Michael Mullarkey

@mcmullarkey.bsky.social

Clinical psych PhD | ML Engineer in sports Subscribe to my newsletter on data, learning, and building: https://buttondown.com/datadash Opinions are my own and don't reflect the opinions of my employer

I’ve alway’s thought it’s a shame that WebAssembly has the word “web” in it. Gives it a connotation that it’s “only for browsers” or something.

 Meanwhile, NASA’s over here with a flight-compliant derivative (SpaceWASM) intended for spacecraft!

BildBild

Enjoyed talking to @psychscience.bsky.social about my group’s new(ish) paper for the Observer. I’ve talked about this on here at an earlier stage. In 4 studies, we found no evidence that asking young adults to cut back on their social media improves well-being. Small thread 🧵 follows:

For Young Adults, Limiting Social Media Use Doesn’t Improve Well-Being

What happens when young adults are asked to slash their social media use to 30 minutes a day? In a new set of studies, most did not follow through, and those who did cut back …

psychologicalscience.org

i swear to god, asking: “what is the problem you are trying to solve? followed by: “what information or data do you need to solve this problem?” are like, some of the single most important questions you need to ask a stakeholder on all kinds of projects, and so many people skip it

I like cultivating generalist skills. I think generalists can bring needed perspectives to fields where niche expertise already exists. And "being a generalist" ≠ "posts confidently about a ton of unrelated topics" When someone does that they will get important stuff wrong + come off weird at best

Calling everything a measurement problem flattens out other difficulties too much And guessing "measurement problem" every time works better than it should

Daniel P. Moriarity@dpmoriarity.bsky.social · 3w ago

Psychology has been riding the intensive longitudinal train for a bit, but what about immunology? Now live @brainbehavimm.bsky.social, our physiometric report on the stability of inflammatory proteins collected every 3 days for 22 days during a natural stressor 1/X

Within the past day we've gotten: - A ~Fable level model at ~Haiku level pricing (GLM 5.3) - An ~Opus 4.6 level model runnable on local hardware (Qwen 3.8 Max 27b) I switched to an open model stack as my daily driver after GLM 5.2 dropped. That move felt like a risk, and so far it's paid off

Charlie Gao@shikokuchuo.net · 4w ago

Some of us at Posit have switched to Kimi K3 as a daily driver. Really proud that my colleagues host this open weights model in Posit AI. I've not had to reach for Fable all week!

Given the OSF news (shout out to them forever, they made what felt impossible possible for a long time) If you could summon a software tool to help you deal with the upcoming changes, what would it be? I promise nothing + know there are great existing tools. I'm just curious where gaps might be

The idea that there's an easy abundance of free resources to learn to code is laughable once you've done what I've done and been a VP of research at a company that helped thousands of people break into tech careers. There's an abundance of *beginner tutorials*

This week on Counting Stuff, how does locating lightning work? There's sensor networks, some run by volunteer hobbyists, that are able to pinpoint where lightning strikes to within a few hundred meters. I dug into how they do that #dataBS www.counting-stuff.com/so-how-does-...

So how does lightning locating work?

Between advanced math and radio sensors, it's possible to locate where lightning strikes from hundreds to thousands of kilometers away. It's wild.

counting-stuff.com

I'm happy to share `ir`, a new command-line tool for running portable R scripts and Quarto documents. It's inspired by my two favorite parts of `uv`: self-describing scripts and tools you can run without installing first. opensource.posit.co/blog/2026-07...

ir 0.1.0: self-describing R scripts and Quarto documents

`ir` is a new command-line tool for running portable R scripts and rendering Quarto documents whose package requirements, and optional R selection, live inside the file itself.

opensource.posit.co

Not every open-source project wants more contributors right now, and that's fine. Four shapes maintainers run — walled castle, clique, university, hippie commune — and your `CONTRIBUTING.md` is saying the quiet part out loud. No wrong answer, just an honest one. blog.stdlib.io/do-you-want-...

Do you WANT contributors?

Four ways to run an open-source project—castle, clique, university, commune—and which one you're really running.

blog.stdlib.io