Posts by peter_mcmahan@mas.to
(DIR) Post #ARvMPmJquAaNssJMBs by peter_mcmahan@mas.to
2023-01-15T02:59:16Z
1 likes, 1 repeats
Wow. RStudio (recently rebranded as Posit) is partnering with Palantir. Yes, the same Palantir with ties to Cambridge Analytica. The same Palantir that partners with and actively supports ICE and other oppressive surveillance. This is really not ok ethically, and can only mean a bad direction functionally for the company's software going forward. I never use RStudio for research myself, but it looks like I'll stop using it in teaching. https://finance.yahoo.com/news/palantir-posit-partner-integrate-rstudio-115900656.html#rstudio #palantir #posit #ice
(DIR) Post #Ajaxwc7lGGLkLJXQXI by peter_mcmahan@mas.to
2024-07-04T18:11:10Z
2 likes, 1 repeats
hey #datascience nerds: can I just gush for a second on how great the combination of #ZFS and #LMDB is?I'm working with a 2.1 TB LMDB database of academic citations. I have half a dozen processes hammering away at this at random. But between ZFS's compression and excellent caching and LMDB's clever use of caching I'm barely hitting the disk at all (99% cache hit rate), and it's only using about 15GB memory.#networkanalysis #sna #computationalscience
(DIR) Post #Aynd9KAvi2Vy8lsHb6 by peter_mcmahan@mas.to
2025-10-02T13:07:49Z
0 likes, 0 repeats
@foone I absolutely adore the chaos energy of this project. But does the metadata on the first disk mean that if I want to start watching at, say, disk 120/234 I'd need to start with disk 1 and then "seek" forward? And that if disk 1 is lost or damaged then the rest of the video is unplayable?
(DIR) Post #AypecBeMyOkQMttJTs by peter_mcmahan@mas.to
2025-10-03T12:34:08Z
0 likes, 0 repeats
@futurebirdThe university I work at has constant workshops on how use "AI" in teaching (!) and how to write prompts for academic work. They have gone all-in on Microsoft so Copilot has been added to our LMS (and of course is pushed in all the MS products we have to use)
(DIR) Post #B14j1puzyBKTAhLuCm by peter_mcmahan@mas.to
2025-12-09T13:52:37Z
0 likes, 1 repeats
@tiago I may give this post to my students next time I teach social networks.Network analysis seems to be especially vulnerable to trying to describe answers without figuring out what the question is first. So many of our methods assume "structure" is a feature only of a network, divorced from what that network is meant to represent, and therefore just a technical problem to be solved.(I'm looking at you, "community detection," but don't think you're off the hook, "centrality.")