Suraj Sahu
LightNews LightNews
banner
surajinacademia.bsky.social
Suraj Sahu
@surajinacademia.bsky.social
97 followers 500 following 6 posts
PhD Candidate at University of California Merced, #MechanoBiology #MulticellularNetworks
Posts Replies Media Videos
Reposted by Suraj Sahu
science.org
"In the end, I have come to realize that being authentic at work is not a weakness, but rather a strength." #ScienceWorkingLife https://scim.ag/49B7hRv
An illustration of a woman wearing an axolotl costume in front of an audience, with text: Embracing my silly side makes me a better scientist. I wish I’d done it sooner
November 21, 2025 at 8:59 PM
Reposted by Suraj Sahu
scientificdiscovery.dev
I'm surprised I only came across it now, but this review on improving communication in data visualization is excellent.
journals.sagepub.com/doi/10.1177/...
The Science of Visual Data Communication: What Works - Steven L. Franconeri, Lace M. Padilla, Priti Shah, Jeffrey M. Zacks, Jessica Hullman, 2021
Effectively designed data visualizations allow viewers to use their powerful visual systems to understand patterns in data across science, education, health, an...
journals.sagepub.com
November 19, 2025 at 11:48 AM
Reposted by Suraj Sahu
stephenturner.us
Does AI-Assisted Coding Deliver? A Difference-in-Differences Study of Cursor's Impact on Software Projects arxiv.org/abs/2511.04427
Does AI-Assisted Coding Deliver? A Difference-in-Differences Study of Cursor's Impact on Software Projects
Large language models (LLMs) have demonstrated the promise to revolutionize the field of software engineering. Among other things, LLM agents are rapidly gaining momentum in their application to…
arxiv.org
November 17, 2025 at 7:01 PM
Reposted by Suraj Sahu
hansonmark.bsky.social
We wrote the Strain on scientific publishing to highlight the problems of time & trust. With a fantastic group of co-authors, we present The Drain of Scientific Publishing:

a 🧵 1/n

Drain: arxiv.org/abs/2511.04820
Strain: direct.mit.edu/qss/article/...
Oligopoly: direct.mit.edu/qss/article/...
A table showing profit margins of major publishers. A snippet of text related to this table is below. 1. The four-fold drain 1.1 Money Currently, academic publishing is dominated by profit-oriented, multinational companies for whom scientific knowledge is a commodity to be sold back to the academic community who created it. The dominant four are Elsevier, Springer Nature, Wiley and Taylor & Francis, which collectively generated over US$7.1 billion in revenue from journal publishing in 2024 alone, and over US$12 billion in profits between 2019 and 2024 (Table 1A). Their profit margins have always been over 30% in the last five years, and for the largest publisher (Elsevier) always over 37%. Against many comparators, across many sectors, scientific publishing is one of the most consistently profitable industries (Table S1). These financial arrangements make a substantial difference to science budgets. In 2024, 46% of Elsevier revenues and 53% of Taylor & Francis revenues were generated in North America, meaning that North American researchers were charged over US$2.27 billion by just two for-profit publishers. The Canadian research councils and the US National Science Foundation were allocated US$9.3 billion in that year. A figure detailing the drain on researcher time. 1. The four-fold drain 1.2 Time The number of papers published each year is growing faster than the scientific workforce, with the number of papers per researcher almost doubling between 1996 and 2022 (Figure 1A). This reflects the fact that publishers’ commercial desire to publish (sell) more material has aligned well with the competitive prestige culture in which publications help secure jobs, grants, promotions, and awards. To the extent that this growth is driven by a pressure for profit, rather than scholarly imperatives, it distorts the way researchers spend their time. The publishing system depends on unpaid reviewer labour, estimated to be over 130 million unpaid hours annually in 2020 alone (9). Researchers have complained about the demands of peer-review for decades, but the scale of the problem is now worse, with editors reporting widespread difficulties recruiting reviewers. The growth in publications involves not only the authors’ time, but that of academic editors and reviewers who are dealing with so many review demands. Even more seriously, the imperative to produce ever more articles reshapes the nature of scientific inquiry. Evidence across multiple fields shows that more papers result in ‘ossification’, not new ideas (10). It may seem paradoxical that more papers can slow progress until one considers how it affects researchers’ time. While rewards remain tied to volume, prestige, and impact of publications, researchers will be nudged away from riskier, local, interdisciplinary, and long-term work. The result is a treadmill of constant activity with limited progress whereas core scholarly practices – such as reading, reflecting and engaging with others’ contributions – is de-prioritized. What looks like productivity often masks intellectual exhaustion built on a demoralizing, narrowing scientific vision. A table of profit margins across industries. The section of text related to this table is below: 1. The four-fold drain 1.1 Money Currently, academic publishing is dominated by profit-oriented, multinational companies for whom scientific knowledge is a commodity to be sold back to the academic community who created it. The dominant four are Elsevier, Springer Nature, Wiley and Taylor & Francis, which collectively generated over US$7.1 billion in revenue from journal publishing in 2024 alone, and over US$12 billion in profits between 2019 and 2024 (Table 1A). Their profit margins have always been over 30% in the last five years, and for the largest publisher (Elsevier) always over 37%. Against many comparators, across many sectors, scientific publishing is one of the most consistently profitable industries (Table S1). These financial arrangements make a substantial difference to science budgets. In 2024, 46% of Elsevier revenues and 53% of Taylor & Francis revenues were generated in North America, meaning that North American researchers were charged over US$2.27 billion by just two for-profit publishers. The Canadian research councils and the US National Science Foundation were allocated US$9.3 billion in that year. The costs of inaction are plain: wasted public funds, lost researcher time, compromised scientific integrity and eroded public trust. Today, the system rewards commercial publishers first, and science second. Without bold action from the funders we risk continuing to pour resources into a system that prioritizes profit over the advancement of scientific knowledge.
November 11, 2025 at 11:52 AM
Reposted by Suraj Sahu
jcb.org
New perspective by @maikbischoff.bsky.social & @mayorlab.bsky.social about how mesenchymal cells use contact-dependent rules to generate swarm-like behaviors, patterns & organ forms. rupress.org/jcb/article/...

