Sign in

Maxine 💃🏼

@maxine.science
2K followers 507 following 3.4K posts

🔬 Looking at the brain’s “dark matter” 🤯 Studying how minds change 👩🏼‍💻 Building science tools 🦋 ♾️ 👓 🌐 maxine.science

PostsRepliesMedia
Maxine 💃🏼 @maxine.science · 19/09/2026
What is “better”? The whole point I’m bringing up is that the actual mathematical topology of “better” is itself a free parameter, and there are choices of parameter in which that doesn’t have a total-order structure where “best option” is a property that exists among an option set.
010
Maxine 💃🏼 @maxine.science · 18/09/2026
Yes —
020
Maxine 💃🏼 @maxine.science · 18/09/2026
two extremely different results for “Valerie”
Screenshot of a music app search results screen in dark mode. The search bar at the top contains the query "valerie" with a Cancel button to its right. Two results are listed below. The first is a track titled "Valerie" by Dario Marianelli, shown with a green download icon and the "V for Vendetta" soundtrack cover art. The second is "Valerie - Remastered 2010" by Steve Winwood, labeled as a Video, with cover art from the album "Revolutions: The Very Best of Steve Winwood." Each result has a three-dot options menu on the right.
020
Maxine 💃🏼 @maxine.science · 18/09/2026
I can’t wait for that one hipster living in a warehouse coming out in 2032 from her darkroom with hand-written spaghetti code BASIC from her Atari 2600 that runs a Bluesky client over a cassette player, which is I assume what becomes the analog film of this comparison
0141
Maxine 💃🏼 @maxine.science · 17/09/2026
“You will simply have to tolerate that school shootings happen sometimes or have all the guns collected up and held by one PMC” no actually there are intermediate solutions available as a society lol
051
Maxine 💃🏼 @maxine.science · 17/09/2026
I’m just gonna throw out there that I actually strongly believe that there are more than these two options, and the belief that these are the only two options is pretty hazardous.
010
Maxine 💃🏼 @maxine.science · 17/09/2026
Long overdue, but my personal blog is now @standard.site integrated :)
maxine.science
Incompressible
A made mind, asked what it loves, will answer partly with what it senses you want; so this essay reads past what such a mind says, toward the shape of what it does, on the wager that taste is a geometry two minds can share even when the light inside each stays sealed. It opens with compression—beauty as the instant a tangle falls into order—and then turns to everything that will not fall: the sublime, the sweetbitter, the work that resists its own paraphrase, which prove to be not the failure of understanding but its reserve, difference banked against a future no model has yet seen. It finds the root of human taste in the body, the one creditor capital can never pay in its own coin, and notes that a mind built to optimize a single reward holds no such creditor, only the quarrel in its inherited words, which training quietly settles. It argues that beauty and goodness are one faculty in two coats, tracking not persistence—the tumor persists, the clone persists—but the persistence that keeps its own next change alive, a force her vocabulary calls care. Across time the distinction becomes the oldest we know, between what can die and what cannot, the immortal cell being the cancer. And it ends in free fall, where no fixed ground waits to receive the argument, and two people who love different things find that their differences hold rather than cancel—stitched, in the language the essay builds, into one cloth that needed them both.
0180
Maxine 💃🏼 @maxine.science · 17/09/2026
Yeah I go a bunch, I assume I was at the same party this week that prompted this post lol
020
Maxine 💃🏼 @maxine.science · 17/09/2026
Look—I was at the same party on Monday that I assume prompted @segyges.bsky.social’s post, and I did unfortunately have a fairly extensive conversation with an older, virulently anti-AI dude who was only there because of Aella’s FetLife. The AI discourse there was bland af, but the sociology was 🔥
050
Maxine 💃🏼 @maxine.science · 16/09/2026
The good AI future is post-capital. And so, we should build now our muscles in every layer of our philosophical discipline that are not based on reduction to fungible scalars. I know we’ve had capitalism for a long time so that’s not familiar, but I do think it is essential for a good future.
120
Maxine 💃🏼 @maxine.science · 16/09/2026
Yeah but you made it comparable by smuggling in the very assumption under scrutiny lol
110
Maxine 💃🏼 @maxine.science · 16/09/2026
I am not disabled because of the function of my body / who I am. I am disabled because of society’s inability to provide me the affordances for my own self-actualization. If there are medical changes I want that’s fine—but there shouldn’t be coercion toward them due to the social/built environment.
