Instructional BiologyMost public conversation about drug development failure centers on cost — the widely cited figure, drawn from published health-economics research, is somewhere north of a billion dollars in average development spend per approved therapy, with the large majority of candidates that enter clinical trials never reaching approval at all. Cost is real, and it shapes which diseases get pursued and which don't. But cost is also the easiest part of this problem to talk about, because it's the part with a number attached.
The harder part to quantify is what happens to the people on the other side of a trial that fails, or a therapy that takes another five years to reach approval because the first three attempts at a mechanism didn't hold up. That cost isn't measured in dollars. It's measured in the years someone spent managing a condition with tools that only partially worked, in the uncertainty of not knowing whether the next option would be better or just different, and in the discouraging reality that a participant's time and effort in a failed trial doesn't disappear — it's simply not converted into anything that reaches other patients.
Why the timeline matters as much as the outcome
A therapy that eventually works but arrives a decade later than it could have isn't the same as a therapy that never works. But for the person managing a condition in the meantime, the practical difference can be smaller than the framing suggests. Late is still a cost. It's just a cost that's harder to put on a balance sheet, because it's distributed across individual lives rather than concentrated in a single line item a company reports.
This is part of why we think the industry's habit of measuring success primarily at the population level — did the treatment arm outperform placebo across the group — is worth examining more closely. A trial can clear the statistical bar for significance and still describe meaningfully different experiences happening inside the same treatment arm: people who responded well, people for whom nothing changed, and people who didn't tolerate it. A p-value tells you the groups differed. It doesn't tell you which side of that difference any one person actually landed on, and that gap is where a lot of the timeline cost above gets absorbed silently.
What we think is worth building toward
Closing that gap is a large part of why we're building the way we are — evidence-linked reasoning that stays connected to individual context rather than collapsing everything into a single population-level average, and a research process designed to surface which questions are actually worth testing next, rather than which ones are easiest to test. None of this shortens the road to a validated therapy on its own. What we think it can do is reduce how often research spends years finding out, the hard way, that a promising-looking mechanism doesn't hold up outside the population average it was measured against.
We don't think there's a shortcut around rigorous, controlled validation, and we're not proposing one. What we're proposing is that the questions guiding which candidates reach that validation stage in the first place can be asked more precisely — and that the difference between asking a better question and asking the same question faster is, over enough iterations, the difference this field actually needs.