Large language models (LLMs) are useful for summarizing publicly available information and brainstorming ideas. However, they typically rely on a limited subset of readily accessible sources, may overlook important publications, and often reproduce conclusions already presented by the original authors rather than critically evaluating the underlying evidence.
At LumiRare, we first identify and analyze the primary literature ourselves. We then compare findings across studies, assess which observations are truly relevant to the specific patient, identify gaps and inconsistencies in the evidence, and distinguish well-supported conclusions from hypotheses and speculation.
The main value of a LumiRare report is therefore not simply collecting information, but proposing scientifically grounded next steps. Depending on the case, this may include alternative therapeutic strategies, disease mechanisms worth investigating, compensatory pathways, suitable experimental models, priorities for future studies, and experiments that can indirectly support or refute competing hypotheses.
In many ways, our reports resemble the scientific reasoning behind a research grant proposal. While AI can help draft text, the quality of a research strategy depends on asking the right questions, selecting informative experiments, understanding the strengths and limitations of different approaches, prioritizing the evidence, and anticipating which results will meaningfully guide the next decisions. That analytical and strategic layer is the core of every LumiRare report.