Just finished this episode. I no longer work for the VA, but I did for 12 years. It’s one system that is offering open access to pharmacogenomic testing for free. I think it started about 4-5 years ago and not only includes the testing component but local expert clinical pharmacists who assist with interpretation of the results.
Thomas, this episode crystallized something I've been writing about from a different angle — the structural pattern where knowledge and capability exist to do the humane thing, but the capital structure routes the benefits upward and the costs downward. It's the story I know from the Ruhr Valley, and it's playing out again with AI.
The pharmacogenomics story is a perfect analog. The science exists. The FDA has the list. Over 150 drugs could be prescribed more safely if genomic data were used. But it doesn't happen — not because of a knowledge gap, but because the economic plumbing doesn't support it. No one with capital at stake has an incentive to deploy it.
That's AI displacement in miniature. The technology exists to augment workers and distribute productivity gains broadly. But the deployment pace isn't set by what's safe for the labor market — it's set by competitive pressure and shareholder expectations. Merck had internal emails showing they knew about the Vioxx heart attack risk and pushed the dose higher anyway. AI labs have internal research showing displacement risks and push deployment faster anyway. In both cases, the people making the decisions never personally bear the cost.
Dr. Topol calls it a moral line that isn't a legal line. That's the gap. And with AI, it's actually wider — pharma at least has the FDA and the possibility of class action suits. For structural economic displacement, there isn't even a regulatory vocabulary yet. No one gets Stevens-Johnson syndrome from losing their job to automation, so there's no visible injury to litigate. Just silence.
Appreciated the conversation and effort to resurrect pharmacogenetics - huge promise as you relate, and such nothingness (at least in the US).
For the most part still playing roulette.
Just finished this episode. I no longer work for the VA, but I did for 12 years. It’s one system that is offering open access to pharmacogenomic testing for free. I think it started about 4-5 years ago and not only includes the testing component but local expert clinical pharmacists who assist with interpretation of the results.
https://www.va.gov/files/2024-10/pgx-prescription24.pdf
Thomas, this episode crystallized something I've been writing about from a different angle — the structural pattern where knowledge and capability exist to do the humane thing, but the capital structure routes the benefits upward and the costs downward. It's the story I know from the Ruhr Valley, and it's playing out again with AI.
The pharmacogenomics story is a perfect analog. The science exists. The FDA has the list. Over 150 drugs could be prescribed more safely if genomic data were used. But it doesn't happen — not because of a knowledge gap, but because the economic plumbing doesn't support it. No one with capital at stake has an incentive to deploy it.
That's AI displacement in miniature. The technology exists to augment workers and distribute productivity gains broadly. But the deployment pace isn't set by what's safe for the labor market — it's set by competitive pressure and shareholder expectations. Merck had internal emails showing they knew about the Vioxx heart attack risk and pushed the dose higher anyway. AI labs have internal research showing displacement risks and push deployment faster anyway. In both cases, the people making the decisions never personally bear the cost.
Dr. Topol calls it a moral line that isn't a legal line. That's the gap. And with AI, it's actually wider — pharma at least has the FDA and the possibility of class action suits. For structural economic displacement, there isn't even a regulatory vocabulary yet. No one gets Stevens-Johnson syndrome from losing their job to automation, so there's no visible injury to litigate. Just silence.
https://wernerglinka.substack.com/p/not-yet