It feels like we’re hitting a new phase of AI in health and science. This week alone, we’re learning about AI models that can design genes, predict how proteins move, and even create two-way brain-computer communication. Far from incremental advances, these feel like breakthroughs that could completely change how we approach medicine and biotech.

How does all this translate into business outcomes? Funding is still flowing, with investors betting big on AI-driven healthcare startups. At the same time, public markets are sending mixed signals, with some AI health companies thriving and others struggling post-IPO.

But it’s early days, and the momentum is undeniable. With AI accelerating discovery at a crazy pace, it’s only a matter of time before these breakthroughs start reshaping healthcare and biotech in ways we’ve never seen before. And honestly, these might be the most exciting AI headlines I’ve seen in any sector (there’s only so many "AI for copywriting/documentation/emails" startups I can take).

On to this week’s developments…

📢 Headlines

AI can now model and design the genetic code for all domains of life 
(Arc Institute) 

The Arc Institute and NVIDIA have developed Evo 2, which is like ChatGPT for DNA, except instead of generating text, it deciphers the code of life. By analyzing DNA from over 100,000 species, this AI can predict which mutations might cause diseases, help scientists understand how life evolved, and even design new genes for research. It’s open-source, meaning researchers worldwide can use it to accelerate discoveries in medicine, conservation, and biotech.

World's first two-way adaptive brain-computer interface (Life Technology)

Brain-computer interfaces (BCIs) have been around for a while, but they’ve mostly been one-way streets, reading brain signals to control devices. Now, researchers in China have built the first two-way adaptive BCI, which not only reads brain activity but also sends information back, allowing real-time feedback between the brain and the device. This could be a huge leap for people with paralysis, helping them regain movement with more natural control, or even letting prosthetic limbs “feel” sensations.

AI tops surgeons in writing post-operative notes (Reuters)

Operative notes are critical for patient care, billing, and surgical quality tracking. But they’re often tedious to write and prone to errors. A new study found that AI-generated reports from prostate surgeries were more accurate than those written by surgeons, with fewer discrepancies (29% vs. 53%). The AI system “watched” procedures, identified key surgical steps, and compiled detailed reports automatically.

MSK Cancer Center and AWS to accelerate AI-driven cancer innovation (MSKCC)

Memorial Sloan Kettering (MSK) is leveraging over a century of oncology expertise and vast patient data to train AI models that could transform cancer care. Partnering with AWS, MSK aims to use machine learning to predict how tumors evolve, identify the best treatments, and uncover hidden patterns in genomic and imaging data.

Healthcare and AI is a hot combination for startups (Crunchbase)

AI-driven healthcare startups are seeing record funding, but their public market performance is mixed. Tempus AI, a precision medicine company, went public in June and now boasts a $11 billion market cap, while Metagenomi, which applies AI to genome editing, has lost over 70% of its value since its IPO. Despite setbacks, investors remain bullish, betting that AI’s ability to speed up drug discovery and diagnostics will fuel the next wave of biotech breakthroughs.

 🔎 Patient Perspective

In most discussions about health AI, we’re hearing mainly from providers, hospitals, and tech companies. We rarely hear patients’ perspectives, and yet they’re arguably the most important stakeholder in these debates. This new section of the Roundup aims to highlight patient voices about AI.

This week, Jennifer Goldsack, CEO of the Digital Medicine Society (DiMe), shared on LinkedIn how she’s using AI for her cancer diagnosis:

See Jennifer Goldsack’s post here and the linked article here.

🧪 Research Spotlight: AI that predicts how proteins move - 100,000x faster

Proteins are constantly changing shape to carry out essential functions in the body, from fighting infections to processing nutrients. But simulating these movements has always been painfully slow, requiring months of supercomputer time just to model a fraction of a second of protein motion. Microsoft Research’s BioEmu-1 is changing that. This new AI model predicts how proteins move, generating thousands of structural variations per hour—100,000x faster than traditional methods—while maintaining the same level of accuracy.

Key takeaways:

• ⚡ Superfast predictions – BioEmu-1 can generate protein structures in minutes instead of months, making drug discovery and biological research way more efficient.

• 🧬 Trained on massive data – The model learned from 200 milliseconds of molecular simulations, 9 trillion DNA base pairs, and 750,000 protein stability measurements, giving it a deep understanding of protein behavior.

• 🔬 Highly accurate – BioEmu-1 matches real-world lab measurements, even for proteins it’s never seen before, meaning it could help scientists design better-targeted drugs and treatments.

• 🌍 Free for researchers worldwide – Microsoft is open-sourcing BioEmu-1 via Azure AI Foundry Labs, making the tool available to scientists everywhere.

Read more about it on Microsoft Research blog and bioRxiv.

💸 Funding 

Abridge raised $250 million to enhance its AI capabilities in automating clinical notes and medical documentation for healthcare providers.

OpenEvidence raised $150 million to develop AI that reads and synthesizes medical literature to help doctors make evidence-based decisions in real-time.

Frontera Health raised $32 million to use AI for improving autism diagnosis and treatment for children.

Keragon raised $7.5 million to develop its no-code healthcare automation platform.

💻 Job Opportunities in Health AI

Health AI Scholar, Samsung Research America Digital Health

Director, AI, Headspace

Product Manager, Platforms, Abridge

That’s all for this week 👋 See you next time.