High Performance Computing, Kubernetes, Python, Cosmology @Planck, now data pipelines for @simonsobs, Cloud @jetstream_cloud. Staff @SDSC_UCSD & consultant.
New guide: SDSC's cyberinfrastructure for AI research — Expanse, Voyager, Cosmos, NRP, and NDP. Step-by-step access for each platform. `https://t.co/7E8tgiqffo` #hpc#ai#sdsc#cyberinfrastructure
Running a Python workshop? You might not need JupyterHub 🐍
JupyterLite runs Python in the browser — no server, no setup. NumPy, SciPy, Pandas, scikit-learn all work via WebAssembly 🧪
https://t.co/rre8USN8ZV
https://t.co/j8WGaZFiRd #Python#Jupyter
New tutorial: Deploy JupyterHub on OpenStack Magnum with OpenTofu.
Kubernetes + HTTPS + custom subdomain — fully automated.
https://t.co/3Sy1SvrZs3
#Kubernetes#OpenStack#JupyterHub
Applications close TOMORROW for SDSC HPC and Data Science Summer Institute! 5 days hands-on: Deep Learning, GPU, MPI/OpenMP. Aug 3-7 San Diego. Limited to 45. Full agenda: https://t.co/ehwVfJithL Register: https://t.co/2VExqULjg5 #HPC#DataScience#AI
Just published: a notebook comparing SPT-3G D1 CMB power spectra with Planck 2018. Shows where primary acoustic peaks give way to secondary anisotropies at high multipoles. Created with a simple prompt to Sonnet 4.6.
2026 New Horizons in Physics Prize!
Congratulations Professor Colin Hill !
"For advances in cosmic microwave background and supernovae cosmology"
"...pushed analyses of CMB data beyond previous limits, producing the most precise tests to date..."
https://t.co/8nN3jZi0cp
Pretty disappointed with Codex for CLI use lately. The 5-hour window now gets burned in about 1 hour of real work. A month or two ago, I almost never hit it.
I’ve been using Gemini CLI more, and Gemini 3 Flash has been impressively good with fairly complex problems.
Applications are open for #FlatironCCA's 2026 CMB Summer School, where participants will learn core concepts and practical tools for CMB data analysis and simulations. Deadline is April 4: https://t.co/hjPZ7XOR5J #astrophysics#science
I wrote a tutorial on running your own AI chat on Jetstream2.
Jetstream2 is a research cloud, and ACCESS is the program that helps U.S. researchers use it. I show how to start a GPU VM, run `llama.cpp`, and chat through a web UI.
https://t.co/xHa9JsrDPw
Confluence is still a miserable place to keep useful content, so here’s a better escape route: export pages and attachments straight to Markdown with `confluence-cli` via the API.
https://t.co/Q9JnBeejo6
I built sphinxnotebookgist, a Sphinx extension for keeping Jupyter notebooks clean in git while rendering executed outputs in docs. It verifies source cells before using a remote executed notebook. https://t.co/hTAcuXexwg
Simplify the maintenance of a Docker Image for #Jupyterhub using repo2docker and #Github actions:
https://t.co/PWCYBpUczO
Notice that inside Github action I am using Kind to actually deploy Zero to JupyterHub in #Kubernetes and test the image inside it.
As an alternative to deploy `nbgrader` on Zero-to-JupyterHub using `ngshare`, on Jetstream 2 we can also rely on the shared filesystem service provided by Openstack Manila:
https://t.co/zDVpuH8XYq
Gemini 3.1 Pro is ramping up in Google Antigravity, starting with Google AI Ultra and Pro users! This model brings our most advanced reasoning and long horizon planning to your development workflow.
New tutorial: deploy nbgrader on Jetstream with ngshare (#Kubernetes). Covers ngshare install, #JupyterHub integration, student workflow, and autograding with expected outputs.
https://t.co/E5SXlF5VA1