HNHacker News
TopNewBestAskShowJobs

frigidwalnut

78 karma · joined February 8, 2025

submissionscomments
frigidwalnut··on GLM-5.3 is now open-weight
Cool! I'm thinking about a local set up. What's your usual tokens/second rate?
frigidwalnut··on Small Models Have Arrived
Sounds interesting. Can you give more details on your workflow and what tasks you use it for?
frigidwalnut··on Now AI agents need what RSS does
What are your favorite feeds?
frigidwalnut··on WH proposes rules giving political appointees final approval on research grants
Sure that’s fine. They funded research that you think is wasteful and you would like to have tools to provide that feedback. The issue with this OMB change is that it is not that. The OMB change is a change that allows the administration to cancel any grant for any reason without answering to citizens about it.

BTW your “believe it or not” is quite condescending. Do you talk to people in real life like that?

frigidwalnut··on WH proposes rules giving political appointees final approval on research grants
I agree. This administration is ground zero for mismanagement of funds and outright corruption. Just look at the director of the FBI and former secretary of DHS. Both have used and continue to use tax payer money for personal use. It should make every tax payer livid.
frigidwalnut··on WH proposes rules giving political appointees final approval on research grants
Private citizens fund scientific research through their taxes. This has been the most practical way to fund science for decades.
frigidwalnut··on Project Hyperion: Interstellar ship design competition
The paper linked in that post proposes a bottleneck at 900Kya in the ancestors of all modern humans. There is a bottleneck associated with the migration out of Africa and the peopling of the world that many populations have, but not all. Based on genetic data the timing is between 100-50Kya, with a lot of the uncertainty coming from converting generation times to years (i.e. how many years on average between parents and offspring). This is a nice reference: https://sci-hub.se/https://www.nature.com/articles/nature213...
frigidwalnut··on Project Hyperion: Interstellar ship design competition
Follow up work has shown that result is a statistical artifact/does not replicate [1,2,3].

[1]https://academic.oup.com/genetics/article-abstract/229/1/iya...

[2]https://academic.oup.com/mbe/article/42/2/msaf041/8005733

[3]https://www.biorxiv.org/content/10.1101/2024.03.25.586640v1

frigidwalnut··on Out of Africa: celebrating 100 years of human-origins research
Thanks for that link. The Wikipedia article you referenced says "This view is now considered discredited by mainstream science."

There are still papers being written about how those analyses were flawed: https://www.pnas.org/doi/abs/10.1073/pnas.2319496121

frigidwalnut··on Trump administration to fire thousands at health agencies
Hello, I submitted this. Thanks for your concern. I've been lurking on Hacker News since ~2014 when I was in college. I made an account and started posting because people on the board have something of an interest in biology and genetics (for example this story on cancer detection from yesterday: https://news.ycombinator.com/item?id=43035147 ) and I wanted to contribute. I posted this because the NIH funds a lot of research related to that and this story reports on a massive change to the NIH, which seemed worthy of discussion.
frigidwalnut··on Out of Africa: celebrating 100 years of human-origins research
The data are public! Why don't you show it?
frigidwalnut··on National Institutes of Health cuts support to universities
Where will these people conduct their research?
frigidwalnut··on National Institutes of Health cuts support to universities
Why do you think that research universities should go bankrupt? What do you think should replace their function in society?
frigidwalnut··on National Institutes of Health cuts support to universities
This is a huge cut. Back of the envelope calculations for UCLA (a big research university with public finance information): $200M cut in operating revenue.

In 2023, UCLA had $270M in indirect costs [1] and they negotiated a rate of 57% with the NIH [2]. So, they had about $473M in direct costs. The new rate would be 15%, which is ~$71M. $270M-$71M = $200M.

[1] Page 24: https://ucla.app.box.com/v/acct-pdf-AFR-22-23 [2] https://ocga.research.ucla.edu/facilities-and-administrative...

(Reposting from a less popular posting)

frigidwalnut··on NIH cuts indirect cost rates to 15% for new and existing grants
This is a huge cut. Back of the envelope calculations for UCLA (a big research university with public finance information): $200M cut in operating revenue.

In 2023, UCLA had $270M in indirect costs [1] and they negotiated a rate of 57% with the NIH [2]. So, they had about $473M in direct costs. The new rate would be 15%, which is ~$71M. $270M-$71M = $200M.

[1] Page 24: https://ucla.app.box.com/v/acct-pdf-AFR-22-23 [2] https://ocga.research.ucla.edu/facilities-and-administrative...