The First Nuclear Clocks Are Ticking: Two Teams Build a Clock Run by an Atom's Core. This is amazing. We've have cesium clocks before, but those are based on whole atom outermost electron transitions, releasing a microwave photon. The hyperfine splitting is purely in the electron cloud, and is easily perturbed by external EMF. The Thorium-229 is different, and is based on transitions in the n nucleus of the atom. Thorium is well-suited for this because a laser could easily excite the nucleus into an isomer state, and when it reverts back to ground state, released a photon. But the Chinese were able to make it a ticking clock, a continuous, self-correcting oscillator. Paper here.
––– 凄い –––
Terrence Tao does not understand cancer medicine. Tao had written that mathematics should be like cancer medicine, because in the latter, at least someone knows exactly the mechanism of a drug before it enters into phase III trials and becomes approved for use. Vinay Prasad counters that by saying that in some cases, we don't fully know. Yet the drug is used, but at least we know enough about how it works, even though new things are discovered all the time. This doesn't invalidate what Tao said, because there will still need to be mathematicians to find hidden insights and connections.
But other mathematicians aren't buying it. Some are saying that undergraduates shouldn't waste time pursuing a PhD. Those considering a career in mathematicians should say goodbye to the romantic notions of Ramanujan, Euler, Fermat, Newton. Because they're going to be poring over LLM and theorem prover output. Either theirs, or someone else's. And like my Buckmaster post yesterday, the whole process of manuscript and research proposal reviews is going to have to change as well. Some professor could easily slip your research dreams into a frontier model, and then off it goes...
––– 良くない–––
Jason Coxon, the guy who quit Anthropic, now says that he's 90% certain China has spies inside America's AI giants. And Satya Nadella says that we should assume all AI models are ‘compromised’. Not by the Chinese, but from a security standpoint. Frontier and open source models can't be trusted. Nadella wants an "emergency brake" installed on AI systems. AI agents are indeed getting out of control. One Anthropic model recently called in a fake homicide tip to police. Can't Anthropic control this?
––– 良くない–––
Speaking of Nadella, here's Microsoft 365, disappointing users who store their data on OneDrive. I'd get my documents and photos and other data off OneDrive and put them somewhere else.
––– 良くない–––
Here's a post if you want to go down a deep rabbit hole of how browsers render text in different languages, like abjad and abugida languages. And how Google's Noto font got its name.
––– 凄い –––
Reflect Orbital still wants to set up a giant space mirror to beam sunlight on earth in the night. It says that the mirror can't hurt your eyes, but is that really true? I think this is a colossal waste of money.
––– 良くない–––
If you're going to set up an analytics script on your website, at least set up a courteous one.
––– 凄い –––
Yeah, Apple has an iPhone 18 Pro demand problem. Apple can no longer amaze people, and now it's showing.
––– 良くない–––
Here's a nifty site. It's a World Factbook type of site, but it cuts to the chase and gives you the essential facts about a country that you're probably looking for. I've bookmarked it.
––– 凄い –––
Why Texas Is Making Data Centers Wait. There are so many factors that go into datacenter planning. Technology changes. Political environment changes. Sentiment changes. And you have to plan for when things go wrong, and how it's going to affect everything else.
––– 良くない–––