iiinova — where your digital self keeps evolving

Pwnallthethings @pwnallthethings.bsky.social · 7h

There are 19 known plausibly habitable planets close enough to earth that if they can pick up radio waves and have life and a tv could have seen the first episode of south park

2123

Pwnallthethings @pwnallthethings.bsky.social · 7h

Tyrannosaurus Rex is closer to the airing of the first episode of South Park than it is to stegosaurus

2183

Pwnallthethings @pwnallthethings.bsky.social · 12/10/2026

The other thing that twitter was really great at was information density because tweets were short. That rich or influential boomers are posting whole essays on X might make the platform in some sense influential, but it also masks that it stopped being twitter and became facebook for tech ceos

1570

Pwnallthethings @pwnallthethings.bsky.social · 12/10/2026

text is very bandwidth efficient, and very searchable, but both of those are increasingly irrelevant concerns

3452

Pwnallthethings @pwnallthethings.bsky.social · 12/10/2026

super spicy take but pushing buttons quickly to make a block of text is a very unnatural way to communicate, and us older generations got very good at it because other ways to communicate were comparatively impractical, and we like it because we're used to it, but it is not intrinsically *better*

2654

Pwnallthethings @pwnallthethings.bsky.social · 12/10/2026

nobody on either platform wants to admit it, but text-based micro-blogging in for the olds. young people are on youtube and tiktok (or equiv)

121342

Pwnallthethings @pwnallthethings.bsky.social · 12/10/2026

Couple years back there was a guy who thought it was cool to shoot up some substations. Anyway, long story short, he has another 23 years left on his prison sentence.

3403

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

(the answer might be yes! idk. I regularly go in planes and vehicles whose safety systems are opaque to me, tho at least for now they are intelligible to /someone/)

280

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

more interesting would be "do you want to get into, say, a plane that was designed and built entirely by AI, and whose designed safety features are entirely unintelligible by any human".

1210

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

sort of the worst possible industry for him to have picked here, given how many ultra-widespread medicines we have that (a) plainly and uncontroversially work (b) nobody knows the mechanism for

3340

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

1. blow up yandex datacenter 2. yandex bots stop searching the internet 3. cloudflare says "russian bots stopped" 4. people on internet who have very different idea of what "russian bot" means get wrong end of the stick about what is being measured or reported here

0403

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

(I don't think he's wrong, but he's also talking his book)

2160

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

ceo of company that builds systems on which agents run, rather than agent alignment: the solution to this is the system on which agents run, not alignment of agents

2220

Pwnallthethings @pwnallthethings.bsky.social · 11/10/2026

Ah yes, X, a nice quiet garden where nobody argues about AI or politics

21272

Pwnallthethings @pwnallthethings.bsky.social · 09/10/2026

On the one hand Grokku the destroyer hath turned his gaze upon the village and destroyed it entirely but on the other hand it you put Vaseline in your eyes and squint while eating handfuls of horse tranquilizers he makes some interesting points about tax reform

1200

Pwnallthethings @pwnallthethings.bsky.social · 08/10/2026

needless to say, you benefit from those higher wages if you are part of the segment that doesn't get laid off; if you're part of the segment that gets laid off you *might* be able to get a better paid job elsewhere in the industry, but many people become underemployed and have big wage loss

070

Pwnallthethings @pwnallthethings.bsky.social · 08/10/2026

I mean, higher efficiency tends to lead to better wages, tho at a lower multiple than the efficiency. But it basically never leads to reduced work weeks

1130

Pwnallthethings @pwnallthethings.bsky.social · 08/10/2026

If Amazon could 10x--hell, 100x--staff output, the vast majority of its staff would still be working 5 days a week. Efficiency gains don't translate into shorter work weeks. It is just not how this works or a thing that ever happens.

091

Pwnallthethings @pwnallthethings.bsky.social · 08/10/2026

I mean, he's literally *just* laid 1000 people off instead of letting ~ 2000 people work 2.5 day weeks because this literally always how companies handle efficiency gains.

1181

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

The clerk has no constitutional role. It's a cute theory but it's a wrong one that confuses the convention for the inherent power of the house

2221

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Then the member elects can vote under their own inherent authority to dismiss the clerk

1260

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Getting very much into the weeds but the member elects can dismiss the clerk

1221

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Yes. Senate rules carry over between sessions. House rules are readopted (or altered) typically as the first vote of the session after electing the speaker

1250

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Bit more niche but also you don't need to be sworn in by the speaker *at all* after an election. You do if you are elected mid-session (by house rules) but immediately after the election you are sworn by the speaker elect by convention, not by requirement

1075885

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

The House is not a continuing body like the Senate. It reconstitutes entirely after the election. No rules or leadership carry over to the next session except only to the extent that the next session affirmatively chooses to readopt rules or leadership

1392790

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

If Democrats win, johnson will not be speaker at the point where they are swearing in the new Congress. They are sworn in by the speaker elect, who is chosen by the member elects

522872586

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Because I can't edit posts, and when you have 200k followers any top level post with an error in it gets a deranged response that isn't worth it

2240

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

In either case, this is a weird problem. AI is producing important math results. That's not really in dispute anymore. But it's producing them in a way that is quite alien to the way math normally operates, and extremely alien to the way it is normally socialized at the professional level.

