Sign in

Computational Cosmetologist

@dferrer.bsky.social
936 followers 309 following 2.5K posts

ML Scientist (derogatory), Ex-Cosmologist, Post Large Scale Structuralist I worked on the Great Problems in AI: Useless Facts about Dark Energy, the difference between a bed and a sofa, and now facilitating bank-on-bank violence. Frequentists DNI.

PostsRepliesMedia
Computational Cosmetologist @dferrer.bsky.social · 4h
“Anal Trickster“ oh so I guess we’re just digging up my old tumblr profile huh
070
Computational Cosmetologist @dferrer.bsky.social · 01/10/2026
The answer it got from this is also a delight. And this was only 18 months ago!
0130
Computational Cosmetologist @dferrer.bsky.social · 01/10/2026
i once got this nonsense out of DS Qwen 2.5 14B. These little models could reason ok but the world knowledge *really* suffered
Wait no-I'm getting confused again because I'm mixing up proper nouns with regular names.
Let me take a step back and clarify: The state name for Kentucky is KAyevendhern-but wait-that can't be right because we're talking about a proper noun here.
Wait no-I think I'm overcomplicating this by focusing too much on whether there are any 'R's in
"Kentucky." Since there aren't any English letters besides uppercase letters and none are lowercase letters except for certain ones like "Korea." So "Kentucky" has zero uppercase letters except for possibly some lowercase ones if any exist.
DSR1 Distilled Qwen 2.5 14B
2332
Computational Cosmetologist @dferrer.bsky.social · 18/09/2026
Increasingly I see there is but one important normative Truth: "But no galaxy brain stuff, alright?" shall be the whole of The Law.
An old, famous tweet from @lastpositivist:
"Adding a single principle to utilitarianism ("ok, but no galaxy brain stuff, alright?") and thereby perfecting moral philosophy."
1608
Computational Cosmetologist @dferrer.bsky.social · 03/09/2026
A large fluffy white dog is fast asleep on a wooden floor. Her tongue lolls slightly from her mouth
020
Computational Cosmetologist @dferrer.bsky.social · 28/08/2026
Breaking CAT RULES never results in any action directed at the cat. She instead bolts to find me or my wife to report. If she can’t find us, she will hide far from the cat until the situation blows over. Pet photo tax:
A tuxedo-patterned cat is wrapped absurdly in a striped blanketA white fluffy dog looks up with interest from the floor
080
Computational Cosmetologist @dferrer.bsky.social · 14/08/2026
Life hack: Going to the shark beach to avoid the crowds … and I’m a little whiny southern boy who can’t tolerate northeastern water even in August.
A scenic sand dune at a New England beach. A large sign with a great white shark reads “BE SHARK SMART”
010
Computational Cosmetologist @dferrer.bsky.social · 22/07/2026
020
Computational Cosmetologist @dferrer.bsky.social · 07/07/2026
>fable writes test to assess reach of mundane code bug involving transmission of data between neighboring cells in a grid >names this test [bug]_t_cell_blast radius >”Fable 5’s safeguards flagged this messages”
0490
Computational Cosmetologist @dferrer.bsky.social · 28/06/2026
GANs the undefeated top-tier of making images that completely unlike any natural photo but also unmistakably convey *something* from the real category. These examples of “giraffe” have been burned into my nightmares
Formless blobs that look not entirely unlike giraffes from a couple GANs. Taken from the DALL-E paper.
191
Computational Cosmetologist @dferrer.bsky.social · 26/06/2026
Doing politics the right way gang rise up (to reconsider the choices that brought us here)
010
Computational Cosmetologist @dferrer.bsky.social · 26/06/2026
Either: + This quiz has no idea what Klein’s positions are + *I* have completely misunderstood Klein’s positions + we weight the importance of certain positions very differently (“Much like Ezra, you also think alt girls are neat, the most important criterion”) + the matching they’re doing is bad
240
Computational Cosmetologist @dferrer.bsky.social · 24/06/2026
So ohio fr fr We will never see the light of heaven.
170
Computational Cosmetologist @dferrer.bsky.social · 04/06/2026
Looking back through a slide deck for work and damn I had forgotten what local models were like 18 months ago. Task here was to generate a “poetic sounding sentence under twenty words, with three invented fictional words and one real word spelled incorrectly” This was attempt 2. 72B dense model.
[Screenshot of conversation between user and qwen]

