Weekly Recap (Nov 26, 2025)
R updates (R Data Scientist, R+AI conference, R weekly), Claude 4.5 Opus, Genesis Mission, AI+science, AI updates (Posit, AI Data Scientist), missing heritability, AI+edu, biotech, AIxBio, new papers
It’s a short week here in the U.S. so I’m publishing my Friday weekly recap a few days early. Thanks for reading, and enjoy the long weekend!
NY Times (gift link): A.I. Is a Bubble. Maybe That’s OK.
Posit Blog: Privacy and AI Assistants.
New paper: Orchestrating Spatial Transcriptomics Analysis with Bioconductor. See also the full e-book free online: bioconductor.org/books/OSTA.
R weekly 2025-W48: Multi-agent orchestration, just another R linter, and managing, branding, and accessibility of R projects.
The R Data Scientist 2025-11-25: Community Roundups, R & AI, Packages & Development, Geospatial Projects, Statistical Methods, Academic Research.
Quarto talks from posit::conf(2025).
Shutdown fallout on NIH: >370 study sections, >24,000 apps were bumped. To keep Jan+May 2026 councils on track, most panels will now only discuss ~30-35% of applications, add a new “competitive but not discussed” tier, and use shorter bullet-style summaries.
Claude Opus 4.5 System Card with benchmarks on bioinformatics, virology uplift, lots of CBRN safety analysis and red teaming results.
White House dropped an Executive Order creating the “Genesis Mission,” a DOE-led national push to use AI and federal scientific data to accelerate discovery. I wrote a short summary a few days ago:
I’ve been trying out a few different platforms for reading/understanding, and writing scientific papers, including Edison (edisonscientific.com), q.e.d. (qedscience.com), and Nature Research Assistant (natureresearchassistant.com). I’ll probably have a longer post about one or more of these in the near future. I’ve really enjoyed Nature’s tool for reviewing my manuscripts prior to submission to get immediate and actionable feedback.
Speaking of AI for science, last week OpenAI shared 13 early experiments where GPT‑5 accelerated research tasks in math/physics/biology/materials; in 4, it helped find proofs of previously unsolved problems. Blog post: Early experiments in accelerating science with GPT-5. Paper: Early science acceleration experiments with GPT-5. This was also discussed in last week’s episode of the OpenAI Podcast.
Posit blog: 2025-11-21 AI Newsletter: Model releases (Gemini 3, GPT-5.1, Kimi K2 Thinking), Token-oriented object notation (TOON), Posit news.
The AI Engineer 2025-11-25: Headlines, Company Engineering Blogs, Applied AI Use Cases, Open Models & Infrastructure, Agents & Multimodal Systems, RAG & Web Retrieval, AI Research Frontiers, Academic Research.
The btw package is now on CRAN. btw is a toolkit that makes it easier to build what you want with ellmer. It provides tools to read R package documentation, inspect data frames, perform git operations, and more. These tools are available through multiple interfaces, including an MCP server, an interactive app, and simple function calls for quick copy-pasting.
Derek Lowe: LLMs for Medical Practice: Look Out.
Sasha Gusev: The missing heritability question is now (mostly) answered.
Chronicle: How AI Is Changing Higher Education.
STAT: Biotech VC predicts more M&A, commercial future for gene editing.
Report Launch: Safeguarding AIxBio Capabilities to Prevent Global Catastrophe.
Owl Posting podcast: Bringing organ-scale cryopreservation into existence (Hunter Davis, Ep #6).
Elliot Hershberg: On AI Infrastructure in Biology Three stories; six parameters.
Asimov Press: The Penicillin Myth. Competing theories seek to explain inconsistencies surrounding Alexander Fleming’s famed discovery.
Finally, a few other papers and preprints that caught my attention this week:
Genome region aware CADD thresholds for noncoding variant prioritization
Transforming Biological Foundation Model Representations for Out-of-Distribution Data
The Era of Agentic Organization: Learning to Organize with Language Models
Human Genome Diversity Project data use and implications for the governance of legacy genomic data
To be or not to be a protein coding mutation, that’s the question!
Data-driven strategies for drug repurposing: insights, recommendations, and case studies
The Genesis Mission sounds intresting. Im curious how the DOE will coordinate all that federal scientific data for AI acces. Thanks for curating these updates!