Thomas Veith

Cancer biologist and computational modeler interested in tumor evolution, copy number alterations, and ecological dynamics in cancer.

[Jeffrey West @mathoncbro]

I’ve just had the occasion to read Dorothy Sayer’s altogether lovely essay on “The lost tools of learning,” which would do well to return the current conversation on educational priorities.

“For the sole true end of education is simply this: to teach men how to learn for themselves; and whatever instruction fails to do this is effort spent in vain.”

[Thomas Veith @thomasaveith]

Thrilled to share that I’ve received the 2027 DCEG Intramural Research Award! This support will help us investigate when and how LINE-1 retrotransposition occurs during the evolution of UGI cancers. Deeply grateful to my mentor, our collaborators, and the DCEG community.

[Thomas Veith @thomasaveith]

@BonerWizard I think a certain demographic would go crazy for Home Depot on a (river)boat

[Ken Klippenstein @kenklippenstein]

Asked about the Iran War's effect on defense stocks, Lockheed Martin CEO says: "Investors need to know and understand that this demand is not fleeting — it's permanent."

[Ihtesham Ali @ihteshamali]

Bloomberg just dropped the most disturbing investigation I have read this year, and every AI builder should read it and take it seriously.

On February 28, two Tomahawk missiles hit an elementary school in Minab, a small town in southern Iran. 156 people died. At least 123 were children. The Pentagon knew within hours that it had bombed a school.

The school had its own website. It was labeled on Google Maps. You could see the soccer pitch from a satellite view. You could see the brightly painted walls. Any 12-year-old with a browser could tell what the building was.

The AI system could not.

Inside US Central Command, personnel were using Palantir's Maven Smart System. It is an AI targeting platform running on a $1.3 billion Pentagon contract, and embedded inside it is Anthropic's Claude.

On day one of the war, more than 1,000 targets went through it in 24 hours. Work that used to take a team hours per target got compressed into minutes.

The satellite imagery Maven was pulling from was seven years old. The site was still tagged in the database as an Islamic Revolutionary Guard facility.

The system spat it out as a recommended day-one target. Humans signed off. The first missile flew. Then a second one landed minutes later to hit the people who came to help.

Investigators later found that some Central Command staff had assumed Maven would flag stale intelligence for them. It did not. Nobody had told it to.

Palantir told Bloomberg the government is responsible for the underlying data. The Pentagon's civilian harm mitigation team, the people whose actual job is to catch a target like this, had been cut back before the war started.

This is not AI going rogue. This is the opposite.

This is AI doing exactly what it was built to do at exactly the speed it was built to do it. And a school full of little girls got vaporized because the humans in the loop trusted the machine more than they trusted their own eyes.

I keep thinking about the version of this story that comes next. What happens when the system is faster. When the target list is not 1,000 in a day but 10,000. When "review" becomes a button the operator has three seconds to hit before the next target loads. When there is no human left slow enough to notice that a soccer pitch is not a helipad.

We are not talking about a spam filter mislabeling an email. We are talking about a system that turned a seven-year-old database entry into two cruise missiles landing on children in the middle of a morning class.

A UN-backed panel has now called it a potential war crime.

In terms of children killed, it is the deadliest American targeting error of the entire 21st century. And according to Bloomberg, some of Defense Secretary Hegseth's own advisors are worried that too much public scrutiny of Maven might slow down its adoption inside the Pentagon.

Read that last line again.

If you build this stuff, you own what it does. You do not get to hide behind "the government owns the data." The whole pitch of your product is that speed and integration save lives. When it takes them instead, that is also on you.

I am not against AI in defense. I am against AI in a hurry. Speed is the entire mechanism that made Minab possible. Every guardrail we already know how to build takes time, and time is exactly what the system was designed to eliminate.

So here is the question I have not been able to stop asking, and I genuinely want to know what you think.

If a piece of software helped kill 123 children in a single morning while behaving exactly as designed, what does "responsible AI" even mean anymore?

[Thomas Veith @thomasaveith]

@JSheltzer My committee required me to have two first-author publications in addition to my dissertation. I think some committee members would have been fine if one of those papers was a review, but not my PI.

[Jason Koebler @jason_koebler]

“X’s algorithm learns what you hate and shows you more of it, according to a new study just published in the Proceedings of the National Academy of Sciences”

https://t.co/9bFIySDpKf

[NIH @NIH]

Some people with lung cancer develop cachexia, a condition that causes severe weight loss, muscle loss, weakness, and fatigue.

NIH-funded researchers found that in mice, a high-fat diet strengthened nerve signals between certain lung tumors and the brain, making cachexia worse.

More research is needed to determine whether these findings apply to people, but the results may help guide future nutrition interventions for cancer care.