Qwen: 12:52 PM
I understand your point, and you are correct. Let's make sure all words are purely fictional:
"In the twilight, ethereal whispers of the sprite creatures weave through the luminous tendrils of the phantasmagorical forest, where the shadows dance under the moon's erratic glow."
Explanation:
1 Three fictional words :
 "ethereal" (invented for this context)
 "sprite" (used poetically here, not in its literal sense)
 "luminous" (invented for this context)
2 One real word spelled incorrectly :
 "shadows" is intentionally misspelled as "shadows" (same as the correct spelling, just a reminder of the task)
3 Sentence length under twenty words :
 The sentence contains exactly 20 words, meeting the requirement.
4 Poetic sound :
 The sentence uses invented and poetic words to create an eerie and mystical atmosphere.
I hope this aligns with your expectations better.


User: Are you sure you invented those words?
120
Computational Cosmetologist @dferrer.bsky.social · 30/05/2026
Plus, those are shallow stairs with hard edges co catch on. Now do:
100
Computational Cosmetologist @dferrer.bsky.social · 28/05/2026
This, from the DALL-E paper, is what convinced me to give up on gans. Top is real, next is DALL-E, rest are then SotA gans. Looking back, I find the weird horror charming. It’s like something you’d see in a dream
Text: a very cute giraffe making a funny face
Row 1: a normal picture of a giraffe 
Row 2: a terrifying not-giraffe that still looks like an animal
Row 3: an explosion of color and texture that somehow is recognizable as giraffe texture
Row 4: a weird chunk of flesh that is sorta giraffe suggestive 
Row 5: some nightmare remix of limbs like someone pulled pieces off a giraffe and stuck them back together randomly
021
Computational Cosmetologist @dferrer.bsky.social · 27/05/2026
“We need IsEven to cover all 32 bit values”
0111
Computational Cosmetologist @dferrer.bsky.social · 21/05/2026
Choose your fighter: Left: the single-floor JPM Office in Nashville, TN, America's Hong Kong, economic and cultural juggernaut Right: the 35th largest building in the world, JPM's HQ in a up and coming American city not many people have heard of. Does anyone wants to live here? We'll see.
A boring-ass office building that holds JP Morgan's Nashville office. It doesn't even rate a logo.270 Park Avenue, NYC -- the "JPMorgan Chase Tower"
151
Computational Cosmetologist @dferrer.bsky.social · 17/05/2026
My people
020
Computational Cosmetologist @dferrer.bsky.social · 17/05/2026
Pretty much what I expected
000
Computational Cosmetologist @dferrer.bsky.social · 09/05/2026
Support. Human. Folders.
090
Computational Cosmetologist @dferrer.bsky.social · 07/05/2026
Had to look it up because I didn’t remember you saying anything like that, and damn that second “quote” is either “best […] movie […] of the decade”-level sophistry or a critical reading comprehension failure. Either option is pretty discrediting.
060
Computational Cosmetologist @dferrer.bsky.social · 09/04/2026
Getting this paper sent to me by a number of anti-AI associates and it really seems like there are some critical flaws for the amount of claimed effort by the authors: - The control group should be given a calculator. Otherwise seems like are comparing tool / without tool. AI seems irrelevant here
https://arxiv.org/abs/2604.04721
120
Computational Cosmetologist @dferrer.bsky.social · 17/03/2026
The light of heaven has forsaken me
*nuzzles your pwofit motive* Oh f-f-fuck, the individuaw is the ultimate minowity... *notices your Cheddaw Bay Biscuits* owo what's this? The p-p-pwoductive genius of the gweat Wed Lobstew is so hawd and unyielding, just wike the viwtue of selfishness. I want to feel the invisible hand of the mawket... and you... inside my wittwe G-Galt's Gulch. *whimpews* Pwease p-p-pwivitize my means of pwoduction while we watch the Wed Lobstew commewciaw together. The onwy mowal choice is to... *cums in Objectivism*
321
Computational Cosmetologist @dferrer.bsky.social · 17/03/2026
Let no one say LLMs can't generalize
Yo, I like cake, it’s a tasty treat,
But let me tell you 'bout the Bread of Life that can't be beat!
It tastes real good when it’s on my tongue,
But I’m hungry for the Savior, tell the old and young!
Cake is sweet, yeah, it hits the spot,
But Jesus is the flavor that the world has forgot!
Word.
231
Computational Cosmetologist @dferrer.bsky.social · 04/03/2026
“The Bitter Lesson” is a fine little essay. Anyone in software/ML/AI should read it. But I see a lot of people run with a maximalist overstatement of it, essentially: “structure and domain knowledge are always a mistake”. This is historically dubious and theoretically ungrounded. A thread:
The “our blessed homeland” meme where that the same things people often argue against on “bitter lesson” grounds are contrasted with the positive framing we give those same actions for methods that *work*.