#CellBio #CellMigration #Morphogenesis #Development #Science
September 23, 2025 at 6:45 PM
Reposted by Suraj Sahu
scientificdiscovery.dev
I don't think the average person is going to learn much by accessing a paywalled scientific paper.

But the current system keeps out journalists, science communicators, policy researchers, and fact checkers from reading into a topic as well.
scientificdiscovery.dev Saloni @scientificdiscovery.dev · Sep 28
In general I think it's hard to combat scientific misinformation when some of the best research is locked behind an academic paywall, while lots of nonsense gets published free for everyone to read in predatory journals.
September 29, 2025 at 6:17 PM
Reposted by Suraj Sahu
mpipks.bsky.social
Bacteria are surrounded by water! New work @natphys.nature.com by @ricardalert.bsky.social @mpipks.bsky.social & collaborators @ Princeton University shows that water capillary forces organise bacterial colonies into gas, nematic streams, and droplet states.

nature.com/articles/s41567-025-02965-y
September 16, 2025 at 6:03 PM
Reposted by Suraj Sahu
amcell.bsky.social
2020. Simple rules for concise scientific writing. Hotaling - aslopubs.onlinelibrary.wiley.com/doi/10.1002/...
Simple rules for concise scientific writing
Click on the article title to read more.
aslopubs.onlinelibrary.wiley.com
September 16, 2025 at 5:29 AM
Reposted by Suraj Sahu
lmrocha.bsky.social
1/2 Very interesting discussion about pace of science (thanks @cyrilpedia.bsky.social!), but one cannot really separate the pace of innovation or disruption from the pervading economy. Given the present technofeudalism, new ideas are harder to translate to society.
www.aisnakeoil.com/p/could-ai-s...
Could AI slow science?
Confronting the production-progress paradox
www.aisnakeoil.com
September 3, 2025 at 5:30 PM
Reposted by Suraj Sahu
ketanjoshi.co
There's a big data centre planned in the bushland next to where I grew up in Sydney. Not small either: a 170 megawatt facility that'll require the demolition of 300 mature trees, many hundreds of years old

www.instagram.com/reel/DNxM_K0...
September 3, 2025 at 10:38 AM
Reposted by Suraj Sahu
ketanjoshi.co
Meta's probably going to release their new energy / env disclosures soon.

(a) In 2024, they consumed 15 TWh
(b) Between 2023 to 2024, they added 3.9 TWh (double the addition of the last period)

Drop your predictions below for (a) absolute consumption and (b) change from 2024, below :)
a chart showing absolute energy consumption a chart showing cahnge in energy consumption
September 1, 2025 at 8:23 AM
Reposted by Suraj Sahu
emollick.bsky.social
We now have some data on the environmental impact per AI prompt:

Gemini: 0.00024 kWh & 0.26 mL water

ChatGPT: 0.0003 kWh & 0.38 mL

...the same energy as one Google search in 2008 & 6 drops of water. Seems to be improving, too: Google reports a 33x drop in energy use per prompt in a year.

But...
August 22, 2025 at 4:40 AM
Reposted by Suraj Sahu
emollick.bsky.social
It appears that the marginal energy used by a standard prompt from a modern LLM is relatively established at this point, roughly 0.0003 kWh (8-10 seconds of streaming Netflix)

Water is more complicated (.25mL to 5mL+), depending on definitions. Training resources are much less clear at this point.
August 25, 2025 at 2:42 PM
Reposted by Suraj Sahu
emollick.bsky.social
There has not been a lot of evidence so far that AI is actually impacting jobs.

