Schrödinger deploys its agentic AI co-scientist Bunsen across Bristol Myers Squibb 🔗

Schrödinger (Nasdaq: SDGR) announced on August 5 a strategic agreement to deploy Bunsen, its agentic AI co-scientist, at scale inside Bristol Myers Squibb's research organization. Bunsen is built specifically for molecular discovery and is optimized to run Schrödinger's physics-based computational methods, planning and executing multi-step discovery workflows and interpreting the results, rather than acting as a general-purpose assistant. The deal expands a long-standing software relationship and adds joint development of new Bunsen functionality, including integration with RetroSynth, Schrödinger's AI synthesis-planning platform. No financial terms were disclosed. It is one of the first cases of an agentic co-scientist moving into a major pharmaceutical company's research organization at scale rather than as a pilot.

Jeff Dean and three senior Google AI leaders leave to found Discovery Loop 🔗

On August 5, Jeff Dean, Google's chief scientist, said he is leaving after 27 years to co-found Discovery Loop, a Palo Alto public benefit corporation built to automate the experimental loop of research. He is joined by Sanjay Ghemawat (Google senior fellow), Oriol Vinyals (a DeepMind VP and Gemini technical lead), and Quoc Le (a Google Brain co-founder), with Dean as CEO. Alphabet is a founding investor and cloud partner, and Radical Ventures and Khosla Ventures are co-leading the seed round. The company plans to begin with machine-learning research itself and later extend into areas including hardware design, drug discovery, and clean energy. The move came alongside a wider Google DeepMind reshuffle, with Demis Hassabis becoming chair of DeepMind and chief scientist of Alphabet. The direct life sciences link is thin for now, drug discovery sits later on the roadmap. What matters is the caliber of talent leaving Google to automate scientific discovery full-time.

GenScript wires an AI design platform straight into its wet lab 🔗

On August 5, GenScript Biotech and Tamarind Bio announced a partnership connecting Tamarind's AI molecular-design platform, a single interface to more than 300 computational biology models, directly to GenScript's synthesis, expression, and testing services. Researchers can submit AI-generated sequences and get experimental data back in as little as four days, according to the companies. The framing is the same one running under this week's phage story: AI can now generate far more candidate molecules than any lab can test, so fast experimental validation is becoming the real constraint. GenScript, founded in 2002, is positioning its wet lab as the validation engine for AI-designed discovery.

Pathos AI adds two cancer programs worth up to $2.2 billion 🔗
Pathos AI, a clinical-stage company that runs its pipeline through an AI platform called Foundry, announced on August 3 that it in-licensed JSKN016, a TROP2/HER3 bispecific antibody-drug conjugate, from China's Alphamab Oncology. Pathos is paying $125 million upfront, with up to $2.093 billion in milestones plus tiered royalties, for rights outside greater China. Separately, Pathos signed a co-exclusive collaboration with AstraZeneca to take AZD4241, a preclinical oral estrogen-receptor PROTAC for ER+/HER2- breast cancer, into the clinic; financial terms of that deal were not disclosed. CEO Iker Huerga said the bottleneck in oncology is not finding molecules but proving they work in the right patients.

The signal: An AI-native biotech is now doing dealmaking at the scale big pharma does, using its own platform as the asset-selection and trial-design engine. Both assets are molecules that already exist, and Pathos is betting its AI earns its return further down the pipeline, in patient selection and trial design rather than in molecule invention. That points to where the industry increasingly expects AI to pay off in drug development, which is clinical execution.

Aiforia raises €20 million in EU venture debt for AI pathology 🔗

Aiforia, a Helsinki-based digital-pathology company founded in 2013, signed a €20 million venture debt agreement with the European Investment Bank on August 3, backed by the European Commission's InvestEU programme. The financing comes in three tranches, starting with €5 million, with each tranche tied to revenue and other interim targets. Aiforia's deep-learning software analyses high-resolution biopsy images to help pathologists detect cancer, and it already markets CE-IVD marked clinical models in Europe.

The signal: This is public European money going into home-grown clinical AI, and it is going to a regulated diagnostic product rather than a consumer wellness tool. The structure matters too. Venture debt tied to revenue milestones treats Aiforia as a commercial clinical business with sales to grow, not as an early research bet. Expect more of Europe's AI diagnostics to be financed through public and development-bank channels as the region works to keep clinical AI capability on the continent.

Hackensack Meridian becomes the first US health system certified for responsible AI use 🔗

Hackensack Meridian Health, New Jersey's largest health network, said it is the first US health system to earn the Joint Commission's Responsible Use of AI in Healthcare certification, announced at the end of July. The Joint Commission launched the voluntary program on June 1. It assesses an organization across five areas: governance, data management, risk and bias reduction, monitoring and validation of AI performance, and transparency and staff training. The certification evaluates how an organization governs AI, and it does not validate individual AI products.

The signal: Governance is becoming something health systems can be formally certified on, which turns responsible AI use from a stated value into an auditable standard. With the Joint Commission noting that more than 80% of physicians already use AI in some form, the pressure is shifting from whether to adopt AI toward whether an organization can prove it is doing so safely. For vendors, this raises the bar. Selling into hospitals will increasingly mean fitting into a documented governance framework, not only clearing a procurement checklist.