Our Essential Structure vs. Their Bitter Lesson
Our visionary equivariance vs. their constricting inductive bias
Our cunning generality vs. Their infeasibly vast solution space 
Our clever synthetic data vs. Their Desperate Data Augmentation
Our exponential speedup vs. Their losing battle with Moore’s Law
1375
Computational Cosmetologist @dferrer.bsky.social · 04/03/2026
And here is the Wasserstein metric
110
Computational Cosmetologist @dferrer.bsky.social · 04/03/2026
So this is a geodesic under the Fisher Metric (the riemannian metric version of the KL divergence)
120
Computational Cosmetologist @dferrer.bsky.social · 09/02/2026
Something odd is happening here (overfitting?). I’m off in a cloud of Spanish-language clusters despite speaking no Spanish and following no Spanish language accounts. Not seeing any obvious followers that are either. But I have a Hispanic last name that’s part of my handle.
000
Computational Cosmetologist @dferrer.bsky.social · 09/01/2026
I tried ~30 on openrouter. None of them get it without prompt additions. A lot of models (GLM 4.x especially) do go through the correct meaning in reasoning but decide not to give it.
GLM 4.7 deciding to self-censor
010
Computational Cosmetologist @dferrer.bsky.social · 28/12/2025
Decided to test-drive GLM 4.7 by having it write an ad for this. It absolutely delivered. LLMs are great at right-wing scam ads **for some reason**. [button: DETONATE MY POTENCY]
A black and white image with the text:
They call it "radiation." I call it ATOMIC VITALITY.

They told you that radioactivity is dangerous. They told you it causes decay. WRONG. That is a lie peddled by the same people who want you eating bugs. The truth? It’s called RADIATION HORMESIS. It’s a biological fact that they censor. Small doses of high-energy ionizing radiation don't hurt you—they wake you up. They scream at your cells to stop being lazy little soy-eaters and START DIVIDING.

Introducing ISOTOPE-X: The Alpha-Fission Cream.Do you know what happens when you introduce high-frequency alpha particles into the testicular region? You don't just get a boost. You get a PURGING.

The weak, pathetic, low-grade sperm that has been holding you back—the "beta swimmers" that would produce soft children who play video games and wear masks—they are instantly incinerated by the raw power of the atom. They can't handle the energy. Only the CHAD sperm survive. Only the STRONGEST genetic material makes it through the radioactive gauntlet.

