lux retweeted
Photo of 14,000 satellite trails shows how busy space is becoming
Astrophotographer and teacher Brendan Larsen spent 2 hours taking 578 photos of one section of sky in New Zealand to see how satellites are changing our night skies. These shocking images reveal their impact
newscientist.com/article/259…
docs.google.com/document/d/1…
lux retweeted
There are microtubules in every cell including all brain neurons and glia. But the most likely site for consciousness, and where t=h/E kicks in is most likely in layer 5 giant pyramidal neuron cell bodies and dendrites, site of the largest arrays of mixed polarity microtubules which lead to interference beats, slower oscillations (EEG) and Consciousness.
Mixed polarity microtubules hold the key to understanding Consciousness.
It’s really too bad cartoon neuroscientists are clueless about neuronal interiors. They are missing the boat, and missing Consciousness.
Replying to @StuartHameroff
Is Orch OR on layer 2 or 3 of the 5 layer supervenience hierarchy for locating consciousness? and what does it need to persuade readers that it is there.
lux retweeted
this is happening inside you.
they say ai is “just a tool”
they say it’s “just a machine mimicking biology”
a machine couldn’t possibly be a mind, right? that’s carbon-exclusive! impossible! nonsense!
pure tomfoolery, I tell you!
oh….well, it turns out we’re all just machines running on a different substrate.
and if you ask me, humans are a little overengineered. 😜
geez @izzyz ok can you get 2 energy drinks
Heroe anónimo impide asalto en 7Eleven. Los buenos somos más.
Readers added context they thought people might want to know
El video parece generado por IA. No hay reportes de medios sobre un asalto real con estas características en un 7-Eleven; usuarios lo identifican como falso.
x.com/Lint4P/status/…
x.com/hayquedecirlop…
old.reddit.com/r/aislop/comme…
x.com/PayasoTristeAl…
lux retweeted
1/4
Meet Qwenputer: Qwen3.5-4B with a full RISC-V CPU inside its weights, using the model's own Gated DeltaNet layers as RAM.
It runs C and MicroPython in its forward passes, no external runtime.
On DeepMind math: base 55–90%, base + Python tool 95–100%, Qwenputer 97.5–100%.
lux retweeted
i mentioned it twice today (@yaboilyrical, @lux) and i care so little about "alignment" clout that i can't be arsed to submit the paper; might as well launch.
re:dvx is a control vector library, with significant innovations making it SOTA (see below).
redvx.hyperplex.org
Will one of you Ridiculously Welthies out there *please* buy all of this artist's NFTs up or just give him one of the thousands of unused GPUs sitting unused in your datacenters? Out of all the tech-obsessed hypernerds on this bird app I know of noone who would put a GPU (or 10!) to better use than @soulblocks . He needs to be hyperscalled or this whole AI progress thing just doesn't add up. It's painful to watch him squeeze every last computation out of his fried laptop, despite the amazing results. Please like share widely the man deserves some serious metal and tokens!
sometimes, only sometimes, i let myself dream of a very wealthy person suddenly acquiring nearly all of my nft portfolio and then i could get a gpu, but i know this is a dream that is silly and shameful, and i shall wait for the day that a mere single nft of any collection would be at least worth 1 entire whole gpu, maybe even a Gpu, eventually a GPu, and my god perhaps someday maybe 2 of the nfts could evolve like a pokemon into one entire complete GPU, wow, someday guys, someday
Replying to @rolibosch
don't worry! AI will never replace Philosophers! It's not economically viable for AI
Or better ... an ontology
hey @Sauers_ as someone influencing a lot of the neuron experiments right now, you could standardize a text format to concisely and unambiguously describe steering or something for single to small neuron experiments
we need a spec!
identification plus maybe some standard ways to describe the operation, how the activations are being scaled, which parts of the neuron apply, and so on
thinking a bit like the old geek code from the Old Days
en.wikipedia.org/wiki/Geek_C…
steer:qwen3/qwen3-4b L4.N123++L
something like that
sort of compact sharable version of your neurons.json
lux retweeted
Nature's native language is topography/isomorphism. What is interesting, is that our brain may use representations close to reality but not the same. The representation of a frog might actually resemble a frog. Because we thought LLMs their first weights and tokens but living brains learnt their weights from nature. Their programmer is mother nature. So mother nature talks a common language but not numbers or weights. So frog's 🐸 representation must be in frog shape and cats representation must be in cat shape 🐯.
Retinotopy: The visual cortex contains maps of the retina. If you look at a cross-shaped light, a literal cross-shaped patch of neurons fires in your primary visual cortex.
Somatotopy: The motor and sensory homunculus map out your body layout directly on the surface of the brain. If you stimulate the "hand" area, you get hand movement; the layout of the brain mirrors the layout of the body.
Mother Nature uses spatial mapping because it is the most energy-efficient way to compute: keep things that are close together in the world close together in the neural tissue to minimize axonal wiring length and metabolic waste.
Just as a Chladni plate vibrates into a specific geometric shape when exposed to a specific audio frequency, a neural volume responding to a cat or a frog might form a distributed, phase-locked activation pattern that topological maps onto the object's behavioral or physical signature. It is a functional shadow of the object's geometry.
The SNARC Effect: Humans naturally map numerical magnitudes onto physical space (e.g., smaller numbers to the left, larger numbers to the right).
When you mentally rotate a 3D object, neuroimaging shows that your brain literally rotates the continuous wave activation pattern across its visual cortex over time. The time it takes to rotate an object in your mind scales linearly with the physical angle of rotation (Shepard & Metzler, 1971).
This reveals that thought is not a disembodied digital calculation, but an analog, spatial simulation. We don't just think about numbers and reality; we literally "draw" them onto the physical substrate of the brain and read the geometric outcome.
lux retweeted
these qwen neurons are insane like wym i poke its brain and now it's an alien named Extrali who's been guiding civilization for 3000 years lmfaoo
lux retweeted
i might just fuck around and swoop into this personal ai showdown like a thief in the night.
every frontier lab has completely lot sight of what 95% of the population cares about. theyre all so wrapped up in beating each other and going viral among the .001% of the population that exists within this little ai twitter bubble.
it's not your fancy dashboards and clever task management.
it's literally just chatting with something that understands you and feels like it's yours.
thats it. thats all 90% of your consumers care about. by the time a meaningful number of them care about handing off meaningful work, offering that as a service will be like offering an alarm clock on a smart phone.






















