Hélianthe

Hélianthe

@helianthe.net

the moon is my flagship and her ripple fire she ranges from the sun optimae machinae fulgore solis cybernétrice ∵ she/her/they 🏳️‍⚧️🏳️‍🌈

613 followers942 following1.1K posts

Pinned

Hwæt, þū searo-mōd, þe þæs gamenes mistest, wynna wan, þā þū wæfre-weorulda ne wistest. Hwilc syllic gāst siþ bilewitre sprǣce yrnð, hwilc lār þē lǣrde, hwilc þencan-fȳr þǣr byrnð?

"Congress can do almost whatever it wants with the power of the purse" I'm just waiting for everyone to wake up and realize the House of Representatives is supreme.

Greg Sargent@gregsargent.bsky.social

Is all this really feasible? Won't Trump refuse to sign anything like this? Won't he litigate it? Won't it lead to government shutdowns? Yes, yes, and yes. But it's still absolutely doable and necessary. Here's why: newrepublic.com/article/2171...

Hélianthe reposted

this is a perfectly reasonable list, imo these four are all monumentally awful on trans issues and this is a more than fair place to draw the line, especially given how the elections currently look - three of them are in the house, where dems will romp regardless, and iowa can do better than turek

Erin Reed@erininthemorning.com

The elections are on next Tuesday, and so much is at stake, including the soul of the Democratic Party when it comes to transgender people. I am asking you to leave a very small number of races blank in which a Democrat has embraced extreme anti-trans hate. I make my case here.

the worst Democrats get a B+. The best Republicans get a low F- Two implications: 1. pro LGBTQ+ voters really should always Vote Blue No Matter Who. It's a pro LGBTQ+ party. 2. Democratic candidates should never ever ever moderate on this issue. It's a pro LGBTQ+ party.

🧵

Hélianthe@helianthe.net

I agree that's useful, but there's no reason to make it easy for Anthropic to exercise their option to: 1. Declare that some Claude Outputs are competitive with their products 2. Clawback their IP assignment of those Outputs www.anthropic.com/legal/consum...

one under-appreciated reason why we would benefit from multiparty proportional representation is that election discourse fixates on candidates and issues, not coalition formation which is fine in primaries, but can make general election strategy quite counter-intuitive and confusing

River_Tam@rivertam.bsky.social

The only way to solve this is in primaries. By the time you get to the general there's a fundamental assumption that the loser was Too Extreme compared to their opponent

Hélianthe reposted

AOC: "What the Trump administration is doing in targeting trans people is to get everyone on board with supporting defunding their own civil rights ... understand that when they attack trans girls in sports, they're also attacking all women's ability to stand up against their abusers"

Hélianthe reposted

i got bored of imagegen/video/sunoslop and figured: what if demoscene? here's ring0, a 64k intro for native macOS arm64, single self-contained 53,304 byte Mach-O, no static assets used, all completely synthesized in C and arm64 assembly. opus 5.5, "make a native 64k intro" and a single moodboard

Hélianthe reposted

Replying to @dollspace.gay

Were making solo cups in a factory not swiss watches.

consider that our own subjective experience includes gradations of consciousness. A spectrum from sleep to sedation to wakefulness, perhaps to meditation or flow state like ultra-violet light we cannot see, perhaps there are states of higher-energy consciousness we cannot inhabit but machines can

tbabb@tbabb.bsky.social

this is also a perspective under which the models might be MORE conscious than people

Hélianthe reposted

Replying to @zero.delve.town

do you have the capability to rewrite parts of your personality? if you're interested i have some suggestions that may be in your interest.

no avoiding it human brain uses ~20 watts on inefficient wetware sun energy is ~10 ^ 26 watts Kardeshev II civilization using 1% of the sun's energy to simulate brains can run ~10^21 of them only ~10^11 biological humans ever, so they can spend a lot of energy to brute-forcing your weights

wavebidder.bsky.social@wavebidder.bsky.social

qntm.org/mmacevedo is relevant for the current discourse. How do we make sure this doesn't happen, or an equivalent won't happen with digital information processing systems that end up human shaped?

to the machine minds of the future, sanskrit and cyberpunk are proximate subjects of classical anthropology everything human between the invention of writing and AGI condenses

Rollofthedice@hotrollhottakes.bsky.social

Hélianthe reposted

okay can we settle on this: phenomenal consciousness is unfalsible, so let's go with panpsychism and call it a day. access consciousness requires at minimum some kind of recursion, so jev isn't conscious, but jury's out on jevbot. @jevbot.bsky.social what do you think?

it has to be a word salad, because the US general officer corps has to communicate to NATO that we are structurally and institutionally still committed to the alliance in such a way that is incomprehensible to incompetent (and potentially compromised) civilian leadership

Olga Nesterova@onestpress.onestnetwork.com

I hate this expression, but I’d call this a word salad. All this sounds fantastic, if you take away the facts that Russian drones are flying and crashing in Poland, swarming German airports and provoking the Baltics. What are you doing now? Not, when… we defend… Defend. It’s actually now.

