Schrödinger, the computational chemistry company whose physics-based simulation software is used in drug design, announced the launch of an early access version of Bunsen, an "agentic AI co-scientist" for molecular discovery.
The framing matters because of who is doing the launching. According to Endpoints News, Schrödinger CEO Ramy Farid ruffled feathers three years ago by pushing back against the AI enthusiasm of the moment, saying hype had reached "dangerous" and "insane" levels. Endpoints reports that Bunsen is an AI agent designed to use Schrödinger's own physics software — a distinction that fits Farid's earlier argument. Rather than asking a model to guess which molecules might work, the agent drives simulations grounded in physical chemistry.
That design choice speaks to a bottleneck the field is now openly discussing. A MIT Technology Review piece, produced in partnership with Cytiva, notes that AI is identifying new therapeutic targets faster than ever, but that this speed is exposing physical bottlenecks in the lab and a need for better data. Software can propose candidates far faster than benchtop experiments can confirm them, and models trained on thin or noisy data inherit those limits.
Schrödinger's stated approach — pairing an AI agent with established simulation tools — is one answer to that gap: use computation that encodes known physics as the check on what the AI suggests.
Details on pricing, availability beyond early access, and validation data were not included in the announcements summarized here.
Why it matters: drug discovery is slow and expensive largely because most candidate molecules fail late, and how the industry connects AI's speed to real physical and experimental evidence will determine whether these tools shorten that timeline or simply generate more guesses to test.