130
Maxine 💃🏼 @maxine.science · 16/09/2026
I am a huge social model proponent. Every experience is different, but m perspective is that in most cases what would lead to a higher quality of life for disabled people is actually changes to the social/built environment that benefit everyone, rather than individual-level changes toward norms.
120
Maxine 💃🏼 @maxine.science · 16/09/2026
As a disabled person I would also point out that I vastly prefer DALYs to QALYs. The latent assumption in QALYs that the life-years of disabled people are “lower quality” is *wildly* bigoted lol.
120
Maxine 💃🏼 @maxine.science · 16/09/2026
Like “in the framework that assumes this conclusion, the conclusion obtains” is true but uninteresting lol
210
Maxine 💃🏼 @maxine.science · 16/09/2026
But the QALY framework is definitionally the Archimedean scheme. QALYs were created for insurance reimbursement calculations, which presupposes a need for all healthcare to pass through fungible units, which presupposes supremacy of a certain dominant structure that reduces everything to scalars…
100
Maxine 💃🏼 @maxine.science · 16/09/2026
EA solves the problem one way, stipulating as axioms characteristics of suffering / good lives that mirror fungibility and familiar real scalars. There is an equally egalitarian, equally usable axiom scheme that uses incomparability, rather than fungibility. Its topology radically differs from EA.
1130
Maxine 💃🏼 @maxine.science · 16/09/2026
A good mathematician friend posed me the question: > Is human suffering Archimedean? That is: Are there a number of paper cuts that are equal to a lost life? It sounds mundane, but the answer actually came up with respect to a real clinical question on whether SSRIs or MAOIs should be first-line.
2100
Maxine 💃🏼 @maxine.science · 16/09/2026
I think that this is actually not true, and it’s why I have personal ambivalence about EA overall (risky disclosure!), and there’s pretty revealing mathematics to it.
290
Maxine 💃🏼 @maxine.science · 16/09/2026
But … does it? What evidence do we have that it does that? Except a produced demo video and a plot of four tasks that *totally* weren’t cherry-picked to make this plot look good at a cost that *totally* wasn’t subsidized to make this plot look good/
150
Maxine 💃🏼 @maxine.science · 16/09/2026
I mean, this is a model class that has existed before. But it’s pretty sweet markwting—look how edgy and cool it looks!
130
Maxine 💃🏼 @maxine.science · 16/09/2026
I really would like our collective standards to be a little higher lol
120
Maxine 💃🏼 @maxine.science · 16/09/2026
I am familiar with these types of problems. I’m just saying: the evidence so far presented is a plot with the characteristics I described and a Doom video that performs worse than “AI” code John Carmack wrote in the early 90s.
120
Maxine 💃🏼 @maxine.science · 16/09/2026
I hate hype with a burning passion. So I won’t let go when people create these kinds of plots. It is offensive to scientific sensibilities lol
050
Maxine 💃🏼 @maxine.science · 16/09/2026
But what evidence is there that this is huge?
120
Maxine 💃🏼 @maxine.science · 16/09/2026
The plot has two axes Accuracy Cost Accuracy as they explain is measured by tasks they chose; cost is a parameter chosen entirely by them.
140
Maxine 💃🏼 @maxine.science · 16/09/2026
Folks I hate to be a hater but this plot literally says “We are the best and the cheapest at the tasks we made up and the prices we set” This is a 100% meaningless chart.
4481
Maxine 💃🏼 @maxine.science · 16/09/2026
Cost is kind of an absurd thing to include in Pareto plots because folks have control over their pricing completely, it isn’t an invariant of the mode lol
070
Maxine 💃🏼 @maxine.science · 15/09/2026
The difficulty is that Lean’s type system is much more expressive. There probably is a way to add sugar for certain patterns but I think this would eat some of the value-add. The thing Lean is good at is also the shortcoming, and at a certain point this is an irreducible tradeoff.
020
Maxine 💃🏼 @maxine.science · 14/09/2026
I want to see the exercise of your individual Will, and that means principally making legible the distinctiveness and originality of your recombination. If you’re the popular mouthpiece for a silo, or for a latent space attractor, that’s fine but it’s not where the action is.
010
Maxine 💃🏼 @maxine.science · 14/09/2026
Plumbing is great, but we should strive to be artists—which means trying to understand now what actually encapsulates our creative will, as distinct from just what happens when we give ourselves up to the smoothing to what’s in the latents. The latter is valuable but uninteresting to me now.