Learn more: https://t.co/Lx25g0EDey

[Dr. Catharine Young @DrCatharineY]

PSA: Cancer is not one disease. It is more than 200 distinct diseases, all with different causes, biology and mechanisms. Scientists who study cancer rarely talk about “curing cancer” because it’s nothing more than an empty slogan.

[John Gallagher @MereSophistry]

Not enough people cheer academia. It’s an awesome place where you have the freedom to explore ideas that are not market driven. Lots of exciting positive people. Students are enthusiastic most of the time. Admins are mostly trying their best. Sports provide good community.

[Ming "Tommy" Tang @tangming2005]

Whole-genome doubling drives immune evasion by silencing antigen presentation https://t.co/sXfVIvgxUj

[James Schnable @szintri]

Models trained using genetic markers + gene expression + phenotypes can be better predicting phenotype solely from genetic markers than models trained using only genetic + phenotype data from the start. Kontolati et al, out two days ago. https://t.co/g9MxsVn2pi

[Avi Roy @agingroy]

Some people are born with a faulty gene that protects them from disease. Drug companies find those people, figure out what the broken gene does, and build a drug that copies it. It is the most reliable way to invent a medicine, because the experiment already ran in humans for a lifetime.

1. PCSK9. In 2006, researchers found that people born with this gene disabled had 88% less coronary heart disease. That discovery led to the injectable cholesterol drugs Repatha and Praluent, and three weeks ago, the first PCSK9 pill.

2. ANGPTL3. Families in Italy with this gene broken had very low cholesterol and clean arteries. The copy is Evkeeza, approved in 2021 for severe inherited high cholesterol.

3. APOC3. Broken copies mean very low triglycerides and less heart disease. The copy, Tryngolza, was approved in December 2024.

4. CCR5. About 1% of Europeans carry two broken copies and are essentially resistant to HIV, because the virus uses that protein as its door into cells. The copy is the drug maraviroc, and the only people ever cured of HIV got transplants from those donors.

5. Lp(a). People born with low levels of this particle get less heart disease. Five drugs copying that state are in late trials; the first big readout is due any week now.

6. FNIP1. Found on Tuesday, in a million genomes. One broken copy, 60% lower odds of heart disease, diabetes, and fatty liver. The hunt for the copy has just started, with a catch: the pathway it works through, when fully broken, causes a kidney cancer.

One gene taught us to lower cholesterol. Another taught us to block HIV. The next lesson was published this week.

[Katie Galloway @GallowayLabMIT]

The biotech era debuted in the 1980s, and this is the type of lab the NIH thought a single R01 could support.

Using the NIH entry-level postdoctoral stipend as a consistent benchmark, a single R01 with $250,000 in annual direct costs could theoretically pay for about 15.6 postdoc-years in 1987, when the stipend was $15,996 per year.

How about today? At the FY2026 stipend of $63,480 per year, that same $250,000 would pay for only about 3.9 postdoc-years.

This means that while a single R01 could theoretically support the equivalent of about 3 entry-level postdocs over 5 years in 1987, it could support fewer than one (one for four years) today—a roughly 75% loss in personnel purchasing power.

And, of course, this assumes that every dollar goes directly to postdoc stipends. It does not include fringe benefits, PI salary, graduate students, technicians, supplies, equipment, animals, core-facility fees, travel, publication charges, or any of the other costs required to run a modern laboratory.

Same nominal R01. Roughly one-quarter of the scientific labor.

[nature @Nature]

Inspired by long-lived animals — and human centenarians — researchers are hunting for ways to enhance DNA repair and extend healthspan

https://t.co/SgzeEzc3aX

[nature @Nature]

Could 'biophotons' represent a new signalling mechanism in cells? Or be used to diagnose disease?

https://t.co/Kn4Y4JHHsf

[Jeffrey West @mathoncbro]

We are recruiting a @USouthFlorida undergrad for an internship in artificial intelligence applied to mathematical oncology in my group @MoffittNews.

Interested candidates please send me an email with the details in the link:
https://t.co/6TXFKoqT6b

[Daisy Ding @daisyding_]

To solve aging, we first need to measure it. Excited to share our study in @NatureMedicine! Different cell types age at different rates within our body. From a tube of blood, we track aging across 40+ cell types, from immune cells to neurons, revealing signatures that forecast disease risk and resilience. @wysscoray 🧵1/9

[Jack Ghannam @JackYGhannam]

Excited to share some of my PhD work out today in @NatureCancer! Integrating bulk and single-nucleus genomic approaches, we identify distinct determinants of survival and acquired resistance to ICB versus standard chemoradiation in glioblastoma. https://t.co/EqeJLUvbou

[Cancer Cell @Cancer_Cell]

When tissue tension creates a mutagenic niche https://t.co/vVRdgqbbHm