FALSIFY: Adversarial Research Pre-Mortem

FALSIFY is an adversarial research pre-mortem built to challenge hypotheses before time, money, and experimental effort are committed. A researcher enters a hypothesis, context, existing evidence, constraints, and the decision they are trying to make. FALSIFY then decomposes the hypothesis into testable claims, identifies hidden assumptions, surfaces confounders and competing explanations, highlights the most dangerous failure mode, and proposes falsification tests that could genuinely weaken the original idea. Instead of ending with another AI chat transcript, FALSIFY produces a structured research workflow with a minimum next experiment, evidence gaps, decision boundaries, and a printable Research Passport. Investigations can also be saved locally and exported as Markdown or JSON. The goal is not to replace literature review, peer review, domain expertise, or experimentation. It is to make uncertainty explicit earlier and help researchers ask better questions before building confidence around a claim. FALSIFY was built using Native.builder and uses OpenRouter free-model inference with structured outputs. It is designed as a focused research tool rather than a general-purpose chatbot.