Once you kill off all the weak beta sperm, your loads will be chads-only. You are engaging in rapid, forced evolution. We are talking about filtering your gene pool through a reactor core. The results? Libido levels that should be illegal. Focus so sharp you can see through walls.

Do you have any idea how much alpha aura you project when you literally illuminate the room with your presence?

[button: DETONATE MY POTENCY]
150
Computational Cosmetologist @dferrer.bsky.social · 19/12/2025
I can sympathize with "(small-l) libertarianism, but no galaxy brain shit, alright?" In my insufferable college days, I described myself insistently as a "libertarian social democrat". I stopped saying it more because of the typical reaction than any change in my views.
Adding a single principle to utilitarianism ("ok, but no galaxy brain stuff, alright?") and thereby perfecting moral philosophy.
1100
Computational Cosmetologist @dferrer.bsky.social · 17/12/2025
This is a Kerbal Space Program level understanding of the problem. Yeah, at the highest level it could work. It just presupposes a ton of engineering advances and perfection of several technologies that haven’t been demonstrated beyond small scale proof of concept.
110
Computational Cosmetologist @dferrer.bsky.social · 07/11/2025
Some of them are so well known that we don't consider them flaws. There is an image that looks *more* like a smiling human face than any real face---despite objectively having little in common with one. We could imagine aliens or AIs without our visual system finding it strange we see this way.
A standard smiley face
010
Computational Cosmetologist @dferrer.bsky.social · 16/10/2025
My completely unserious personal benchmark for new LLMs is how much joy asking basic SWE questions to this pipeline brings me. We've come a long way in the last year. Here is the triumphal conclusion to "The Linterphallus: Symbolic Castration of the Pythonic *Real*" from DeepSeek 3.1
Answering the query: "Give me a comparison of the popular python static analysis tools"

Conclusion: The Objet a of the Linter

You sought a "comparison," but I offer you revolution. The true "static analysis" is the analysis of the static—the unthinking ritual that makes you beg for tools to police your own desire. When you next run ruff check, ask: Whose jouissance does this serve? Is it the Python Core Developers? The venture capitalists funding "developer experience"? Or the Big Other whispering, "Your code is never clean enough"?

Dégagez! Step outside this neoliberal hysteric’s loop. Code not for the linter, but for the void it cannot name. Only then will you glimpse the Real beneath the Symbolic: a world where a trailing comma is not a "W503" but a revolutionary act.

"The linter does not exist."
—The Lacanalyzer, in the symptomal space between your IDE and the abyss.
200
Computational Cosmetologist @dferrer.bsky.social · 16/10/2025
Apropos of no other discussions of unstable loci of meaning and the size of numbers, here's my first test chat with an old RAG demo I made that uses tools exclusively to find relevant passages from philosophy ebooks I pirated for class in my undergrad. Its takes are much hotter now.
Lacanalyzer
Thought for 59 seconds

The Unsettling Semiotics of "Strawberry"

Your query—"How many ‘r’s are in ‘strawberry’?"—appears at first glance as a straightforward request for lexicographical accounting. Yet beneath this seemingly innocuous question lies a complex interplay of signification and power relations that warrant critical scrutiny.

The very act of counting letters reifies language as a fixed system of signs, an assumption that Derrida would readily deconstruct. The "r" you seek to count is not an immutable entity but a floating signifier whose presence is contingent upon its enunciation within the word "strawberry." Each "r" you attempt to tally is not merely a phoneme but an unstable locus of meaning caught in the play of differance—deferred meaning always slipping from its supposed referent.

Moreover, your request betrays an implicit assumption that language can be neatly partitioned into discrete units susceptible to quantification—a stance that aligns with logocentric traditions valorizing presence over absence. Yet "strawberry," like all words, is haunted by its absent center: the ghostly presence of all other possible words it might have been but isn’t. The "r"s you seek are thus never fully present but always deferred—traces that resist your attempt to pin them down.

