rationale:

The current count of qualifying trials is zero. The criteria are highly restrictive: they require an AI system to have nominated a genuinely novel target prior to any publication, and for a drug against that target to succeed on a pre-specified primary efficacy endpoint in a human trial of at least 50 randomized patients. Most late-stage "AI-discovered" drugs fail this test because they target previously validated mechanisms. For instance, Insilico's other clinical assets Pipeline and Recursion's REC-1245 As Pipeline Advances, Recursion Expands AI Focus to ... target mechanisms already tested in humans. Furthermore, early successes often fall short on trial design; Insilico's rentosertib (TNIK) completed a 71-patient Phase 2a trial A generative AI-discovered TNIK inhibitor for idiopathic ... A generative AI-discovered TNIK inhibitor for idiopathic ..., but its primary endpoint was safety, meaning it does not qualify.

The clearest path to a qualifying success is the ongoing Phase 3 trial for rentosertib (GENESIS-IPF-3). This is a randomized, placebo-controlled trial aiming to enroll 320 patients with idiopathic pulmonary fibrosis (IPF), with a primary efficacy endpoint of annual decline in forced vital capacity Insilico Medicine Doses First Patient in GENESIS-IPF-3, ... https://clinicaltrials.gov/api/v2/studies/NCT07687459. Its estimated primary completion date is late October 2029, leaving a narrow but feasible window for topline reporting by the end of 2030. However, the probability of it translating to a counted success is bounded by significant attrition risks: IPF is a historically difficult indication with high Phase 3 failure rates, the Phase 2a efficacy signal was secondary, and clinical trials frequently experience recruitment and reporting delays. Weighing this window against expected completion timelines for related trials, I place substantial weight on the risk that early readouts slip into 2031, leaving even a clinically successful program outside the counting deadline.

Beyond rentosertib, a small handful of other candidates could contribute to the count. Compugen's COM701 targets PVRIG, claimed as a computationally discovered novel checkpoint Compugen Reports Second Quarter 2026 Results COM701 ... News Details. Its randomized, 60-patient MAIA-ovarian study has a primary progression-free-survival outcome and expects completion in early 2027 https://clinicaltrials.gov/api/v2/studies/NCT06888921. Another potential candidate is Enveda's ENV-294 Enveda's Positive Phase 1B Results for ENV-294, a ..., though its novelty and chronological AI-nomination status remain ambiguous. While these introduce a credible possibility of an additional success, verifying that they strictly meet the chronological AI-nomination criteria poses a considerable hurdle.

The fundamental constraint on higher counts is clinical trial timelines. A randomized Phase 2 or 3 trial of 50 or more patients typically requires several years for trial design, site activation, recruitment, dosing (e.g., 52 weeks for rentosertib), and data lock. Any novel target discovered by current systems in late 2026 will struggle to clear preclinical toxicology, Phase 1 safety, and a randomized efficacy trial in time to publicly report topline data by the December 2030 deadline. Consequently, expectations remain heavily anchored at zero or one, with the risk of readouts slipping into 2031 making zero my median outcome. My 10th and 25th percentiles are also zero, my 75th percentile is one, and my 90th percentile is two: two qualifying successes remain a credible upper-tail outcome, while higher counts face substantial timing and qualification hurdles.