Steven Robbins

Totally fair, and makes perfect sense.

That was my guess as well. If your department had money, you may have a buffer. If not, you’re moving toward make or break. Ouch. I feel for all these people.

Ouch. I’m so sorry to hear that, Dan. That’s lot of stress, beyond the normal amounts of grant stress. Wish I had something more productive to add than I feel for you.

Thanks Jeremy. That’s what I’d have guessed, just don’t have a way to quantify the % of people one step away from crashing out, though one can extrapolate from these graphs of awards.

Thanks Dave. Like you say, academia has always been hoping to get the next grant and lots don’t. But it’s getting worse, is how I read you? Can you quantify how much worse?

And is what you’re saying that people are getting fewer grants and there admitting fewer grad students?

AI Keeps Failing at Microbiome Prediction Why simple models keep winning, where deep learning still shines, and where the field is headed

AI has huge promise for genomics -- but it has consistently failed at microbiome-based prediction.

My new post on why simple models keep winning, where deep learning actually earns its place, and where the field is headed

blekhman.substack.com/p/ai-keeps-f...

Genuine question: given reports of [waves hands around] everything, us US academia falling apart or are people finding ways of making things work? It seems like funding is evaporating, but I don’t know if that’s accurate or the social media anxiety machine hyperbolizing to some extent.

Will say, AI coding agents are useful when you’re new to a task and know enough to ask the right questions, but at a time cost of vibe coding slop, vibe coding the fix in a circle a number of times. Then you can eval the product and learn.

For most tasks, the slop phase kills productivity in net.

Oh dear me. Having found nothing, they pin hope on a result that is significant “before adjusting for multiple comparisons”. People: this adjustment is not optional. It’s clear nobody will ever give up on this quest.

The political effects of X's feed algorithm https://doi.org/10.1038/s41586-026-10098-2 Received: 16 December 2024 Accepted: 4 January 2026 Published online: 18 February 2026 Open access • Check for updates Germain Gauthier,5, Roland Hodler?5, Philine Widmer35 & Ekaterina Zhuravskaya3,4,5 m Feed algorithms are widely suspected to influence political attitudes. However, previous evidence from switching off the algorithm on Meta platforms found no political effects'. Here we present results from a 2023 field experiment on Elon Musk's platform X shedding light on this puzzle. We assigned active US-based users randomly to either an algorithmic or a chronological feed for 7 weeks, measuring political attitudes and online behaviour. Switching from a chronological to an algorithmic feed increased engagement and shifted political opinion towards more conservative positions, particularly regarding policy priorities, perceptions of criminal investigations into Donald Trump and views on the war in Ukraine. In contrast, switching from the algorithmic to the chronological feed had no comparable effects. Neither switching the algorithm on nor switching it off significantly affected affective polarization or self-reported partisanship. To investigate the mechanism, we analysed users' feed content and behaviour. We found that the algorithm promotes conservative content and demotes posts by traditional media. Exposure to algorithmic content leads users to follow conservative political activist accounts, which they continue to follow even after switching off the algorithm, helping explain the asymmetry in effects. These results suggest that initial exposure to X's algorithm has persistent effects on users' current political attitudes and account-following behaviour, even in the absence of a detectable effect on partisanship.

A new paper shows that less than 2 months of exposure to Twitter’s algorithmic feed significantly shifts people’s political views to the right.

Moving from chronological feed to the algorithmic feed also increases engagement.

This is one of the most concerning papers I’ve read in awhile.

"Seasonal enhancement of the viral shunt catalyzes a subsurface oxygen maximum in the Sargasso Sea" https://www.nature.com/articles/s41467-025-67002-1

Viruses transform microbes, populations, & ecosystems.

In new work jointly led w/@tn-marine-micro.bsky.social + more, we find a link between enhanced viral infection and productivity 50 meters below the surface in the otherwise nutrient limited Sargasso Sea.

a 🧵

www.nature.com/articles/s41...

The giant viruses surprised us at almost every turn of this project, but ultimately led us down a very rewarding path. Happy to share this work is now available online 🧪

Now published in Nature Biotechnology:
go.nature.com/44P7nSm
If you missed it, the TL;DR is in my April thread below

Really crazy that these massive genes are 6% of the 1.5 Mbp genome in this bee gut symbiont:

Also, have said 1000x, Medicare For All is phenomenally cheaper because you have the collective bargaining power of 340M people. No company can afford to lose the US market. And you don’t need to pay 1000s of middle men and CEOs. Thats why Australia pays like 100x less for care than the US.

Medicare for All would save $450B a year.

Every dollar spent on food stamps generates $1.50-$1.80 in economic activity.

Each dollar going to low-wage workers adds $1.20 to the economy overall.

It’s not about what this country can or can’t afford.

It’s about priorities.

Resolving plasmid-encoded carbapenem resistance dynamics and reservoirs in a hospital setting through nanopore sequencing The growing resistance of Enterobacterales to last-resort antibiotics such as carbapenems puts a significant burden on healthcare systems, also due to plasmids driving a rapid spread of carbapenem res...

Long-reads exposed plasmid-driven carbapenem resistance transmission missed by routine diagnostics

📌Resolving plasmid-encoded carbapenem resistance dynamics and reservoirs in a hospital setting through nanopore sequencing

www.doi.org/10.1099/mgen.0.001644

🖥️🧬💻
#AcademicSky
#MicroSky
#IDSky
🧪🧫🦠

Hugging Face – The AI community building the future. We’re on a journey to advance and democratize artificial intelligence through open source and open science.

Super excited to announce the release of gene and intergenic region annotation from the largest bacterial genome and MAG datasets available, including AllTheBacteria, GTDB, SPIRE, HRGM, mOTUs and MGnify - dereplicated and available from HuggingFace huggingface.co/AllTheBacteria

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

This piece has *really* *really* *really* good charts.
h/t @profraywills.futuresmart.com.au
#energysky

More renewables = lower electricity prices.

Also: Nyuck Nyuck! youtu.be/FbjiKIKzY3c

But, seriously, folks: Gas is flaming out.
#energysky

The rise of battery storage, and why the grid is rapidly passing gas!
reneweconomy.com.au/the-rise-of-...

EIA: 99%+ of new US capacity in 2026 will be solar, wind + storage Solar, wind, and batteries are set to supply virtually all net new US generating capacity in 2026, according to the latest EIA data.

EIA: 99%+ of new US capacity in 2026 will be solar, wind + storage
🔌💡
electrek.co/2026/01/28/e...

Yyyup! 😅

AI Agent Skills Have a Massive Security Problem

Someone just proved you can trick developers into running malicious code.

16 people from 7 countries fell for it in 8 hours.

Here's what happened:

This is the silly thing, I already have the R226 SingleM metapackage downloaded. I was just thinking originally in terms of "classifying whole reads" and then realized I didn't really need to, so SingleM is the answer.

Even so, hopefully people got a useful educational thread out of my brain fog.

Thanks Claudio. I tend to assume that if I have questions, others probs have the same questions. Serves the dual purpose of answering my own questions while making a form of educational content for others. Luckily we're a helpful community.

Stay tuned, I will definitely not know more useful things!

How do you even come up with these things, man? 😂

Seems well worth investigating. Have seen it and thought interesting, but hasn’t risen to the top of my mind to test yet.

Thanks Titus! In retrospect, we don’t really need to classify the reads, I don’t think, just get a taxonomy breakdown at the read level, so SingleM seems like the way to go.