When you write your blog post with claude, I start hitting these phrases and then start only skimming and looking for them, and stop reading what you have to write.

You have to bet on the IPO knowing only the information at the time, not knowing what the future growth of the stock was.

"Hey, maybe we shouldn't send men over the top only to die in a barrage of machine gun fire..."

"That's just victim blaming--it isn't our behavior that needs to change; it is the enemy's behavior that needs to change."

Victim-blaming is another term that is apparently becoming a meaningless vibe.

I'm definitely hoping you succeed. I'd love to be able to write a small language (maybe that didn't do absolutely every last thing you can do in Rust, but was ergonomic and easy to use) that had a borrow checker and didn't require writing a nuclear reactor.

The point is to inflict economic losses on Yandex and the companies using those datacenters, which hits the elite around Putin. That is the fault line to try to establish and then exploit/widen.

I mean, if Russia is striking datacenters in London/Dublin/Frakfurt/Paris/Amsterdam, then we've got really big problems.

And is the US even allied to Europe, or is Europe going it on their own? They might do better to move to datacenters in China at that point.

Barring that kind of a situation, moving European data back to Europe makes the most sense.

Putting datacenters in the Baltic states might not make the most sense, tho.

Start a company that does ALPR and facial recognition, but narrowly targeting politicians.

An event at one orbital altitude is also not confined to that altitude; debris can reach higher altitudes and affect satellites in those orbits.

Nope, the Russian public is mostly entirely checked out of politics and will roll with any inconvenience. You won't see any rioting in the streets due to this.

The point is to inflict economic harm on the elite around Putin--the owners of Yandex, the other companies using those datacenters, etc.

The point is not to take out the average Russian's internet search index.

A more important goal is for these strikes to get headlines and serve as Western-facing propaganda, with "hitting Russia's Google" and "hitting Russia's Amazon" being easily digestible for those in Europe and America. That still beats the effect this will have on the average Russian.

Russia isn't a Western Democracy.

(We did nominally succeed in regime change in the country that didn't attack us, though...)

When the Soviets went into Afghanistan, the mujahideen did not have the domestic capability to build Stinger missiles themselves...

My brain can't help itself in looking for AI tells and gauging if writing sounds like AI or not.

Then once it latches onto the idea that the writing is probably AI, then it feels like half my brain is just reading the text looking for the AI smells.

Calling it "AI smells" like "code smells" might actually be what is going on in my brain, and I start code reviewing it at that level rather than reading it for what it does.

I really would prefer people not just post slop. I don't care if you use it to clean up your own writing, but don't just let it rip.

No one had ever made the Kessel Run in under 12 parsecs until Han Solo, either.

https://arxiv.org/abs/2610.08144

> Autoformalisation is increasingly used to verify mathematical texts, including those generated by AI, as in OpenAI's announced proof of blow-up of solutions to the Navier-Stokes equations. In this process, an AI system translates the text from a natural language (NL) into a formal language such as Lean. Once this translation is done, the argument expressed in the formal language can easily be mechanically verified. The purpose of this article is to demonstrate why this process may offer no confidence in the original NL argument, owing to the various difficulties in performing the translation semantically faithfully. In particular, we highlight that the problem of resolving ambiguities in mathematical NL text, which is necessary in order to provide semantically faithful translation, is arbitrarily high up in the Solvability Complexity Index (SCI) hierarchy/arithmetical hierarchy (the SCI =∞). Hence, informally, providing semantically faithful AI autoformalisation is harder than any computational problem including the Halting problem (which has SCI =1). To demonstrate the effect of this result we provide several examples of AI mistranslations of NL statements and proofs into Lean in practice, resulting in mismatches between NL proofs and their Lean `verifications'. These include OpenAI's announced Navier-Stokes proof. In particular, we show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to the Navier-Stokes equations.

And I don't think that paper addresses it, but if the LLM can find a bug in Lean and exploit it to prove something, there's a good chance it will find it and not report it. So if you've got some million-line proof in Lean, spit out by an LLM, you still can't quite trust it, even after validating the problem transcription.

(This is the same category of problem as the huggingface hacking incident, where the LLM finds and exploits an unintended cheaty loophole)

That's just an argument that you shouldn't do anything if you can't do it perfectly.

We know we've got it pretty well close to accurate, for the bulk of the observable universe.

If c changes, then chemistry changes, and stuff like hydrogen absorption lines shift. This is how scientists have looked for changes in c in the early universe.

which are not supermarkets.

I doubt the answer to this is "none".

And how many of them are just exploiting some loophole that will need to be closed in the problem definition?