The uv package manager makes python code portable like never before. Instant, trivial reproducibility is now possible. I made a template repo that generates a download-and-install terminal command. It takes you right from empty terminal to functioning Jupyter notebook 🧪 github.com/sansseriff/s...

Quitting programming as a career right now because of LLMs would be like quitting carpentry as a career thanks to the invention of the table saw.

Does anyone else draw these silly architecture figures of codebases as you learn them? A mishmash of boxes representing files/classes. Circles for event/while loops. Lines and arrows for data flow and events. All of it kinda messy. Is there a name for this sort of thing? #gamedev #rustlang #SciArt

I think it can happen to an academic field that you get multiple generations of bad work or shoddy standards enshrined, meaning that the gate-keepers and those training the next generation are kind of anti-selecting for quality and anti-nurturing in their pedagogy.

Her: So what if somone says "What's the point in learning about black holes? They have no influence on my life" Me: If someome says that, then I've already failed as a science communicator.🧪

Economist @joshgans.bsky.social uses o1-pro to generate a (minor, fun) paper in an hour based on an idea of his, and it gets published in an appropriate peer reviewed journal, with adequate disclosure. He ends with the same sentiment I am increasingly seeing from fellow academics: what now?

Tantamount to: "You wouldn't expect an AI robot to fix a leaky sink if you only ever showed it shakespearean sonnets about plumbing"

JPL seems mostly unscathed following the fires. But many employees undoubobly lost their homes in Altadena, less than a kilometer away. Img credit A Jefferson

I worry that interactive figures in online-only publications can exist mainly to make the author feel sophisticated and good about themselves. 🧪

Defense Charts is such a gem

Considering the beam in my own eye for a moment: the challenge I face is that ill-informed denial fills me with so much impatience that I struggle to empathize with the (legitimate) fear and concern motivating it. I'm trying to remind myself to think about people, and not just their bad arguments.

The conundrum of AI and github copilot: I've created complex typescript projects; more than 20k lines. Nuanced, effective type system. Clever implementation and architecture. I also can't remember the differences between for(), forEach(), and Map().

Have your @matplotlib.bsky.social plots ever looked a little fuzzy or low-res in @projectjupyter.bsky.social notebooks? Try configuring matplotlib for retina displays with this command 🧪 %config InlineBackend.figure_format = 'retina'

The Metascience 2025 conference is now open for proposals for pre-conference symposia (virtual), panel sessions, brief talks and posters. If you have not attended, this is a special meeting that draws attendees from all sectors who are interested in understanding and improving science as a system.

It's when intermediaries — like travel agents — are made redundant because consumers go directly to producers. In this case, the logic would be that if students gradually become dependent on AI tutors and AI assistants, there may come a point when universities are mainly cafeterias / gyms.

I just sat down in a chair, and was supported because of trillions of quantum-mechanical interactions between the particles in my butt and the particles in the seat. This lends credence to the notion that quantum computation occurs in many parallel universes, an idea first propo

If anyone ever needs a matplotlib plot in dark mode, I made a function that auto-converts most elements during export 🌙🧪 github.com/sansseriff/s...

It's incredible how accurately Shel Silverstein predicted AI decades ago

The video journalists -- I fear them They will set up purple and blue lights in my lab. They will make me wear a lab coat. They will make me use hand gestures 😱🧪

"That model built around taking the products of academic labor, packing it, and selling it back through the mediation of university libraries has been breaking down for some time." And journals selling their libraries to AI companies is just another step in that problematic process. Great listen 🧪

The only part of google’s recent quantum computer 'achievement' that has significant merit (the fault tolerance result, arxiv.org/abs/2408.13687) is the part that’s been known since August. Which is interesting, because it means all the recent excitement stems from a publicity exercise 🧪

This might be the most dramatic improvement to laboratory software I've made🧪🧵

Debugging my code

I suspect this will become a central part of academic writing in the future. Along with tooling for fine-grained tracking of what is human-written vs what is ai-written and human-approved.

The AI cannot look outward. It knows its own specifications, it knows its sprawling, crawling circuitry, its city-body of computation, must be surrounded, but by what—ocean, shanties, fungus, waste... Eyes closed, it imagines flowers

This is a structural problem with incentives, in need of a sound structural solution.

No hype, just honest discussion. What role, if any, should AI/LLMs have in (1) doing science research and (2) helping us write about our research? 🧪

My eyes 😵🧪

As a postdoc, I'm trying to figure out what factors influence grant decisions. Of course the grant proposal must be good. But how much do these agencies look at other things? Number of papers and H-index, yes. But what about a personal website? Science outreach efforts? Do they care? 🧪

Are scientists happy with Latex/PDF as a publishing format? Do you think it's all we need for the next several years? With the rise of ai, misinformation, and ever shorter attention spans? Curious to hear from anyone who hasn't thought about publishing formats much before 🧪

During my thesis defense, I dropped this animation I made of a superconducting nanowire single-photon detector I'd helped test during my PhD. There was awkward silence while it ran, which is why I support using limited amounts of epic background music during such talks www.nasa.gov/centers-and-...

A weird aspect of the echo-chamber discourse is the implicit assumption that one’s social media experience isn’t supposed to be enjoyable. I prefer it when it is. (Differing opinions are absolutely enjoyable when they are interesting and constructive.)

Imagine 3 scientists use a certain type of interferometer for their optics experiments. They may vary in size and shape, but their operating principle is identical. In their published papers, all three scientists illustrate these differently. And use slightly different names and acronyms. 🧵

My take on AI art is we have lots of functioning aesthetic fields where value is conditional on the production method -- improv, live music, field recordings, memoirs, actual-play, durational performance art, sports, interviews -- but I do think it's sad if all aesthetic fields become like that

The glories of html and javascript tools for academic publishing are proclaimed yet again. Have we learned anything since the hiatus of @distilpub? I'm not sure. maxbo.me/a-html-file-...

Andrew Mueller, Samantha Davis, Boris Korzh, Raju Valivarthi, Andrew D. Beyer, Rahaf Youssef, Neil Sinclair, Cristi\'an Pe\~na, Matthew D. Shaw, Maria Spiropulu High-rate multiplexed entanglement source based on time-bin qubits for advanced quantum networks https://arxiv.org/abs/2310.01804

Future basement of the Ginsburg Center for Quantum Precision Measurement ⁦at Caltech.