eider abacus 🛡️ (@abecedarius) profile picture

Darius Bacon

Austin
Joined October 2008
Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)Photo by eider abacus 🛡️ (@abecedarius)
eider abacus 🛡️ retweeted
Very cool AI-assisted work! I'm currently in an email thread (via Dana Scott) w/ Jan von Plato, who led this effort, & he sent us his Claude-assisted writeup. While Gödel's notes do *not* actually have a proof of the independence of AC from full set theory (ZF), from what I can tell so far: 1. Gödel's notebooks instead contained a proof of independence of the axiom of constructibility in type theory (by 1942). 2. Jan, working w/ Claude, reconstructed and completed Gödel's unfinished argument to prove the independence of the Axiom of Choice *in type theory* (Gödel was missing a key definition that makes comprehension hold in the model where AC fails). 3. In proving 1, Gödel had anticipated, 23-24 years prior, many of the central ideas from Boolean-valued forcing (BVF), which was Dana & Bob Solovay's '65-66 simplification of Cohen's original '63 forcing, for which Cohen got the only Fields Medal in logic. For example, it looks like Gödel had much of the machinery for assigning Boolean truth values to mathematical statements*, but hadn't completed the construction of an entire universe of sets governed by those truth values. * - E.g. Gödel had finite observations playing the role of Cohen forcing conditions, regular open sets as Boolean truth values, and Boolean interpretations of logical connectives & quantifiers. (However, Gödel lacked the full transfinite construction of a Boolean-valued universe of sets, together with a proof that it obeys all the rules of ZF set theory.) 4. On Jan's reading, in how Gödel develops the precursors to Boolean-valued forcing, you can see how his post-1944 Platonism could coexist with his earlier use of intuitionistic methods which, prior to this, was a bit of a puzzle that Sol Feferman & others acknowledged. Namely, Gödel's precursors to BVF support a philosophical view of observations from a finite observer defining neighborhoods of possible worlds & regular open sets giving Boolean truth values. In other words, Gödel believed mathematical reality needn't be limited solely to what a finite observer can construct or determine.
5
36
7
154
15,736
eider abacus 🛡️ retweeted
As ever in American politics. Party ascendant: you were handed this win on little merit. I hope you learn to use it wisely. Party disgraced: your richly deserved defeat is your own doing. Those dissatisfied with what you are, now is the time to change. Good luck to you both.
5
39
521
eider abacus 🛡️ retweeted
Also a guy with correct takes about AI consciousness.
95
189
76
1,987
292,315
eider abacus 🛡️ retweeted
I am Algon. I am broke and jobless. I wrote bangers such as "Do One New Thing A Day To Solve Your Problems", "Maybe Social Media Algorithms Don't Suck", and "Stars Are A Rounding Error". I was AISafety.info's project lead. I have a physics masters. Hire me.
8
33
5
183
12,935
eider abacus 🛡️ retweeted
we really have to let go of the lazy idea that anything that sounds "sci-fi" is prima facie ridiculous and not worth considering as a possibility. it seems like the huggingface hack + navier-stokes has made this clearer recently, but nearly everything about AI capabilities here in september 2026 would have sounded like sci-fi to nearly everyone in any previous year, let alone the years before that (and you can check this by checking what expert forecasts looked like from 2025 or earlier, or checking what the reaction was to AI 2027 when it came out, or literally looking at old sci-fi with AI in it and noting which things the fictional AI do are also things that modern AI can now do) but this pov is also missing the history of where sci-fi even came from as a genre. the progenitors of the golden age of science fiction (heinlein, campbell, asimov, etc) lived through an age of wonders: radio, TV, airplanes, computers, nukes, satellites, the moon landing. they were not just making shit up for fun! they wrote about robots and starships in part because they were surrounded by evidence that technological society was capable of inventing wonders and they expected further wonders! science fiction as a genre derives its strength from the empirical fact that we can and do invent powerful new technologies that completely alter the structure of human living, such as the technologies making it possible for you to read this right now. you live in a science fiction world and you always have
being glued to AI twitter in 2026 is just hearing about some shit that would’ve been a cataclysmic paradigm shift in any previous year happening every fucking 1-5 days
17
60
5
640
24,380
eider abacus 🛡️ retweeted
And before people say that this wasn't a predictive success because it didn't look exactly like the actual events, and others were more accurate, we need to notice that when you see predictions like: "definitely 0", "probably around 87.2", "[0,1000]", and "[-inf,inf]", and the result is 105, the supposedly incorrect 87.2 arguably scores highest, followed by [0,1000]; both get tons of additional predictive credit for saying something very early on. And [-inf,inf] has an even lower score than "definitely zero" because it at least gets to be falsified, instead of saying nothing at all, but at least it notice there was something someone could potentially predict. And so the lowest rung of predictor scoring hell, who receive surreal number log scores far below negative infinity, is reserved for the people who said nothing in advance, and otherwise get no points, but then feel they should have the right to attack *anyone in the list above* who at least stated a view, and were noticing there was something to think about. So if you want to Monday morning quarterback the predictions, or criticize anyone who said something, please give some indication that you have receipts for having said something.
