Julianne Dalcanton

Julianne Dalcanton

@dalcantonjd.bsky.social

Astrophysicist at the Center for Computational Astrophysics & U. Washington.

2.1K followers367 following532 posts

Though I think not unfamiliar trouble if viewed through a labor/class lens. Academics, and especially STEM-flavored academics, often don’t see themselves as “labor”, and part of our discomfort is that the same forces are coming for us that have come for other “skilled tradespeople”.

John Stott@jpsastro.bsky.social

Indeed, I think science is in trouble

Like many scientists, I’ve been mulling over the impact of LLMs on our fields. I’ve not yet seen it articulated, but I think there is a largely unspoken “invisible labor underclass” framework at play.

View full thread

I am way out of my expertise zone with this framing, but people who think about systems of class and economies and labor have plenty of more rigorous and informed things to say.

But this is what I’m mulling at the moment. I just don’t see how this winds up being fiscally or socially stable, and “the LLMs will do it” does not seem like a satisfying answer.

Like many scientists, I’ve been mulling over the impact of LLMs on our fields. I’ve not yet seen it articulated, but I think there is a largely unspoken “invisible labor underclass” framework at play.

View full thread

And the zillions of pull requests that the Agentically Enabled will be generating for key open source projects. Who exactly is supposed to be verifying and implementing them? Who exactly is supposed to be wading through to find which ones are misguided?

Somehow all these models of our future assume that there is some OTHER group of very highly trained, insightful individuals who function as an underclass that will somehow be satisfied with endlessly reviewing the products of the Real Scientists to make all this work out.

Like many scientists, I’ve been mulling over the impact of LLMs on our fields. I’ve not yet seen it articulated, but I think there is a largely unspoken “invisible labor underclass” framework at play.

The Real Scientists of the future are envisioned to be those pushing agentic LLMs to the forefront of rapid knowledge discovery. And then that work is supposed to be tediously verified and translated by who exactly?

The hundreds of papers collaborated into existence with LLMs? The Real Scientists will be leading these, and the magic of peer review will separate the wheat from the chaff. Who, exactly, is supposed to be executing this tedious search for the wheat?

Julianne Dalcanton reposted

Here's a chart showing who funds basic research in the US. It's easy to tell that private funding cannot make up for significant cuts to government funding.

We have had constitutional amendments before. We were extremely close to passing the Equal Rights Amendment. Is there any concerted effort around a Right to Privacy amendment? Surveillance tech (Flock, Palentir, etc), AI scraping of personal communications, bathroom policing are begging for it.

Trump swag has been decreasing at NYC tourist shops, but what I find most baffling is their failure to capitalize on Mamdani and the graphical branding of his campaign.

Aaron Huertas@aaronhuertas.bsky.social

Public opinion update: No Trump shirts for sale at the boardwalk tshirt shops.

Just passed a Certain Kind of Man filming his righteous take that the aristocrats in the French Revolution were wrongly targeted, since they were the real wealth and job creators. I am not making this up. It came out of his actual mouth as he huffily tossed his expensive hair.

Julianne Dalcanton reposted

Our Undergraduate and Graduate Research Showcase is back! Cash prizes to be won: 1st place - $1000.00 2nd place - $750.00 3rd place - $500.00 Register to present your research using the links in the thread below! #BlackSpaceWeek2026 #BlackInAstro

I am not ready to face the level of cringe in this to dare watching it, but Frank is a lovely interviewer, and there is backstory about why I haven’t shut up about Arp systems for 30 years.

🔭 New AAS Journals video! Julianne Dalcanton (Flatiron Institute/University of Washington) discusses new Hubble imaging of 216 interacting galaxies from the Arp and Arp–Madore catalogs, revealing rich structures that will support future studies with JWST and ALMA. 📺 youtu.be/eewSudl4w34

Julianne Dalcanton reposted

You have more in common with the people of Iran than you do with anyone in this administration, any member of the technocracy, anyone who is profiting from this war, or any billionaire. When you see Iranians forming human chains outside key infrastructure sites, you should feel a sense of kinship.

Julianne Dalcanton reposted

My mother: You’re responsible with what you post online, right? Always respectful? Me, about to post my definitive ranking of "most cancellable impact basins:" Yeah, absolutely

#astromethods #astrocoding #astrocode ☄️🔭

Julianne Dalcanton@dalcantonjd.bsky.social

Please spread the word! In addition to becoming an important part of the MESA project, the person in this role would have opportunities to collaborate on scientific software across CCA, which has a vibrant, interactive community of scientists and code developers.

Please spread the word! In addition to becoming an important part of the MESA project, the person in this role would have opportunities to collaborate on scientific software across CCA, which has a vibrant, interactive community of scientists and code developers.

Flatiron Institute@flatironinstitute.org

We're seeking a research software engineer to develop, maintain, and contribute to community-oriented software instrument tool MESA (Modules for Experiments in Stellar Astrophysics). Learn more and apply: https://apply.interfolio.com/182988 #science #astrophysics #hiring

Universities that had IT infrastructure gave up much of it to switch to “free” tools from Google (Gmail, Drive, etc), who then changed from an unlimited plan to a throttled-by-data-size-unless-you-pay model after the transition to Google was baked in. The same thing will absolutely happen here.

What happens when researchers (especially those with data/code/text-heavy activities) have been conditioned to rely on hours a day interaction with advanced models have to deal with whatever monopoly companies demand they (or their institutions) pay? How exactly is that gonna work?