61485

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Do we want to live in a world where mathematics progresses beyond the speed and capability of human insight? I think for mathematicians the answer is likely to be "no". For fields that use mathematics--which include important practical fields--the answer is probably annoyingly contingent.

51074

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

An interesting Q is whether the insight is *necessary* for utility anyway. For a lot of mathematicians that insight is a lot of the point. For a lot of practical STEM fields it is often a nice-to-have rather than a need-to-have. And can AI substitute the insight? And if so what is mathematics *for*?

3621

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

What makes this complicated is that frontier labs aren't wrong either. Those facts are useful independent of the insight, because they can be mined for those insights. And the facts can be valuable even without the insight of *why* in other STEM fields too.

3572

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

but theorems output as the collateral exhaust fumes of model evals give you facts, not insights. At least not directly. Only by studying those facts intensively can you relate those facts back to genuine insight. And that's the heart of the fissure between labs and the mathematical community.

111818

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

e = mc-squared isn't a neat-o fact because it's a fact and facts are fun, it's neat-o because it enables a whole new way to think about energy and matter, and see the universe, and thoseenable you to make predictions--and later test those predictions--against observations about the universe.

1835

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

And those results let you understand the field in a more profound and deeper sense that enables more interesting observations later. Or another example: that energy is related to mass isn't just a fun fact, it's a *profound* observation about the universe as a whole that enables new theory-making

3893

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Some mathematicians like proving things, to be sure, but a lot of the profession is trying to understand deeper relationships between sub-fields. That there is a relationship between prime-order groups and prime fields is not just a cute fact; it tells you something fundamental about both sub-fields

21056

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Some is very real: this is releasing proofs at a rate that exceeds professional mathematics' ability to socialize and understand the underlying mechanisms going on in the proofs. Professional mathematics does prove things in math, but it is still not (just) about proving things with math.

11013

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

I think that's also why you have professional mathematicians and labs in some serious degree of upset. Professional mathematicians are not wrong here: this is undermining their profession in ways that are quite serious. Some is superficial: Professional mathematics has used proofs as social capital

2972

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

I bet that's how you get to what happened today: a bulk release of significant results. Released to avoid that internal conflict. Results that tbqh probably deserve better than to be mass-dumped. But these aren't proofs to prove them. They're the exhaust fumes of the math eval factory.

312711

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

I think this is where those competing interests come back into play. One valid option is to just delete the results. Not your job. Not the core job. They'd already made their point. On the other hand, this is a big pile of results. Maybe someone can use them? Is sitting on frontier results right?

1662

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

In that case OAI kinda got burned. I think a lot of ways in which they and others handled that badly. But what to do now? They're sitting on a big pile of pretty major results. What do you even do with that? Mathematicians are mad at you for the whole NS fiasco but your job is not be a proof-machine

2583

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

But pumping the gas doesn't just solve famous problems that get newspaper headlines--like the Navier Stokes millennium prize problem--they solve a bunch of other problems collaterally too.

1530

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

So what do you do if you're a lab that has a model that can solve problems at that level? For a start you pump the gas on it a bit, to see how far it can go. Focus hard on a few of the very famous problems. Famous problems get attention, and, as the famous AI adage goes, attention is all you need

3903

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

But the models kept going, and eval runs now blow through some of those genuinely serious math problems at the frontier of the actual mathematical research community. So internal to the labs they end up with piles of genuinely serious math proofs that are (mostly) collateral output to the evals

1744

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

The problem is they kinda exhausted those; the models got smarter, and the new "hard problems" to eval against stopped being artificial test problems and started becoming just lists of famous unsolved problems

1730

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Historically, those evals were math problems that didn't really engage with the mathematical community directly. It's things like "solve this 4th grade homework problem", and then later "solve this math olympiad". Problems that were admittedly *impressive*, but not in any way *important* to math.

1611

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

That's not quite true of all the experts they hire tho. Some of those experts are professional and very good mathematicians in their own right. For those staff there's often a very genuine and sincere incentive about the math results directly, even if for the labs institutionally it's more abstract

2650

Pwnallthethings @pwnallthethings.bsky.social · 07/10/2026

Evals are hard problems with a success criteria, and in this case, a success criteria is *the same* as solving the problem. Findings proofs is basically a collateral output of running those evals at all. The labs I think institutionally care about the evals more than the proofs themselves

2901