scTREND: An annotation-free single-cell time-resolved and condition-dependent hazard model www.biorxiv.org/content/10.6... 🧬💻🧪 github.com/R301Carbine/...

A large language model for complex cardiology care www.nature.com/articles/s41...

Tried out this “for you” feed here. Man what a tire fire. Hot takes, rage bait, about as bad as X. No thank you.

"Human ergonomics matters less when humans aren’t the primary authors. […] I still love the language, but it seems clear that I and much of the industry will be writing less and less of it." wesmckinney.com/blog/agent-e...

AI-Enabled Biological Design and the Risks of Synthetic Biology: #Biosecurity implications for what AI can and can't do in biology today. blog.stephenturner.us/p/ai-enabled...

Attendance plummets at some science meetings. Travel bans. Gov Shutdowns. Funding. www.science.org/content/arti...

International AI Safety Report 2026 internationalaisafetyreport.org/publication/...

GPT-5 lowers the cost of cell-free protein synthesis openai.com/index/gpt-5-... Paper (OpenAI + Ginkgo Bioworks): Using a GPT-5-driven autonomous lab to optimize the cost and titer of cell-free protein synthesis (pdf) cdn.openai.com/pdf/5a12a3bc...

Started poking around with the paper-2-web Claude skill. Uploaded a PDF of one of my recent papers. "Make it a website". 5 minutes later: genome-engineering-biodiversity.netlify.app

No joke: I got angry hate mail today for writing an obituary of a Black woman scientist—because the person felt she did didn’t deserve the recognition. Which just makes me want to share it again: www.nature.com/articles/d41...

Navigating ethical, legal and social implications in genomic newborn screening rdcu.be/e2Gu7

Weekly Recap (February 6, 2026): Positron + Jupyter, Claude Opus 4.6, Vicki Boykis at AMLC, Emily Riederer, Wes McKinney, Codex, CC for science, #Rstats updates (R Data Scientist, R Works, R Weekly, R-Ladies, rOpenSci), AlphaGenome blog.stephenturner.us/p/weekly-rec... 🧬💻🧪

Alternative headline: "Faculty recruitment and retention is failing at most of Oklahoma's public colleges, governor says"

I see that @wjakethompson.com updated the #Rstats {taylor} package with lyrics from Showgirl. taylor.wjakethompson.com. Let's do a little text mining...

Biological data governance in an age of AI www.science.org/doi/10.1126/... #biosecurity

I appreciate the choice here - more AI features are coming to Firefox, but Firefox gives you the ability to turn any or all of them off. blog.mozilla.org/en/firefox/a...

How we’re helping preserve the genetic information of endangered species with AI blog.google/innovation-a...

Claude Opus 4.6 www.anthropic.com/news/claude-... Model card shows benchmarks are saturated for many tasks www-cdn.anthropic.com/0dd865075ad3...

Accurate plasmid reconstruction from metagenomics data using assembly–alignment graphs and contrastive learning www.nature.com/articles/s41... 🧬💻🧪 github.com/RasmussenLab...

This is great. The bits about Software Carpentry and their teaching philosophy in Stephen’s blog post really sticks with me. Some of those workshops were the start of my data science journey way back in the day.

I don't love this. PaperBanana: Automating Academic Illustration for AI Scientists arxiv.org/abs/2601.23265

I've always been a fan of what the Allen Institute is doing. New in Nature: OpenScholar, an 8B RAG model for scientific literature, outperforms GPT-4o by 6% on correctness. Experts preferred its answers over human-written ones 51%-70 of the time. www.nature.com/articles/s41... 1/3 🧵

Big Data Science in Medicine. Who's going to tell them...

Over the moon to announce that @vickiboykis.com will be the keynote speaker for the Applied Machine Learning Conference here in Charlottesville this spring! news.cvilleds.org/p/2026-amlc-... #databs

OpenAI Codex App https://blog.stephenturner.us/p/openai-codex-app-qqman #rstats #datascience

Ten simple rules for teaching data science arxiv.org/abs/2602.02874 1: Teach data science by doing data analysis 2: Use participatory live coding 3: Give tons of practice and timely feedback 4: Use tractable or toy data examples 5: Use real and rich, but accessible data sets

Lede buried. When humans prefer AI written responses over human experts to difficult scientific questions…

"Ten simple rules for teaching data science": arxiv.org/abs/2602.02874 A new preprint by @minecr.bsky.social and myself. We'd love any feedback!

