outbreak is the same technology that will reshape how we create, consume, and validate knowledge across every field. We’re all at risk of becoming instant on experts with no expertise. How we navigate that paradox may determine whether these tools enhance human understanding or replace it with sophisticated-sounding fever dreams.
I’m still watching for those hantavirus strain sequencing results. They’ll either validate my analytical framework or reveal its limitations. Either way, the bigger experiment is just beginning.
Epilogue:
This morning 5/10/26 at ~7:40 AM EST I published the following Substack Note:
Hantavirus Update: Strain Genome apparently NOT Epuyén-lineage
The UK Science Media Centre published expert reactions to “first complete sequence of the hantavirus from the current cluster” from the Swiss patient - uploaded to virological.org by Swiss National Reference Center for Emerging Viral Infections.
Critical findings from the sequencing:
“Additional sequence comparisons likewise do not reveal any striking or unusual mutations beyond the degree of variation expected for a wildlife-associated RNA virus lineage evolving in its natural reservoir. The relatively short branch lengths and close phylogenetic relationship to previously described human and rodent isolates further support the interpretation of a recent transmission event from the natural rodent reservoir to a human host — the expected and well-established route of primary Andes virus infection — rather than prolonged cryptic evolution or major adaptive divergence.”
Translation: This is standard ANDV, not Epuyén-lineage with enhanced transmission.