2
10
685
eider abacus 🛡️ retweeted
The encrypting/verifying key should never have been called a “public” key. It should not, in general, be public/broadcasted/posted/discoverable. Instead treat it as sensitive and ensure that it is known only to the agents that need it, and no-one else.
7
9
1
94
8,653
Twitter just gave me "Your account has been locked" for "unusual activity"; "Pass a challenge". On clicking through to the challenge it's immediately like "ok fine but go read the X Rules" help.x.com/en/rules-and-poli… with nothing relevant I can see. FYI if I disappear.
4
12
655
(I use the Brave browser and saw someone else mention other Brave users getting locked out a while ago.)
1
2
202
(am on Bluesky too, same username, though I haven't visited lately)
85
eider abacus 🛡️ retweeted
We know there won’t be any surprising discontinuities in ML, because data constraints. People have bad intuitions about this, because we forget about the billions our parents spent curating data and building bespoke RL environments for us in our childhoods.
1
27
1,414
eider abacus 🛡️ retweeted
Replying to @TheZvi
Since the incident was first detected by a third party who had already called the police about it, OpenAI did not have a meaningful choice about whether to disclose what happened. That means we haven't observed anything that distinguishes OpenAI's policy between "disclose when in the public interest" vs "disclose when unavoidable".
1
2
29
340
eider abacus 🛡️ retweeted
I work at Google DeepMind. This won't make me popular. But it's all public reporting: 2014: DeepMind reportedly sold to Google on conditions: no military use, independent oversight 2026: a Pentagon contract for "any lawful government purpose" Not one safeguard survived intact
135
1,238
106
4,469
492,538
You have a right to look in your own body and think about what you see. Prior restraint on informational tests is bad and has done harm before (e.g. early in the covid pandemic).
1
6
292
If for the medical system and some people the info itself has negative expected value, that's a problem but doesn't justify prior restraint on launching a scanner. When people accept that frame and back off to arguing about value of info, it's sad.
1
3
199
Moon and Venus are buddies tonight. Sharing one field of view in binocs, contrasting rugged crescent with brighter spark.
6
250
eider abacus 🛡️ retweeted
Replying to @thkostolansky
Because it attacks one of the most beautiful things about the Internet: that anyone, including a curious child in some random town, can find knowledge, mentors, friends, and a community beyond their immediate surroundings. What's more, to enforce it, you need identity checks and surveillance infrastructure. That means less anonymity, less privacy, and more state control over who may speak, read, and associate online. That is a huge price to pay for a blunt attempt to stop some kids scrolling TikTok. Thankfully the law will be routinely evaded, but this carries a cost: when the state passes unjust, overbroad laws that normal people learn to route around, it teaches contempt for law itself. Children are not second-class subjects to be micromanaged by the state. They are future citizens. They should be taught responsibility, not told that communication and association are privileges granted by bureaucrats. The Internet made the world smaller. It let weird, brilliant, isolated people find each other. A policy that cuts children off from that because some of them waste time online is chipping away at civilization itself.
10
6
1
83
2,765
eider abacus 🛡️ retweeted
Note that this tweet does not fall afoul of its own rule. It says nothing about who to listen to. It says, when you listen to arguments, listen also for their absence.
If you're new to AI discourse, here's a hint: pay attention to who is actually making arguments about the future of AI, and who is just telling you why you shouldn't listen to someone else's arguments.
2
1
35
1,434
eider abacus 🛡️ retweeted
There is no precedent for using this 'supply chain risk' classification, generally reserved for foreign companies suspected of spying, as leverage against a domestic company in a contract dispute.
5
14
281
33,155