The FDA’s attempt to pull the approval of the drug Tavneos over supposedly manipulated study results is a chilling reminder of how hostile Washington has become to the founder mindset. It is a fundamental misunderstanding of what it takes to ship a product in the modern era.
If you look at the pharmaceutical industry from first principles, a clinical trial is really just an onboarding funnel. You have a cohort of users, a new feature, and a desired engagement metric. When a consumer software startup notices a dip in their conversion rates, they don't scrap the entire app. They tweak the dashboard. They smooth out the metrics. They tell a better story. But when the visionary builders behind Tavneos applied this exact same framework to their FDA submissions, the regulatory state panicked.
I was having matcha in Hayes Valley last week with a three-time accelerator alumni who recently pivoted from decentralized finance into synthetic nephrology, and his frustration was palpable. The smartest people I know are increasingly terrified to build in the biological space. They look at what is happening to a drug that was approved, shipped, and is currently generating revenue, and they wonder why they should even bother trying to cure diseases if the government is going to nitpick the Excel spreadsheets that got them there.
We have to stop treating physical reality as an immutable constraint. When the FDA accuses a biotech company of manipulating study data, what they are really criticizing is a company's roadmap. The founders weren't lying about the drug's efficacy; they were simply presenting the FDA with a forward-looking projection of how the molecule would perform once it achieved product-market fit inside the human immune system. Punishing them for being optimistic is anti-progress.
If we restrict our portfolio companies to reporting only the things that actually happened in physical space, we are effectively capping our total addressable market at whatever the universe currently allows.
The discourse around faked data completely ignores the sheer amount of sweat equity required to falsify a Phase III clinical trial. Do you have any idea how grueling it is to manually adjust mortality rates across a massive dataset while maintaining statistical plausibility? That is relentless execution. That is founders sleeping under their desks to ensure the placebo group perfectly aligns with the investor pitch deck. To dismiss that kind of raw hustle as mere fraud is deeply disrespectful to the grind.
I recently spoke with a founder who spent Thanksgiving weekend locked in a co-working space, manually migrating adverse trial reactions into a folder labeled 'Anomalies.' He didn't do that because he is a criminal. He did it because he is obsessed with product velocity. He understands that if you pause your sprint cycle every time a patient's kidney fails, you are going to lose your competitive advantage to a foreign biotech firm that doesn't share your squeamishness.
We celebrate founders who distort reality in every other vertical. When a consumer hardware CEO convinces the public that a phone without a keyboard is a feature, we write ten thousand essays about his genius. But when a biotech founder attempts to distort the reality of whether or not a pill causes liver failure, suddenly they are threatened with a subpoena. The hypocrisy is staggering.
And what about the end users? The FDA claims it wants to protect patients by pulling Tavneos off the market, but this betrays a complete ignorance of user psychology. Have regulators considered that patients want to feel hopeful? To someone with a rare autoimmune disease, manipulated data is just another word for a premium user experience. By demanding rigorous empirical proof, the FDA is actively degrading the patient's right to believe they are participating in a successful launch.
We prefer to think of the altered control group data not as a fabrication, but as a high-fidelity simulation of the reality we intend to build in Q4.
The obvious solution here is to abstract the biological humans out of the clinical trial stack entirely. The fact that we are still relying on wet, unpredictable mammalian bodies to generate regulatory data is a massive bottleneck. If you need a dataset proving that a compound reverses inflammation, you shouldn't have to wait three years for a randomized cohort to metabolize it. You should be able to spin up a cloud server, prompt a large language model to simulate five hundred perfectly cured patients, and submit the outputs to Washington.
When an algorithm hallucinates a legal precedent, we call it a training error. But if an algorithm can hallucinate a complete recovery from severe vasculitis, who are we to argue with the output? The model is simply predicting the most optimal biological scenario based on the prompt we gave it.
If Amgen is forced to pull Tavneos over a few creatively optimized data points, it will trigger an economic freeze across the entire ecosystem. Venture capital will dry up for any founder brave enough to ship a therapeutic before the science is fully baked. We will be stuck in a stagnant loop of only approving drugs that actually work.
The regulatory state needs to realize that objective truth is just a legacy protocol, and a highly inefficient one at that. We are trying to build the future of human health. The future does not have time to wait for a control group.