Trial Pre-Mortem
A clinical-trial pre-mortem agent: it tells a drug developer whether a trial is worth running and where it will die - then says "run this" or "do not run." - potentially saving hundreds of millions of dollars and letting
Core trick
LLM reasons over a trial protocol against past failed mechanisms to flag design risks and output a run or do-not-run verdict.
Description
A clinical-trial pre-mortem agent: it tells a drug developer whether a trial is worth running and where it will die - then says "run this" or "do not run." - potentially saving hundreds of millions of dollars and letting clinical trial developers focus on better designs or repurposing efforts.
README excerpt
# Trial Pre-Mortem (PSP) ## The problem In rare neurodegeneration, a single clinical trial costs years and nine figures — and program after program has died re-running a mechanism that **already failed**, sometimes despite demonstrably hitting its molecular target. Progressive Supranuclear Palsy (PSP), a 4-repeat tauopathy with no disease-modifying therapy, is the clearest example: a string of anti-tau antibody trials engaged the target and produced **zero clinical benefit**. The most valuable thing a tool can tell a drug developer is often *"do not run this trial as designed"* — before the spend, not after. This is a **pre-mortem**, not a dashboard. Given a protocol synopsis (population, mechanism, endpoints, powering, design), it doesn't score-and-display — it **reasons**: it proposes findings about where the trial will die, critiques them against public ground truth, and **kills any finding it cannot substantiate from a retrievable source.** The output is a ranked kill-list plus either a defensible redesign or a reasoned "do not run." ## Who it's for Clinical-development strategists and biotech execs in rare neurodegeneration deciding whether to greenlight a program, and translational investors sizing the risk — the people about to commit years and capital to a mechanism that may have already failed in disguise. ## How it works A **builder** (Opus 4.8) retrieves public data (ClinicalTrials.gov v2 + web search/fetch) and drafts per-axis findings. An **independent ver