biomni

A General-Purpose Biomedical AI Agent

0x9B74d455f6623341cCF04EE24b88Bd237D21a576

[biomni @biomni]

BIOMNI.

[CZ 🔶 BNB @cz_binance]

Read my book, Freedom of Money. From my childhood to founding Binance, challenges we faced, wandering the Earth, stories (from my perspective) about Luna/UST, FTX/SBF, dealing with US DOJ, SEC, going through jail (as the only person in US history to ever go to jail for a single violation of BSA), then full and unconditional pardon, and more...

Audiobook: https://t.co/Ow76tim5G4

English: https://t.co/UxgYxYJ3NF
Chinese: https://t.co/ItFd8FEyuK

All proceeds go to Charity.

First IRL book signing event in Dubai soon. Due to limited physical books available, will make the first event a small one.

[Stanford University @Stanford]

Stanford materials scientist Will Chueh, director of @StanfordEnergy, is merging AI and energy research to accelerate the world’s shift to sustainable, affordable power.

#ResearchMatters
https://t.co/l3SaFw9TtS

[Stanford University @Stanford]

Stanford and the U.S. Department of Energy’s @SLAClab will lead six Genesis Mission projects that will harness AI to accelerate advances in critical areas including battery waste, understanding the evolution of the universe, and plant genetics.
https://t.co/OWoeE2Ri0I

[biomni @biomni]

Official website

https://t.co/vhQHeRKym1

[NVIDIA AI Developer @NVIDIAAIDev]

👀 New livestream this Friday. Join our discussion with @Stanford researchers for an exciting look at Biomni — a next-generation biomedical AI agent transforming how labs explore data and design experiments.

Discover how Biomni, powered by NVIDIA DGX Spark, turns complex biomedical workloads into fast, automated insights — giving teams a powerful launchpad for rapid experimentation, evaluation, and seamless scaling.

📆 Friday, February 13, at 9 a.m. Pacific
🎥 Watch here on X, or on YouTube: https://t.co/fTOLeESuUa

[Thomas H. Chapin IV @tomchapin]

Biomni

[Biology+AI Daily @BiologyAIDaily]

Biomni: A General-Purpose Biomedical AI Agent

１．Biomni is a general-purpose biomedical AI agent that autonomously performs diverse biomedical research tasks, ranging from bioinformatics to experimental protocol design, without relying on predefined templates.

２．Its standout capability is executing full end-to-end scientific workflows: generating hypotheses, retrieving tools and datasets, reasoning through plans, writing code, and interpreting results—all autonomously.

３．Biomni achieves this by constructing a unified biomedical action environment (Biomni-E1) comprising 150 expert-curated tools, 105 software packages, and 59 databases covering 25 biomedical subfields.

４．An LLM-powered action discovery agent systematically analyzed 2,500 recent publications to define this action space, making Biomni the first agent with an explicitly structured, validated biomedical environment.

５．Biomni uses a novel code-centric agentic architecture (Biomni-A1) that dynamically retrieves relevant resources, generates executable code, and plans adaptively based on feedback during execution.

６．On three general biomedical QA benchmarks (HLE, LAB-Bench DbQA, and SeqQA), Biomni outperformed strong generalist baselines and even matched or exceeded expert-level human performance.

７．Across eight realistic biomedical tasks—including gene prioritization, drug repurposing, single-cell annotation, and microbiome profiling—Biomni showed 400% average improvement over base LLMs and significant gains over coding agents.

８．Case studies show Biomni autonomously analyzing wearable CGM and temperature data from 30 participants, uncovering personalized metabolic phenotypes and novel thermogenic patterns.

９．It also performed complex multi-omics analyses of 336,000 cells from embryonic skeletal tissue, identifying new regulatory TFs like AUTS2 and ZFHX3 potentially involved in human skeletal development.

１０．Biomni successfully designed and executed wet-lab validated molecular cloning experiments, rivaling expert scientists in accuracy and speed, including sgRNA design and Golden Gate assembly workflows.

１１．Importantly, Biomni provides a user-friendly graphical interface allowing scientists to submit natural language queries and receive detailed, traceable outputs—no coding required.

１２．While powerful, Biomni still faces challenges in nuanced clinical reasoning and historical knowledge integration, pointing toward opportunities for reinforcement learning, multimodal data support, and continual environment expansion.

１３．Biomni offers a scalable vision for AI-augmented biomedical discovery, where virtual agents collaborate with human scientists to drive innovation across research and healthcare.

💻Code: https://t.co/SUJjShxWPb
📜Paper: https://t.co/XS4Alfg3Q9
#Biomedicine　#AI　#LLM　#Bioinformatics　#DrugDiscovery　#PrecisionHealth　#MultiOmics　#SyntheticBiology　#BiomedicalResearch

[Jay Rughani @JayRughani]

Phylo's Biomni Lab lets biologists collaborate with AI agents to get more done, with rigor, at scale.

Excited to support @KexinHuang5 @jure @YuanhaoQ @lecong and the @phylo_bio team

[Young Kwang Chae, MD, MPH, MBA @youngkwangchae]

The future of cancer research is agentic AND reproducible. 🧬🤖
@jure from @Stanford presenting "Biomni." #AACR26 It’s more than a chatbot—it’s an integrated lab environment that truly closes the wet-dry loop.

The breakthroughs:
✅ Standardized Reproducibility: Automated, end-to-end workflows minimize human error and "tool-hopping" variability. 🔄
✅ Automated Lab Work: Executes complex molecular cloning protocols in ~10 mins. 🧪
✅ Robotic Integration: Connects directly with hardware (like PyLabRobot) for physical execution. 🦾
✅ Massive Efficiency: Data analysis cut from weeks of manual effort to 35 minutes. ⏱️
AI isn't replacing scientists; it’s scaling our impact, consistency, and precision. 🚀
@AACR @ProjectBiomni @OncoAlert

https://t.co/SjKeIQQZCS

[Andrew Dunn @AndrewE_Dunn]

EXCLUSIVE: a16z and Menlo Ventures are backing a new AI startup called Phylo, founded by two recent Stanford PhDs

Phylo's founders developed Biomni, a popular AI agent for biomedical research that debuted last year. I talked with them on what's next: https://t.co/ujkHKNPZPg

[biomni @biomni]

updated
https://t.co/vhQHeRKym1