This paper finds early signs that this may be starting, showing a decrease in entry-level job openings relative to experienced ones & in the jobs where AI is mostly automates, not augments, work (like coding).
August 26, 2025 at 3:03 PM
Reposted by Suraj Sahu
emollick.bsky.social
We really have not made a lot of progress on explaining the core deep mystery of LLMs:

How does a model using matrix multiplication to predict the next word manage to simulate human thought well enough to do all the very human-like things it does? And what does that mean about us and our thinking?
August 28, 2025 at 4:20 AM
Reposted by Suraj Sahu
cyrilpedia.bsky.social
'Matter fact, why even go to college or live in a dorm when you could stay in your mom’s basement for four years, and have ChatGPT complete and upload all of your assignments? It would be so much more efficient than what we’re doing now.'
www.mcsweeneys.net/articles/how...
How I Learned to Stop Teaching and Love AI
Like a fool, I used to resist, but no more. In the past, I railed against the rise of AI. I preached small-minded sermons to students who had to si...
www.mcsweeneys.net
August 27, 2025 at 1:46 AM
Reposted by Suraj Sahu
newyorker.com
An M.I.T. study found that 95% of companies that had invested in A.I. tools were seeing zero return. It jibes with the emerging idea that generative A.I., “in its current incarnation, simply isn’t all it’s been cracked up to be,” johncassidysays.bsky.social writes.
The A.I.-Profits Drought and the Lessons of History
Like the steam engine, electricity, and computers, generative artificial intelligence could take longer than expected to transform the economy.
www.newyorker.com
August 25, 2025 at 2:43 PM
Reposted by Suraj Sahu
ketanjoshi.co
"The most accurate model was the reasoning-enabled Cogito model with 70 billion parameters, reaching 84.9% accuracy. The model produced three times more CO2 emissions than similar sized models that generated concise answers"

"Reasoning" models are just brute-forcing info like Bitcoin mining
Hidden Cost of Smart AI: 50× More CO₂ for a Single Question
Some AI models burn through 50× more carbon to think through a question, yet don’t always offer better answers.
scitechdaily.com
August 21, 2025 at 7:40 AM
Reposted by Suraj Sahu
insilicoplants.bsky.social
aibio-uk.bsky.social AI in the Biosciences (AIBIO UK) BBSRC Network @aibio-uk.bsky.social · Aug 7
🚨Opportunities from AIBIO-UK!

🔬 Apply for our AI & Biosciences Sandpit (Nov 18–19) – up to £30k funding!
📝Apply: forms.office.com/e/S8eqNF4X31

🎓 Express interest in upcoming training scholarships in AI: forms.office.com/e/h5SsZqwy5v

#AI #Biosciences #ResearchFunding #Training #AIBIOUK
Microsoft Forms
forms.office.com
August 20, 2025 at 9:33 PM
Reposted by Suraj Sahu
sciencecolin.bsky.social
For those of you looking for science jobs, I've been collecting links to job boards and fellowships as I've come across them and figured others might find them useful. Feel free to reach out if you know of something that would be good to add to the list!

docs.google.com/spreadsheets...
Post Grad Job Search
docs.google.com
August 19, 2025 at 3:16 PM
surajinacademia.bsky.social
Would be great if the graduate members would be open to host members.
biophysicalsoc.bsky.social Biophysical Society @biophysicalsoc.bsky.social · Aug 18
Want to attend #bps2026 in San Francisco but need help with funding to get there? BPS offers travel awards for students and scientists at every career level.

Deadline to apply is October 3
(and you have to submit your abstract by October 1!)
BPS2026 Travel Awards | Biophysical Society
Travel Award FAQs
buff.ly
August 19, 2025 at 12:42 AM
Reposted by Suraj Sahu
acritschristoph.bsky.social
Yeah, LLMs for coding are like copy-and-paste-from-stack-overflow on steroids. It's a great way to quickly get started, learn new things, and do small tasks, but too much without thinking will create a mess.

That said in a recent ~1000 line project, I'd guess it made me ~25% faster
wcratcliff.bsky.social Will Ratcliff @wcratcliff.bsky.social · Aug 12
Not sure how many scientists here have tried Claude Code or similar command line coding assistants. I had a complicated family property tax problem that was best solved by a brute force Monte Carlo simulation approach, so I spent a few days coding up and analyzing a model with Claude Code.
August 12, 2025 at 2:32 PM
Reposted by Suraj Sahu
wcratcliff.bsky.social
Here is one more crazy statistic: in the course of coding this model, I used >200 *million* tokens of compute (last column)! I don't know what the kWh equivalent of this is...but this is probably moving the needle significantly on my net CO2 emissions.
August 12, 2025 at 2:36 PM
Reposted by Suraj Sahu
clauswilke.com
Yup.
clauswilke.substack.com/p/phd-level-...
PhD-level intelligence or the graduate student from hell
AI companies love to tout that their models are approaching—or have reached—PhD-level intelligence. This is blatant nonsensical marketing geared towards an audience that deeply misunderstands what a P...
clauswilke.substack.com
August 12, 2025 at 2:18 PM
Reposted by Suraj Sahu
wcratcliff.bsky.social
Not sure how many scientists here have tried Claude Code or similar command line coding assistants. I had a complicated family property tax problem that was best solved by a brute force Monte Carlo simulation approach, so I spent a few days coding up and analyzing a model with Claude Code.
August 12, 2025 at 12:47 PM