Rahul Dave

If you get a divorce just modify the engraving to make it cosh^2x and sinh^2x

A man in a hot air balloon realized he was lost. He lowered his altitude, spotted a woman in a boat below and shouted to her: “Excuse me, can you help me? I promised a friend I’d meet him an hour ago, but I don’t know where I am.” The woman consulted her portable GPS and shouted back, “You’re in a hot air balloon approximately 30 feet above ground elevation of 2,346 feet above sea level. You’re at 31 degrees, 14.97 minutes north latitude and 100 degrees, 49.09 minutes west longitude." The man rolled his eyes and yelled, “You must be a Democrat." “I am,” replied the woman. “How did you know?” “Well,” answered the balloonist, “everything you told me is technically correct. But I have no idea what to do with your information. Frankly, you haven’t been much help to me.” The woman smiled and yelled, “You must be a Republican.” “I am,” shouted the balloonist. “How did you know?” “Well,” said the woman, “You don’t know where you are or where you’re going. You’ve risen to where you are thanks to a large quantity of hot air. You made a promise you have no idea how to keep and you expect me to solve your problem. You’re in exactly the same position you were in before we met, but somehow now it’s all my fault.”

🔥

The best way to unblock Hormuz is to go back to the world before the continental drift

We now have a version of Omarchy where everything is working out-of-the-box with the new Dell XPS Panther Lake machine! Webcam, sound, display refresh. Huge thanks to @dell and @intel for this collaboration! Excited to get this released in time for first customer machines 🤘

Elon saved us from Twitter the way that asteroid saved us from dinosaurs

I hope this one catches on.

An eatery in Pune 38

Jim Hacker: Humphrey, we have to do something about Iran. Sir Humphrey Appleby: Prime Minister, the government is already doing a great deal. Jim Hacker: Such as? Sir Humphrey Appleby: Monitoring developments, coordinating with allies, reviewing contingency plans and expressing concern. Jim Hacker: That all sounds like nothing, Humphrey. Sir Humphrey Appleby: On the contrary, Prime Minister. In diplomacy it is vital to appear active without becoming involved. Jim Hacker: The Americans are bombing things, the Iranians are firing missiles, the Strait of Hormuz is practically closed and we’re… appearing active? Sir Humphrey Appleby: Precisely. Jim Hacker: Innocent people are dying, Humphrey! Sir Humphrey Appleby: Yes, Prime Minister. That is why the Foreign Office is drafting a very strongly worded statement about it. Jim Hacker: A statement won’t stop a war. Sir Humphrey Appleby: No, Prime Minister, but it will ensure that we are on record as having been extremely concerned while it was happening. Bernard Woolley: If I may, Prime Minister — the Cabinet Office has identified six possible courses of action. Jim Hacker: Good! What are they? Bernard Woolley: We can condemn the escalation, call for restraint, urge negotiations, support our allies, assist defensive operations or participate directly. Jim Hacker: And what do they recommend? Sir Humphrey Appleby: Supporting our allies. Jim Hacker: That sounds suspiciously like participating. Sir Humphrey Appleby: Oh no, Prime Minister. Participating means fighting. Supporting merely means allowing others to fight from places that technically belong to us. Jim Hacker: Humphrey, if Iranian missiles hit one of our bases, we’ll be in the war anyway! Sir Humphrey Appleby: Yes, Prime Minister, but we shall have entered it with the invaluable diplomatic advantage of being surprised. Bernard Woolley: It’s generally considered the safest way to enter a war, Prime Minister. Jim Hacker: How on earth can that be safe? Sir Humphrey Appleby: Because if the war goes badly, we can say we never meant to join it. And if it goes well, we can say we were there all along.

Well, I did a thing. I hope it's useful. I mapped every four-year college in the U.S. higher education along two dimensions — institutional resilience & post-college market position — using eight indicators from federal data along with a new measure of institutional AI exposure.

a flow of salmon around a sink of bears

Few know this, but I (George) was the only person in history to get a perfect score in CMU compilers, which is likely the best compilers course in the world. Combine that with crazy low level knowledge of hardware from 10 years of hacking. Then add a team of people who are talented enough to push back on my dumb ideas and clean up the implementations of the good ones. The team who keeps this whole operation running, software, infrastructure, and product. I love how there's no hype in deep learning compilers. It was one of the most annoying things about self driving cars, all the noobs who burned through billions on crap that was obviously dumb, and the companies who deserved to go bankrupt years ago if not for government bailouts (Tesla and China will devour them all). In this space, the competition is @jimkxa at Tenstorrent, @clattner_llvm at Modular, and @JeffDean at Google. Three of the living legends of computer science. And companies like @nvidia and @AMD, who are definitely live players, making single chips that have more power than the whole Internet two decades ago. This space is so fun to play in. If you haven't, read the tinygrad spec. It's all coming together beautifully.

