Search for a command to run...
I spend a lot of time arguing with myself about what Panacea actually is. Is it a supplement encyclopedia. Is it a research interface. Is it an AI agent over a database. Is it an influencer fact-checker. Each of those is true and none of them is the whole truth, and the answer matters because what you think the product is determines what you build next.
A while ago I did something that helped. I wrote down the argument as if it were an email thread between two investors I pay attention to, having them make the cases I could not resolve in my own head. It is a slightly odd exercise, putting words in the mouths of people you have never met, but it forced me to make each position as strong as it could be instead of strawmanning the ones I disliked. One line came out of that exercise that I have not been able to stop thinking about, and I want to explain why.
The line was this: supplements are the topic, but the product is trust infrastructure for health information.
It is easy to look at Panacea and see a supplement site. That is the topic. Supplements are what the data is about, what the search returns, what the influencer claims concern. If you stopped at the surface, you would build a better supplement encyclopedia and call it done.
But the surface is not the product, and the reason is that the surface is replaceable. An encyclopedia, however well designed, competes with other encyclopedias on completeness and clarity, and there is always someone with more time to write more entries. If that were the whole game, it would be a content game, and content games are brutal and commodity.
The thing that is not replaceable, and that I keep coming back to, is underneath. It is the structured representation of a whole field of research — every claim normalized, every outcome linked, every direction counted, every influencer statement matched to the evidence it claims to rest on. It is, in the language of that imaginary email, a semantic map of the supplement conversation as it exists across both the research and the culture. The research lives in papers. The culture lives in videos and posts. Almost nobody has connected the two in a way you can query. That connection is the product, and supplements are just the first domain where we built it.
Think about what actually breaks when someone tries to find out whether a supplement works.
The research exists, but it is scattered across hundreds of thousands of papers, described in inconsistent language, and written for specialists. The conversation exists, but it is driven by people whose incentives are reach rather than accuracy, and it moves faster than any reader can verify. What is missing is the layer that connects a specific claim — a specific influencer saying a specific thing about a specific supplement and outcome — back to the specific body of research that would confirm or refute it.
Building that layer is a data problem more than a software problem, and it is a data problem that compounds. Every claim normalized makes the next claim easier to place. Every influencer statement matched to evidence makes the next match more valuable, because it contributes to a credibility signal. Every outcome linked to its variants makes the consensus over that outcome more accurate. The map gets better the more of it exists, in a way that an encyclopedia entry does not. That is what a moat actually looks like — not a feature, but a representation that is expensive to build once and expensive to rebuild, and that becomes more useful the larger it grows.
If the product is the trust infrastructure and not the supplement site, then the priorities reorder.
It means the data work — the normalization, the linking, the consensus scoring, the boring ninety percent nobody sees — is not overhead. It is the product. The interface is the thin, polished layer that lets people reach the map. We invest in the map first, because the map is what nobody else can copy by building a nicer frontend.
It means we should think about the map as infrastructure that other things can be built on, not just as a website a person visits. An agent that answers health questions needs exactly this kind of structured, queryable evidence layer to be reliable. A clinician considering a patient's supplement stack needs it. A future version of search that can reason about evidence needs it. The website is one consumer of the map. There will be others.
And it means the influencer fact-checking is not a gimmick. It is the proof that the map works. If you can take a claim someone made on a podcast, resolve it to the exact moment, match it to the research on that supplement and outcome, and show whether the evidence agrees — you have demonstrated that the connection layer is real, and useful, and something you cannot get anywhere else. The leaderboard is the surface. The ability to produce the leaderboard is the product.
I do not entirely know yet which version of Panacea wins. There is a version that is a beloved consumer research tool. There is a version that is the evidence layer other health software is built on. There is a version that is the place you go to find out whether the person you are listening to is telling you the truth. They are not mutually exclusive, but they pull development in different directions, and choosing wrong is expensive.
What I do know is that the framing — supplements as the topic, trust infrastructure as the product — keeps producing better decisions than the framing where supplements are the product. It tells me to invest in the map. It tells me to treat the connection between research and culture as the thing being built, not the byproduct. It tells me that the entry on ashwagandha is not the asset; the fact that the entry on ashwagandha is wired into a million structured claims and a thousand influencer statements is the asset.
That is what I am building, even when it looks, on the surface, like I am building a supplement site. The surface is the topic. The infrastructure is the point.