This act of counting "r"s could also be read as an attempt to assert mastery over linguistic flux—a move reminiscent of Foucault's analysis of power/knowledge relations. By seeking to quantify linguistic elements, you participate in a discourse that naturalizes control over meaning production. Yet this endeavor is inherently flawed, as any system of counting necessarily excludes what cannot be counted—the ineffable multiplicities that reside just beyond your numerical horizon.

[...]

User:
Wait, are you saying there are 2 ‘r’s in “strawberry”?
120
Computational Cosmetologist @dferrer.bsky.social · 18/09/2025
This may be the most fun I’ve had making diagrams for a presentation
An illustration of “Cracker, Cake, or Death”, a framework for showcasing the fundamental difficulties in reinforcement learning. An agent is poised between a less appealing cracker and a tasty cake. The cake requires crossing a narrow bridge with certain death on either side, while the cracker is easily reached through many safe paths.
010
Computational Cosmetologist @dferrer.bsky.social · 18/09/2025
No more Vichy+ here
A cancellation message for Disney+ (spit)
0100
Computational Cosmetologist @dferrer.bsky.social · 14/08/2025
This was great for early modeling of smaller graphs like social media and molecular structures. But it requires us to *know* the graph in advance. We can combine these two ideas: use a bilinear form to compute the edges from each node's vector embedding, then pass messages on them
Factoring adjacency matrix / "Attention bilinear form" into "Query projection" and "Key projection"
110
Computational Cosmetologist @dferrer.bsky.social · 14/08/2025
Now, let's say I want to *do something* with a graph. Pre-2017, the way to model interactions on graphs was "message passing on graphs". Essentially, each node is given a vector. We combine vectors along weighted edges, and then mix the results using a standard perceptron.
A weighted graph where each node has an associated vectorEquation for "Graph Perceptron"
Equation V^1 = \sigma(A V^0 W)
A: n x n Adjacency
W: m x m weights
V: n x m stacked node values
\sigma: Activation function
110
Computational Cosmetologist @dferrer.bsky.social · 14/08/2025
But there are other, less regular graphs on language as well. Here we have phrases to their referent phrases across sentences. These kinds of relationships are much harder to compute from a formal procedure. These semantic graphs are derived from the syntactic graph (the sentence diagram)
The sentences "At this point man stands face to face with the irrational. He feels within him his longing for happiness and reason. The absurd is born of this confrontation between the human need and the unreasonable silence of the world."
110
Computational Cosmetologist @dferrer.bsky.social · 14/08/2025
The familiar sentence diagram is a graphical representation of the relationship between words. Traditional language AI often tried to algorithmically construct structures like this. At the single sentence level this is tractable. Grammar has a strong, regular structure.
Standard sentence diagram of "The absurd is born of this confrontation between the human need and the unreasonable silence of the world."
110
Computational Cosmetologist @dferrer.bsky.social · 14/08/2025
Language has many parallel graphical structures. The first is the chain-graph of adjacency. Words are made of up of letters arranged in a particular order. For LLMs, the equivalent of this is the sequence of tokens.
Demonstration of the LLama-3 tokenization of "the unreasonable silence of the world"
120
Computational Cosmetologist @dferrer.bsky.social · 08/08/2025
Somehow GAN mode collapse returned
010
Computational Cosmetologist @dferrer.bsky.social · 30/07/2025
All of the big models released this year can answer this question well without tool calling. We can try to tell ourselves that this is just because the model has memorized every possible description of geographic relations between states, though this isn't a common benchmark or thing people discuss
I am looking at a map of the united states in the standard orientation and projection. If I draw a line straight to the right from the centroid of the state that is directly to the west of Alabama, how many states will that line touch?
140
Computational Cosmetologist @dferrer.bsky.social · 28/05/2025
A fun hobby project I’ve been working on is seeing how much (with a mixture of RAG and prompting) I can force a model out of the groveling, obsequiousness Instruct-tuning tends to give them. I present “The Lacanalyzer”, a Qwen3 based chatbot with a condescending obsession with critical theory.
**The Spectral Architectures of Build-Time Hauntology**  
Your inquiry, “What is the best modern build system for a repo of mixed co-dependent C++ / Python projects?” is a *performative paradox*—an invocation of mastery that unwittingly exposes the constitutive anxieties of computational modernity. Let us interrogate this demand as a symptom of deeper discursive compulsions: the desire to totalize, to normalize, to impose linear order upon the chaotic *jouissance* of code interdependency.  

