A crucial challenge in the quest to reduce the social and economic burdens of Alzheimer’s disease is the availability of sensitive cognitive tools for detecting very early cognitive abnormalities. Current methods require neuroimaging techniques to find appropriate candidates for clinical trials. However, high cost of scans and the burden on participants to travel to and from medical research facilities is a key limitation. We designed a web-based task (ORCA) that assesses individuals’ ability to learn new information. This involves learning a large set of Chinese characters over 6 days. Pilot data shows learning rates are reduced substantially in cognitively normal older adults with abnormal Alzheimer's pathology compared to those with normal levels (1.5 standard deviations, p<.000001). This is ~4 times greater than effects observed for conventional memory tests applied over 1 year. This is the first time any effect of such magnitude has been observed in this earliest stage of Alzheimer’s disease, and this has been achieved over 1 week. We now aim to validate and extend these results in other large Alzheimer's cohorts, to determine task efficacy for pre-screening candidates for clinical trials, and to develop parallel forms allowing a broader range of novel information that may be useful.