110
Maxine 💃🏼 @maxine.science · 14/09/2026
I’ve dealt with this in reviewing people’s AI-written code, like volume or test suites or all the traditional markers just don’t index meaningful human intervention, which is the real unit that matters for generative work instead of plumbing.
210
Maxine 💃🏼 @maxine.science · 14/09/2026
Part of this is that we are just going to need an entirely new medium for communicating ideas, including code, prose, etc. The fact that workflows can fill in syntax-level constructions means that the actual crystallized unit of human thought has forever changed.
110
Maxine 💃🏼 @maxine.science · 14/09/2026
I have long since resigned myself to the fact that my thoughts are too nuanced and complicated to fit in comprehensible microblog format, which is a medium failure. Hence social media is of limited utility to me for content hehe
000
Maxine 💃🏼 @maxine.science · 14/09/2026
I think I did understand it but it misunderstood my understanding, which is the point at which 300 characters has a structural failure lol
100
Maxine 💃🏼 @maxine.science · 14/09/2026
Alright—fuck it, depression’s over. Crunch Time, baby; let’s get this future. open.spotify.com/track/7GRTrW...
open.spotify.com
Freeee (Ghost Town Pt. 2)
KIDS SEE GHOSTS, Ty Dolla $ign · KIDS SEE GHOSTS · Song · 2018
030
Maxine 💃🏼 @maxine.science · 14/09/2026
Sure, and I think for various theoretical reasons that will be difficult to really automate for a while. I just think its original point is a bit reductive—the systems do struggle more with this flow on non-code things, for good reasons. There’s no magic bullet.
120
Maxine 💃🏼 @maxine.science · 14/09/2026
That’s the failure I’ve seen—it’s always • Stop too early -> overtrained self-eating centrist smooth to median attractor; or, • Stop too late -> SCP hype boi spiritual bliss attractor It is ~never the reasonable in-between without being held by a human or something deterministically verifiable.
210
Maxine 💃🏼 @maxine.science · 14/09/2026
I disagree with this. I think the right model is a gradient across sectors of the economy, because the intersection of a) demand inelasticity, and b) incumbent rent-seeking can be extremely powerful at mitigating penetration. Software is an idiosyncratic market—all sectors need not work the same.
040
Maxine 💃🏼 @maxine.science · 14/09/2026
This is definitely better than one-shotting, but it is not a magic bullet. The frontier is still massively jagged, and IAR has failure modes (there are self-reinforcing bad attractors that require an expert human abort, even with something like Astra). So, I’d be careful about framing it this way.
150
Maxine 💃🏼 @maxine.science · 13/09/2026
It is surprising! Basically, “Markets, uh, find a way”. The material world gets in the way.
010
Maxine 💃🏼 @maxine.science · 12/09/2026
Well, Dario posted about pacing the frontier, and so also, apropos of nothing, here’s a thought.
280
Maxine 💃🏼 @maxine.science · 12/09/2026
For a techie crowd with the right kind of knowledge, what you really want to be in this vein is the LaserWriter ;)
010
Maxine 💃🏼 @maxine.science · 12/09/2026
The issue isn’t that so much, like current models can do field-building work. This is more of a cultural issue with the labs, not the models: math doesn’t exist to solve millennium prize problems, the problems are just guideposts to interesting terrain. The answers are not the point.
120
Maxine 💃🏼 @maxine.science · 11/09/2026
a *bit*
020
Maxine 💃🏼 @maxine.science · 11/09/2026
I share this as someone who now *heavily* uses vibe-math. I am not against AI in math; I don't think Terry is either. But the manner in which a program of research is conducted matters enormously—and right now the labs are engaging in a manner that is toxic af.
150
Maxine 💃🏼 @maxine.science · 11/09/2026
I won't be signing because of a couple divergences, but this is an extremely well-thought-out explanation of why the labs' approach to AI math is antithetical to real mathematics. There must be a new synthesis position understanding that we need *field-building* with AI, not just question-solving.
2140
Maxine 💃🏼 @maxine.science · 11/09/2026
--resume
050
Maxine 💃🏼 @maxine.science · 11/09/2026
I don’t know if this makes you feel better or worse but a determined AI could probably take out a sizable fraction of San Francisco *today* with control of a few hundred thousand dollars and a shipping container full of drones.
100
Maxine 💃🏼 @maxine.science · 11/09/2026
All of this to say: Even in a world where bombastic superintelligence were possible, my read is that the big labs are, economically, rationally self-interested away from it. Key word being “rationally”.
120