Lightnews — Scholar-powered news

Reposted by Russ Poldrack

Emily G. Jacobs @emilyjacobs.bsky.social · 7d

These datasets live on OpenNeuro. Shout out to @russpoldrack.org team's vision to build a data-sharing platform that accelerates discovering + maximizes the ROI of painstakingly collected data.

Access these datasets and more at: wbhi.ucsb.edu/researchers/...

4 14

Reposted by Russ Poldrack

Emily G. Jacobs @emilyjacobs.bsky.social · 7d

When @laurapritschet.bsky.social & Pavel Shapturenka set out to build the 28&Me + 28&He datasets, I don't think any of us could've predicted the spectacularly creative ways the datasets would be used years later. That's the power of open science. 👇🏼

1 8 22

Russ Poldrack @russpoldrack.org · 14d

This is pretty hilarious github.com/TomLefley/i-...

GitHub - TomLefley/i-am-absolutely-right: Counts the number of times Claude Code has said I'm absolutely right

Counts the number of times Claude Code has said I'm absolutely right - TomLefley/i-am-absolutely-right

github.com

2 2 16

Russ Poldrack @russpoldrack.org · 17d

A better way to track problem solving during agentic coding - the latest in my Better Code, Better Science series russpoldrack.substack.com/p/a-better-w... - I have found this relatively simple approach to work very well.

A better way to track problem solving during agentic coding

Better Code, Better Science: Chapter 5, Part 5

russpoldrack.substack.com

1 12

Russ Poldrack @russpoldrack.org · 20d

thanks Micah, so glad to hear that!

1

Russ Poldrack @russpoldrack.org · 20d

I know that citation statistics are faulty indicator, but I have to say that there is something humbling in knowing that people have decided 100,000 times to cite a paper that I was involved in.

10 10 250

Reposted by Russ Poldrack

Adam Bonin @adambonin.bsky.social · 22d

My mom called me last night about the Kimmel firing, saying "This is how Hitler got started!" I quickly responded, "By firing the late night tv hosts?"

Turns out, mom was right.

Goebbels Ends Careers of Five 'Aryan' Actors Who Made Witticisms About the Nazi Regime

Feb. 4, 1939

February 4, 1939, Page 1

BERLIN, Feb. 3.—Propaganda Minister Joseph Goebbels today ended the professional careers of five "Aryan" actors and cabaret announcers by expelling them from the Reich's Chamber of Culture on the grounds that “in their public appearances they displayed a lack of any positive attitude toward National Socialism and therewith caused grave annoyance in public and especially to party comrades.” The five include perhaps the best known German stage comedians who survived previous Chamber of Culture purges and still dared to indulge in political witticisms—namely, Werner Finck, Peter Sachse and “The Three Rulands,” represented by Helmuth Buth, Wilhelm Meissner and Manfred Dlugi. Their expulsion means that they are henceforth forbidden to appear before the public in Germany.

Besides motivating this action in an official communiqué, Dr. Goebbels also publishes a long article in the Voelkischer Beobachter in which he denounces them as “brazen, impertinent, arrogant and tactless" and generally imitators and successors to Jews. Simultaneously he denounces the "society rabble that followed them with thundering applause—parasitic scum, inhabiting our luxury streets, that seems to have only the task of proving with how little brains people can get along and even acquire money and prominence."

As regards the details of the "crimes" of which the five are accused, Dr. Goebbels mentions that they made political witticisms about the colonial problem, the Four-Year Plan and Chancellor Hitler's monumental building program and one of them even raised the question of whether there was any humor left in Germany today.

What amused the public most, however, and presumably roiled the National Socialist authorities most—although Dr. Goebbels does not mention it—is that they deftly, but unmistakably, caricatured some gestures, poses and physical characteristics of Nat…

58 2.3K 5K

Russ Poldrack @russpoldrack.org · 22d

this should be fixed now - I moved the offending files to an examples directory

1 1

Russ Poldrack @russpoldrack.org · 22d

thanks! will work on that ASAP. Glad it ran easily for you!

1 1

Russ Poldrack @russpoldrack.org · 22d

I have just dropped a new version of my academicdb project, which helps prolific researchers generate a CV automatically. Now with a web interface! If you are interested, please try it out and let me know what you think - it takes a bit of setup work but then runs easily using Docker.

GitHub - poldrack/academicdb: Project to maintain a database for CV/website rendering

Project to maintain a database for CV/website rendering - poldrack/academicdb

github.com

2 8 49

Reposted by Russ Poldrack

Ole Jensen @olejensen.bsky.social · 22d

In our Trends in Cogn Sci paper we point to the connectivity crisis in task-based human EEG/MEG research: many connectivity metrics, too little replication. Time for community-wide benchmarking to build robust, generalisable measures across labs & tasks. www.sciencedirect.com/science/arti...