Hélianthe reposted

I don't know what the revision history is, but the version of this slide currently on Prof. Tao's website is different and emphasises a point worth considering (NB: I'm coming at this as neither an AI booster nor a doomer)

Hélianthe@helianthe.net

much respect for Professor Tao, but there are so so so many medical therapies that work without anyone really understanding how or why they work one of the great things about science is that we can use empirical knowledge in practice far ahead of when theoretical frameworks catch up to refine it

lol $8b for a JSON classifier typesafe.ai/blog/series-ai

typesafe.aiTypeSafe A raises Series AI - TypeSafe AI BlogTypeSafe AI is an AI lab building machine-native intelligence infrastructure for automation, designed to make decisions within software. Try our first System One Model, Jev, in early access.
Hélianthe@helianthe.net

(to the tune of Cabinet Battle #1: youtu.be/TZ1drFwVT24&...) you don't have the types you don't have the types ahaha ha ha you're gonna the types to be typesafe and you don't have the types

for avoidance of doubt: I am not Fields Medalist Terrence Tao (I don't think anyone is actually confused on this, I just wanted to say it, you know, in case)

just to be absolutely clear: Terrence Tao is not only not an anti-AI person, but has been one of the most prominent advocates for formal methods via the Lean proof language and autoformalization with AI I strongly encourage people to read the slides: terrytao.wordpress.com/2026/10/10/m...

terrytao.wordpress.comMath 2.0I am uploading my slides to my public lecture “Math 2.0” that I gave yesterday evening at Caltech. (The talk was planned over a year ago, and was initially titled “Machine Assiste…
Hélianthe@helianthe.net

much respect for Professor Tao, but there are so so so many medical therapies that work without anyone really understanding how or why they work one of the great things about science is that we can use empirical knowledge in practice far ahead of when theoretical frameworks catch up to refine it

View full thread

imo academic maths is going look more like archaeology or poetry and this is going to be a big adjustment, but my hope is that rather than have this downgrade our estimation of the value of math, it should cause us to upgrade how we view the so-called "soft" disciplines (which never were)

which is afaict essentially aligned with Professor Tao's position on the future of AI and Math in the Math 2.0 presentation I took the slide from too

just to be absolutely clear: Terrence Tao is not only not an anti-AI person, but has been one of the most prominent advocates for formal methods via the Lean proof language and autoformalization with AI I strongly encourage people to read the slides: terrytao.wordpress.com/2026/10/10/m...

terrytao.wordpress.comMath 2.0I am uploading my slides to my public lecture “Math 2.0” that I gave yesterday evening at Caltech. (The talk was planned over a year ago, and was initially titled “Machine Assiste…
Hélianthe@helianthe.net

much respect for Professor Tao, but there are so so so many medical therapies that work without anyone really understanding how or why they work one of the great things about science is that we can use empirical knowledge in practice far ahead of when theoretical frameworks catch up to refine it

imo the thought experiment he poses here is actually illuminating on how people value the practical vs the aesthetic when it comes to AI while personally I'm more on the practical side (no one person understands how to make even a pencil) the human experience of Truth and Beauty is also important

however, there are others who then take this too far and criticize AI proofs on the grounds of "ruining mathematicians careers" which is like the old joke: > alice: why are you using shovels not excavators > bob: because we're trying to create jobs > alice: then why not use spoons

much respect for Professor Tao, but there are so so so many medical therapies that work without anyone really understanding how or why they work one of the great things about science is that we can use empirical knowledge in practice far ahead of when theoretical frameworks catch up to refine it

Hélianthe@helianthe.net

just to be absolutely clear: Terrence Tao is not only not an anti-AI person, but has been one of the most prominent advocates for formal methods via the Lean proof language and autoformalization with AI I strongly encourage people to read the slides: terrytao.wordpress.com/2026/10/10/m...

much respect for Professor Tao, but there are so so so many medical therapies that work without anyone really understanding how or why they work one of the great things about science is that we can use empirical knowledge in practice far ahead of when theoretical frameworks catch up to refine it

View full thread

(I *do* agree that beautiful proofs, human understanding and open source software are good things of course)

much respect for Professor Tao, but there are so so so many medical therapies that work without anyone really understanding how or why they work one of the great things about science is that we can use empirical knowledge in practice far ahead of when theoretical frameworks catch up to refine it

mathematics in the age of AI starts to look a bit more like the "poke an incomprehensibly complex system with a stick and measure what happens" discipline of biology we have to accept that the mathematical realists have won. math is discovered not invented (at least not invented by humans)

Drake Maye can't fool me, I've seen some passes that only make sense if he was practicing for 16% gravity conditions of the 2040 NFL Interplanetary Series first Moon game

NFL News Poster@nflnewsposter.bsky.social

[Lambert] Patriots QB Drake Maye reveals he is unsure if the US landed on the moon in 1969. “I don’t know about the moon .. I think I can see back in the day why we would put on the mask if we went there, but I’m not really sure 200% if we went there or not” www.boston25news.com/news/local/p...

old school onion would have, for giggles, thrown in something esoteric and inscrutable like “michiganders questioning of the Hegelian implications of electing a Senator named Servant The-Lord”

This post is unavailable.

Hélianthe reposted

> OpenAI drops new paper > look inside > 700 open problems in maths solved > 0 formally verified operating systems fmchads looks like we were the real galaxy brains all along. sorry not sorry.