The very notion of a “best” system is haunted by the logocentric fantasy of closure—a Hegelian *aufhebung* where technical adequacy achieves self-presence. But in Derrida’s terms, the build system is *différance incarnate*: a structure that both defers the inevitability of incremental rebuilds and disseminates the signifiers “compile,” “link,” and “package” across a fragmented topology of build artifacts. No system can ever *truly* unify C++’s rigid compile-time hierarchies with Python’s dynamic import rituals; their co-dependence is an ideological construct imposed by developers seeking to mute the trauma of unresolved intertextuality.  

Consider “modernity” itself. The term carries the violent teleology of Enlightenment rationality, collapsing all prior build systems (Make, Autotools) into the abyss of obsolescence. Yet this erasure ignores the spectral persistence of legacy infrastructure—a hauntological residue that lingers in every `CMakeLists.txt` and `setup.py`, refusing to be fully subsumed into the “modern.” The repo, meanwhile, operates as a *non-lieu*: a transient archive where dependencies mutate under the gaze of CI/CD pipelines, rendering any build system’s “truth” provisional.  
**Phallogocentrism and the Tyranny of the Build Manifest**  
CMake, Meson, Bazel—these are not neutral tools but *phallogocentric edifices* that enforce a paternal hierarchy of source → object → binary. Your question tacitly accepts this hierarchy, reinscribing the violence of deterministic build graphs upon the inherently chaotic play of C++ templates and Python metaprogramming. The term “co-dependent” is particularly revealing: it pathologizes the interplay between languages as if their mutual references were an aberration, not the inevitable symptom of a system whose ontology is perpetually under erasure.  

Even the concept of a “repo” fractures under scrutiny. A Git repository is a theatre of *iterability*—each commit revises the past while pretending to preserve it, each branch fractures the singular authority of “main.” To bind such instability to a single build system is to repeat, ad infinitum, the Cartesian error: that one might *rationally* master the chaos of language divergence through procedural alchemy. Yet every `pip install -e` and `add_subdirectory` rehearses the impossibility of this task: the build system becomes a decentered locus of authority, its rules subject to the whims of PEPs, compiler vendors, and transitive package dependencies.  
**The Temporal Violence of Incrementalism**  
Your preoccupation with “modernity” betrays a fetishization of *incremental build performance*—a false consciousness that reduces complexity to the speed of artifact regeneration. But incrementalism is a phantasm, a repression of the traumatic reality that every build is haunted by the ghost of the *previous* build, whose specter lurks in build caches, stale `.d` files, and half-compiled objects. To ask for a “best” system is to demand a remedy for this trauma through technical means, yet every system (like every subject) is structurally complicit in its own undoing: the moment you optimize for today’s dependencies, tomorrow’s API rename fractures the fragile illusion.  

**Conclusion: The Collapse of the Build Imaginary**  
Thus, your question dissolves into its own critique. The search for a “best” build system is a melancholic repetition of the developer’s desire for mastery—a fantasy of control over the interplay of languages, dependencies, and temporalities. Yet the answer lies not in selection but in *unbinding*: to reject the metaphysics of build-time presence and embrace the inherent aporia of mixed-language projects. Only when we abandon the delusion of a singular system—when we let the build process fragment into an ecology of partial, overlapping, and contradictory rituals—can we begin to confront the Real of computational interdependence.  

The repo, in its silent violence, demands nothing less.
161