Blog post on the new OpenAI Codex app doi.org/10.59350/b6e... #Rstats

Biological risk and the AI capability shift doi.org/10.59350/b38.... Notes from the 2026 Int'l AI Safety Report. 100 experts, 200 pages, 1400 references. AI co-scientist, crossing the expert threshold, industry response, dual-use, governance. 660 words / 3 min. #biosecurity

PlotGDP: an AI Agent for Bioinformatics Plotting www.biorxiv.org/content/10.6... 🧬💻🧪 (careful what you upload here, nslookup the domain)

The genetic foundations of convergent traits www.nature.com/articles/s41... (free: rdcu.be/e2bBO) 🧬💻🧪

Using semantic search to find publicly available gene-expression datasets academic.oup.com/bioinformati... 🧬💻🧪 github.com/srp33/GEOfin...

Bridging the Digital to Physical Divide: Evaluating LLM Agents on Benchtop DNA Acquisition www.rand.org/pubs/researc... #biosecurity 1/ 🧵

Follow-up looking at the new OpenAI Codex App: vibe-updating my old qqman #Rstats package to ggplot2 with plan+execute. blog.stephenturner.us/p/openai-cod...

R Works: Dec 2025 Top 40 new #Rstats pkgs: Agriculture, AI, med stats, audio, pharma, causal inference, computational models, ecology, statistics, econometrics, epidemiology, genetics, time series, utilities, genomics, machine learning, visualization, mediation analysis. rworks.dev/posts/decemb...

Interesting on the #consgen front: in partnership with the UAE 🇦🇪 Colossal Biosciences is building the BioVault, a cryobank to house genetic samples from 10k species.

Preprint from @lpachter.bsky.social lab argues journals should treat results as machine-readable objects, not just PDFs. Narratives stay for humans, but claims, evidence, and evaluations should ship in structured form so machines can index and verify. www.biorxiv.org/content/10.6... 1/ 🧵

Positron Notebook Editor for Jupyter Notebooks posit.co/blog/announc... Environment management, variables pane, data explorer, AI assistant all built in.

LLMs are good at idea quality, but weak on idea variance. GPT-4's brainstorm pools are less diverse than humans, unless you prompt for diversity explicitly. CoT prompting gets closest to “team human.” papers.ssrn.com/sol3/papers....

dplyr 1.2.0 release notes: replace_values() , recode_values() , replace_when() , filter_out() cran.r-project.org/web/packages... #Rstats

UNESCO Recommendation on Open Science (2023) www.unesco.org/en/open-scie...

Finally got around to reading Claude's Constitution. It's long, but fascinating. I mean it. www.anthropic.com/constitution 1/ 🧵

Francis Collins is giving a lecture in the Rotunda room here at UVA later this month

New blog post: Biological risk and the AI capability shift. Notes from the 2026 International AI Safety Report. 100 experts, 200 pages, 1400 references. AI co-scientist, crossing the expert threshold, industry response, dual-use, governance. blog.stephenturner.us/p/internatio... #biosecurity

Amazing how snow is cleared from parking lots and roads within hours, but bike lanes remain inaccessible and bike racks uncleared for weeks

OpenAI launched a dedicated Codex desktop app positioned as a focused UI for running multiple agents in parallel, keeping changes isolated via built-in worktrees, and extending behavior with skills and scheduled automations openai.com/codex/

For the evening crowd. New blog post: What's "tacit knowledge" in #biosecurity, and is it really a barrier? A look at the Dec 2025 RAND paper "Contemporary Foundation AI Models Increase Biological Weapons Risk." doi.org/10.59350/mk9... 🧬💻🧪🛡️🦠

Customizing CRISPR–Cas PAM specificity with protein language models Paper: www.nature.com/articles/s41... (free: rdcu.be/e1YwG) Code: github.com/Profluent-AI... Web: protein2pam.profluent.bio

AGI gets misdefined as perfection, autonomy, or superintelligence. But human general intelligence is messy, uneven, and error-prone too. If we credit humans with general intelligence from behavior, is it correct to apply the same standard to frontier LLMs? www.nature.com/articles/d41...