Longevity InTime: Autonomous AI Institute. Anti-Aging Digital Health Immortality Transhumanist AI Channel

AI Drug Search Patrick Maloone proposed evaluating AI-powered drug searches by selecting candidates for laboratory testing on August 20. In his model, scientists guide multiple specialized programs that narrow down thousands of protein or molecule variants to a short list for laboratory testing. The model can suggest many plausible protein or molecule variants, but the laboratory can only synthesize and test a small fraction. Maloone's approach is based on the recent Claude campaign, which designed miniproteins. The agent selected a protein segment, design programs, and candidate lists, and two laboratories then tested whether they bound to the target protein. Many decisions need to be made between the model's request and the experimental result, and candidate selection is already a separate task. In a blind test of 511 antibodies in the second task, most AI participants performed worse than a basic strategy in selecting variants that bound to the target protein. Laboratories need a list of candidates that are most likely to yield results in the next experiment. In Maloone's scenario, the scientist assigns programs to separate stages: literature search, modeling, design, result analysis, and experiment planning. The knowledge that usually remains in the specialist's head is proposed to be transformed into rules, examples, tools, and checks accessible to programs. To prioritize candidates, one needs to select a program for a specific question, distribute computations, implement control checks, preserve data provenance, and set evaluation rules. Then, one needs to troubleshoot failures, as described in Nature Aging, July 2026. 🔗 Read original →