Is Trump’s Iran War Like a Katrina Moment? It’s no great insight to say Trump’s impulsive Iran War has been...

Is Trump’s Iran War Like a Katrina Moment? talkingpointsmemo.com/edblog/is-tr...

You know there's some guy who does that mrna vaccine for pancreatic cancer is going "well acktuallly..."

Anyway airports suck and they water down the drinks.

Gil Duran tweet: TLDR: Fascism in response to Palantir's long fascists screed on X.

"Your Account is Suspended" Message on X

The CEO of Palantir posted a fascist manifesto on X.

I pointed out that it was fascist—which resulted in a permanent suspension from X (my second time!).

So, when you hear the tweeters complaining that BlueSky is intolerant, remember why many of us came here in the first place.

the penne opticon

Yeah and frankly being in airports feels soul destroying at the best of times. Lounges have even gotten worse.

Also love an airport like Burbank where I can just show up and be ready to fly in 10 minutes.

Yeah, though I would say it's continually surprising what these systems are capable of.

Though they aren't *Just* next token anymore, right?

An Ode To: Trashcan Man (The Stand) YouTube video by TheSpringwoodSlasher

www.youtube.com/watch?v=LA25...

"My life for you!"

Yeah I don't think aesthetics works that way -- I imagine the result is the 'premium mediocre' version.

By way of contrast, I know that some imagine generation systems like Midjourney were actually trying to optimize for a particular *aesthetic* rather than just decomposable best.

I don't think I would ever describe Claude as thinking. But since reasoning is something that can be done formally, I would in some domains describe it is that. This may be a source of confusion here.

Also, water in a pipe maze can be Turing complete!

It's also true for math, where there are proof systems that can vouch for your proof.

You could in theory do Mechanical Turk style verification for other solutions as well. "Claude find me just the right shade of blue that 98/100 people approve of!"?

Well with security you really could have a series of questions that could be answered -- both by the LLM or deterministically. Does this crash the program? Produce exploit code that pops calc.exe, and so on.

But I don't know if anybody is debating that it's the same as how humans reason/think/figure things out? Rather than whatever is doing is capable of solving problems, and doing what most of us would call reasoning through them.

Certainly no humans could be primed by the way a question was asked!

But more seriously, for sure you're right:

But we can't forget there are also things it will figure out if you re-roll the prompt a dozen times!

Right, the bit I don't get is how you can say it isn't producing original results. That it got there by means a human does not possess is ... sort of the point?

For reference, the prompt for this was something along the lines of "I think there's a security bug in this code; can you find it?".

Hey looks Rainbows End is happening. :|

War on general purpose computation

Right but this is not just combinatorics here (that would take until the heat death of the universe to find any one bug of this complexity!).

In fact, Mythos was not even hooked up to a fuzzer to find this particular bug [!], if Anthropic's paper is to be believed.

Deep question. We know they're Turing complete by now (bsky.app/profile/timk...), but just zero idea about something that restrictive.

Also fwiw I don't think Claude Mythos "understands" things in a way that human do, so I shy away from using that language. Rather it has a model of the discipline which can be used as an exploratory tool. Like shining a flashlight in a vast, multi-dimensional space.

This is not necessarily an endorsement of AI in general, or even AI in specific (maybe it will *never* be cheap enough to do this at scale, but I wouldn't take that bet), but if you want evidence AI can produce original results, here it is.

Note Provos is semi-famous in infosec and (needless to say) is not a bad programmer.

It was striking for me because not only is this in a space where people get money for finding security bugs, in an OS widely regarded as secure, it was, put plainly, a new thing. And in my discipline!

Assessing Claude Mythos Preview’s cybersecurity capabilities | Niels Provos | 32 comments The past comes back to haunt you. TIL that clearly I am a bad programmer: Anthropic posted on the abilities of their Mythos model to find security vulnerabilities today: https://lnkd.in/gcjSRcSS - On...

Probably the best example is Mythos finding that 27 year old OpenBSD bug.

It's notable because here we have the guy who wrote the code talking about this brand new exploit being found and developed. It's striking because OpenBSD is widely considered well-studied and secure.

A man wearing an eye of sauron shirt (heh) carrying a sign that says 'Your Code, Their Cages' with the Palantire Logo. Next to him is a woman with a sign that says 'Stop the Weapons of Math Destruction'.

Similar feelings.

how is this literally happening in real life

'if you don't support pig butchering scammers using Palestine to steal money from the gullible, you're pro-genocide' is the product of extreme brain cooking

>unarmed pleasure vessel inexplicably capable of running a blockade

What did he know?

This is good feedback, particularly as a diabetic on metform + glipizide.

My concerns have been about the claims it affects things like appetite for risk as well as appetite for food.

I believe I could do with no desire for food for a few years, but dunno about long term.

I also have my doubts about the twelve (I guess?) companies who are getting access to Mythos.

So I guess my read is this is well intentioned, but if one believes they can't keep it safe for very long, I hope it comes for the rest of us sooner rather than later.

Throttling makes sense to me, but only insofar as we believe Anthropic can keep Mythos under wraps until we go through the equivalent of a coordinated disclosure. Same company leaked Claude Code in the last quarter, so I just have a hard time believing their security process survives the attention.

Scoop: NSA using Anthropic's Mythos despite blacklist The government's cybersecurity needs are outweighing the Pentagon's feud with Anthropic.

scoop w/ @mccuri.bsky.social: The NSA has been using Anthropic's Mythos despite the Pentagon's instance that Anthropic is a "supply chain risk," according to two people.

The NSA was one of the 40+ orgs that got initial access to the model weeks ago, one source added.

www.axios.com/2026/04/19/n...