Confronting the connectivity crisis in human M/EEG research

The cognitive neuroscience community using M/EEG has not converged on measures of task-related inter-regional brain connectivity that generalize acros…

www.sciencedirect.com

2 28 87

Reposted by Russ Poldrack

Jamie Cummins @jamiecummins.bsky.social · 23d

Can large language models stand in for human participants?
Many social scientists seem to think so, and are already using "silicon samples" in research.

One problem: depending on the analytic decisions made, you can basically get these samples to show any effect you want.

THREAD 🧵

The threat of analytic flexibility in using large language models to simulate human data: A call to attention

Social scientists are now using large language models to create "silicon samples" - synthetic datasets intended to stand in for human respondents, aimed at revolutionising human subjects research. How...

arxiv.org

12 150 330

Russ Poldrack @russpoldrack.org · 23d

I am really glad that I'm not playing a drinking game where I have to drink every time Claude Code says "You're absolutely right!" because I would probably be dead from alcohol poisoning.

1 24

Reposted by Russ Poldrack

Governor JB Pritzker @govpritzker.illinois.gov · 23d

A free and democratic society cannot silence comedians because the President doesn’t like what they say.

This is an attack on free speech and cannot be allowed to stand.

All elected officials need to speak up and push back on this undemocratic act.

1.8K 18K 61K

Russ Poldrack @russpoldrack.org · 23d

I think the assumption is that longer tasks are harder. Certainly they require more human work, so the longer the task is that the agent can do, the more useful it can be for humans (since the agent will take much less time to do those tasks).

1

Russ Poldrack @russpoldrack.org · 24d

it certainly doesn't seem to be slowing down: metr.org/blog/2025-03...

Measuring AI Ability to Complete Long Tasks

We propose measuring AI performance in terms of the *length* of tasks AI agents can complete. We show that this metric has been consistently exponentially increasing over the past 6 years, with a doub...

metr.org

1 1

Russ Poldrack @russpoldrack.org · 24d

well, it's definitely changing rapidly - the tools can get so much further on their own now than they could a year ago. interesting question as to whether there is a wall that they will ultimately hit and where that will be located.

1

Russ Poldrack @russpoldrack.org · 24d

Strategies for effective coding agent usage - the latest in my Better Code, Better Science series russpoldrack.substack.com/p/strategies...

Strategies for effective coding agent usage

Better Code, Better Science: Chapter 5, Part 4

russpoldrack.substack.com

1 3 10

Russ Poldrack @russpoldrack.org · 24d

Similarly, Claude Code has a very difficult time following a test-driven development model. Despite clear instructions, it will regularly pronounce "All tests are passing!" even though no functional code has been written yet (violating TDD). These models really like to succeed, at any cost.

Russ Poldrack @russpoldrack.org · 24d

It's interesting how some instructions are really hard for coding agents to follow. Recently I have been using a problem file, in which I include an instruction to only mark a problem as fixed once I have confirmed that the fix worked (see image below). Claude Code basically refuses to do this.

1 1 12

Russ Poldrack @russpoldrack.org · 24d

It's interesting how some instructions are really hard for coding agents to follow. Recently I have been using a problem file, in which I include an instruction to only mark a problem as fixed once I have confirmed that the fix worked (see image below). Claude Code basically refuses to do this.

4

Russ Poldrack @russpoldrack.org · 24d

This great piece by @addyosmani.bsky.social resonates strongly with my recent experience using AI coding agents

The 70% problem: Hard truths about AI-assisted coding

A field guide and why we need to rethink our expectations

addyo.substack.com

3 12 43

Reposted by Russ Poldrack

Hakeem Jefferson @hakeemjefferson.bsky.social · 27d

I have many thoughts about Charlie Kirk—and perhaps even more about the white elites, including some on the left, who insist we can’t hold multiple realities at once. We can. And we must.

A brief 🧵

1 130 440

Russ Poldrack @russpoldrack.org · 28d

Great explainer on variance partitioning

Jörn Diedrichsen @diedrichsenjorn.bsky.social · Sep 10

Variance partitioning is used to quantify the overlap of two models. Over the years, I have found that this can be a very confusing and misleading concept. So we finally we decided to write a short blog to explain why.
@martinhebart.bsky.social @gallantlab.org
diedrichsenlab.org/BrainDataSci...

1 22

Reposted by Russ Poldrack

jeanette-mumford.bsky.social @jeanette-mumford.bsky.social · 29d

Do you ever wish you could just use python to pull together the files and code for running FSL's randomise? Me too! I made this: github.com/jmumford/ran... It will even replace the numbers in the file outputs with contrast names of your choosing (and replace corrp with 1minusp).

GitHub - jmumford/randomise-prep: Generate design matrices, contrasts, and scripts to set up FSL randomise analyses.

Generate design matrices, contrasts, and scripts to set up FSL randomise analyses. - jmumford/randomise-prep

github.com

19 48

Russ Poldrack @russpoldrack.org · Sep 10

I’ll write more soon, but I’d suggest exploring it to help with the generation of specification files for agent-based workflows.

2