Human organoids that mirror their corresponding organs in cell-type diversity present an opportunity to perform large-scale screens for compounds that protect disease-affected or damage healthy cell types. However, such screens have not yet been performed. Here, we generated 20,000 human retinal organoids with GFP-labeled cone photoreceptors. Since degeneration of cones is a leading cause of blindness, we induced cone death and screened 2,707 compounds with known targets, for those that saved cones or those that further damaged cones. We identified kinase inhibitors that protected cones in both the short and longer term, HSP90 inhibitors that saved cones in the short term but damaged them in the longer term, and broad HDAC inhibition by many compounds that significantly damaged cones. This resource provides a database for cone-damaging compounds, and it describes compounds that can be starting points to develop neuroprotection for cones in diseases such as macular degeneration.