[X.com] by @iruletheworldmo

it’s over

turns out the rl victory lap was premature. new tsinghua paper quietly shows the fancy reward loops just squeeze the same tired reasoning paths the base model already knew. pass@1 goes up, sure, but the model’s world actually shrinks. feels like teaching a kid to ace flash cards and calling it wisdom.

so the grand “self-improving llm” dream? basically crib notes plus a roulette wheel: keep sampling long enough and the base spits the same proofs the rl champ brags about, minus the entropy tax. it’s compression, not discovery.

maybe the endgame isn’t better agents, just sharper funnels. we’ve been coaching silicon parrots to clear increasingly useless olympiad hurdles while mistaking overfit for insight. hard not to wonder if we’re half a decade into the world’s most expensive curve-fitting demo.

Image from tweet

View original on X.com

[X.com] by @NormSNLJokes

And finally, legendary pool hustler Minnesota Fats passed away Wednesday.

You know, now he's probably up in Heaven, racking them up for a game with Saint Peter.

Or maybe he's in Hell, where demons gnaw at his flesh, and the agonies of the damned never cease.

Either way, he'll be missed!

Image from tweet

View original on X.com

[X.com] by @Plinz

The word “science” has two meanings, which are often confused. Traditionally, it stands for the rational pursuit of knowledge, while the contemporary use refers to institutionalized consensus seeking via peer review. The switcheroo seems to have happened between 1960ies and 90ies

View original on X.com

[X.com] by @Musa_alGharbi

As I see it, the White House made two big miscalculations here.

First, they went waay to extreme on the demands, and had too many of them, to the point they became self-refuting and compliance became literally impossible, as I noted. In such a scenario, the only possible response is non-compliance. Had they made demands of Harvard that were similar in scope and structure to those made at Columbia (which were also extreme and unacceptable from my POV), there is a decent chance that Harvard could've folded the same way Columbia did.

BUT, their second miscalculation is that, after Columbia capitulated, rather than acknowledging this, thanking them for their cooperation, unlocking the funds, etc. -- instead, they tried to turn the screws even more and ratcheted up their rhetoric and demands. THIS sent a signal to Harvard and other schools that even if it were possible to comply with the Administration's orders (which, again, in this case it literally is not possible), there is no point. Compliance gets you nothing. So why comply?

An administration that wanted to encourage other universities to rapidly fold like Columbia would've tried to create an incentive for this, by rewarding schools who provide easy compliance, and seeking to make an example of those who resist. Instead, they chose to make an example out of Columbia despite the university doing everything they asked and striking a completely conciliatory posture.

Both of these moves left Harvard with few plausible options other than defiance. It's a total self-own on both counts.

View original on X.com

[X.com] by @Musa_alGharbi

I'm struck that this guidance includes and mandates the imposition of many new ideological litmus tests while also demanding that the university, "abolish all criteria, preferences, and practices, whether mandatory or optional, throughout its admissions and hiring practices, that function as ideological litmus tests."

How on earth is a university supposed to comply with the demands articulated below without policies that function as ideological litmus tests? I mean, they explicitly call for reducing or eliminating "activist" scholars, and for mandating the hire of people on the basis of ideology if a department is perceived to have too little ideological diversity...

No wonder it was rejected. It's literally impossible to actually comply, because the demands are, themselves, blatantly contradictory.

Quoted Tweet:

John Sailer @JohnDSailer

NEW: The Trump admin sent an updated list of demands to Harvard on Friday (April 11), which significantly expanded on the earlier letter it sent on April 3. This apparently prompted Harvard to reject the admin's demands outright, even with hundreds of millions of dollars on the https://t.co/zO1yHbCWLb

Image from quoted tweetImage from quoted tweetImage from quoted tweetImage from quoted tweet
View original tweet Mon Apr 14 18:04:30 +0000 2025

View original on X.com

[X.com] by @PsychRabble

Social Psych News Flash: JUST ACCEPTED AT META-PSYCHOLOGY Registered Replication Report Famous Study Finding (N=127, shown below) "STEM faculty biased against women" -- so famous its been cited over 4000x & got the lead author invited to Obama's White House -- not merely fails to replicate (N=1208), but FULL BLOWN REVERSAL - STEM faculty were biased against men. More to come. Link to RRR in reply. Shown, ratings of male vs. female lab mgr applicants (w/identical credentials).

Image from tweet

View original on X.com

[X.com] by @typesfast

Flexport's team was able to reverse engineer the formula the Administration used to generate the "reciprocal tariffs."

It's quite simple, they took the trade deficit the US has with each country and divided it by our imports from that country.

The chart below shows the predictions of this formula plotted against the actual new tariff rates.

Image from tweet

View original on X.com

[X.com] by @Camp4

Last summer, during a climbing trip in France, I reinjured my back (not while climbing). 12/10 pain and I had to be wheeled through the airport.

The flight home was an experience I never want to relive. I resolved to go all-in on rehab.

If you’d have told me that 9 months later I’d be able to do this, I’d have called you crazy. After all, I’m a tall, old guy—I’ve never even been able to touch my toes.

Jefferson Curls are perhaps the single best exercise for strengthening the intricate scaffold of muscles that support the spine. It also increases range of motion throughout the posterior chain.

These days almost everyone has weak hamstrings, glutes, and back from sitting too much.

If you have chronic back, hip, or knee problems the solution probably isn’t surgery or chiropractic or meds, it’s movement with resistance. There are no shortcuts—just consistent hard work.

Invest in spine health before it becomes a problem—your 50-year-old self will thank you.

View original on X.com

[X.com] by @Hesamation

Top 10 highly technical YouTube channels to learn AI from scratch:

1\ Andrej Karpathy - a blend of general and technical content, Zero to Hero playlist is a must-watch https://t.co/tNNLTyYT86 @karpathy

2\ Umar Jamil - highly technical, implements ML and LLM techniques from scratch https://t.co/AP9ecr04wq @hkproj

3\ Simon Oz - technical low-level machine learning videos https://t.co/7StZUevNXp

4\ Tunadorable - paper review, implementation, triton https://t.co/vwOhYtui07

5\ GPU Mode - technical interviews and walkthroughs about anything related to GPUs https://t.co/gfRuxCeBEn

6\ AI Jason - AI experiments, software design, and new techniques beautifully explained https://t.co/7O7DRRxSQa @jasonzhou1993

7\ Ferdinand Mom - everything related to distributed training & inference https://t.co/gmV7vBQixg @FerdinandMom

8\ Welch Labs - unique in-depth look at machine learning complexities like nobody else https://t.co/ovEt9WnEx2 @welchlabs

9\ Artem Kirsanov - neuroscience and machine learning from a different look, great visuals https://t.co/sJBmDbP7Qy @ArtemKRSV

10\ David Ondrej - new models, building apps with AI, practical for developers https://t.co/BEOr0MgHag @DavidOndrej1

This list is for a practical and technical audience. It's so hard to hand-pick just 10 channels, there are so many great ones out there making great content.

If you know any other channels, LMK.

Image from tweet

View original on X.com