A Startup Used AI to Make a Psychedelic Without the Trip (www.wired.com)

Mindstate Design Labs announced a promising Phase I result for MSD-001, an AI-designed oral formulation of 5‑MeO‑MiPT (“moxy”) that produced robust psychoactive effects without the classic hallucinogenic “trip.” Using models trained on biochemical profiles and more than 70,000 trip reports (from clinical datasets, forums, social media and other sources), the company designed a compound that selectively targets the serotonin 2A receptor with fewer off‑target interactions. In a 47‑participant trial, MSD‑001 was safe and well tolerated across five doses, produced heightened emotion, associative thinking and brighter perception but not hallucinations or ego‑dissolution; onset was ~30 minutes, peak at 1.5–2 hours. Researchers validated effects with standard psychedelic scales, eye‑tracking and brain imaging that showed brain‑wave patterns similar to psilocybin, indicating the drug engages intended neural mechanisms. For the AI/ML community this is a concrete demonstration of multimodal, data‑driven phenotypic drug design: models mapping molecular signatures to subjective experience enabled a targeted compound that decouples neuroplasticity‑promoting effects from intense perceptual disruption. The work highlights opportunities (precision psychopharmacology, expanded treatment populations) and challenges (noisy, biased trip‑report data, interpretability, and tough regulatory paths—e.g., recent FDA scrutiny of MDMA trials). MSD‑001’s “psychedelic‑tofu” strategy—using a stripped‑down agonist as a modular base to combine with other agents—could reshape how AI is applied to design therapeutics that tune conscious states while raising important safety, ethical and validation questions going forward.