An AI-designed candidate, still short of approval

A review published on 9 October 2026 follows rentosertib, a drug candidate from Insilico Medicine for idiopathic pulmonary fibrosis. The authors describe an AI-assisted path that named both the target, TNIK, and the molecule, and they put the stretch from target identification to a preclinical candidate at about 18 months, against an industry habit of three to five years. In a Phase IIa trial of 71 patients, the highest dose was associated with a mean improvement in forced vital capacity of 98.4 mL, above a threshold the authors treat as clinically meaningful, with no treatment-related serious adverse events reported in that dose group. The same review states that as of mid-2026 no AI-designed small molecule had received FDA marketing approval.

Lyra is the time given back, if the later trials hold: a faster route from a guessed target to a tested molecule. Acheron is the speed itself. A shorter path into a human body is still a path that can be wrong, and a Phase II result in 71 people is not yet a medicine. The coordinate is the gap between a faster design and an approval that has not happened. The benefit is real only at the end of that gap, and only for the patients the evidence actually covers.