Britain is 6% Muslim. Germany 5%. France 10%. Sweden 9%. Belgium 7%. At this rate of Islamic conquest, Europe will be majority Muslim sometime around the year 2847. I’d pencil in some mild concern for around 2600 and see how things look then. Now. The refugees. Since someone asked who’s paying for all this. Let’s follow the money back a bit further. America invaded Afghanistan, spent 20 years there achieving absolutely nothing, then left in such breathtaking chaos that people were literally hanging off aircraft. It then invaded Iraq over weapons that turned out not to exist, killed somewhere between 150,000 and a million people, and converted a functioning country into a sectarian hellscape. This is before we even get to the drones over Yemen, Somalia and Pakistan. The people washing up on European shores are, in very large part, the direct human wreckage of American foreign policy. America created the disaster. Europe is housing the survivors. And America is on the internet asking why Europe keeps letting people in. Remarkable cheek, really. As for eliminating indigenous culture: the United States actually eliminated its indigenous people. Deliberately. With rifles and government paperwork. Europe took in Syrian doctors. These are not comparable situations, and pretending they are requires a truly heroic indifference to history. The culture is fine. France still has the cheese. The Louvre is still there. Bach is still there. Nothing has been eliminated except, apparently, the ability to read a percentage.

Saw an Indian guy with Proud MAGA in the bio. Thought he must be a Trump supporter. Turns out he meant Make Andhra Great Again 😂

@eatonphil Pixi! Uses uv rust libraries underneath, binds conda and pip, great for polyglot projects

what's everyone using for python dependencies outside of pip and uv?

In as much as its knowledge is an interpolation between concepts it knows, i dont find this surprising. But what happens if you take it one layer higher and train it on how to write languages? But surely some of thqat is in the training set, so wouldnt one expect something more by in-context exposure to the languages BNF or other representations? A meta-learning, so to speak?

Ha — this is very similar to a homework assignment in my PL class last fall! cs.brown.edu/courses/csci17…

There is also pixi. uses the uv runtime underneath but arguably has a more mature implementation of dependency groups (using their notion of features) and a great task runner (like npm run) but with make like dependencies and the such. I have found it better for science projects like those with notebooks and so on and so forth, things that wont be packages or libraries, and for teaching. But that was not uv's point anyways. and uvx replaces pipx nicely and juv does the same for jupyter, so for many things it's actually simpler to use uv. Why do i bring this up? There are also two political aspects to this (witness the opposition to ghostty creator @mitchellh joining the board of Vercel): the openai/anthropic beef on the dept-of-war/hegseth thing but on the other hand openai is far friendlier to open source and OSS models in particular. So even in my own mind, while i cant say i am happy with openai's recent behavior, i am not sure what to think. Just to acknowledge this is a concern going through a lot of our minds and we are more confused than ever, and that having other choices which do what uv does but perhaps not at the extremely low level of friction uv provides is a good thing...

Thoughts on OpenAI acquiring Astral and uv/ruff/ty simonwillison.net/2026/Mar/19/op…

This is very very nice to see, multi-vector models really seem like a class of their own here!

@antoine_chaffin @lateinteraction do we understand why? Lot of the browsecomp problems seem to have pretty generic words necessitating good recall to start but good ranking order to not kave to make the top "k" very large. Is it the multi-embedding providing somewhat longer contexts helping the ranking with the late interaction helping the recall? I am trying to understand the mechanism by which this works so nice...

Antoine and team had trained a nice ColBERT late interaction model last year... Now they decided to try it on BrowseComp+, the canonical "deep research" task. Guess what, it's not only the strongest method by far but also basically solved the task (~90%). Who would have thunk!

BrowseComp-Plus, perhaps the hardest popular deep research task, is now solved at nearly 90%... ... and all it took was a 150M model ✨ Thrilled to announce that Reason-ModernColBERT did it again and outperform all models (including models 54× bigger) on all metrics