O

[Chris Offner @chrisoffner3d]

Inherited vs. self-made billionaires.

[O @MateElis4]

@bryan_johnson bruh he lived till 99 years old being born in 1924 dying in 2023 I’m not sure even your guaranteed to live that long with only the technology we have now. if you were born the same time you would’ve died just the same. death is inevitable.

[Rossst.03 @Rossst_03]

Kian Katanforoosh, Stanford AI lecturer (Forbes 30 Under 30):

"Wall Street will pay you $500K a year to build these models. I'd rather teach them to you for free."

this free stanford lecture holds the entire "AI predicts the market, 80% win rate" pitch the 2026 quant threads are selling you. and the man teaching it didn't take the fund money either, he co-built stanford's deep learning class, gave it to millions online for free, and started an AI company instead of a hedge fund.

at the board he builds it from scratch: a neural net doesn't predict the future, it learns the expected outcome across thousands of inputs at once, patterns no single indicator could hold. stack enough weak guessers, let them vote, the noise cancels and the signal survives. that's the whole "100 AI agents auditing the market" idea, minus the marketing.

backpropagation has been public since 1986. hinton won a nobel for it in 2024. random forests came out of leo breiman's free 2001 paper. none of it is secret. it's the same stack i mapped in the article above, old and free and sitting in a textbook the whole time.

and here's the honest part the win rate hides. a model that scored 80% on past data is describing the past, not promising the future. ensembles cut variance, they don't turn a weak edge into a real one, and the market shifts under the model in ways the training set never saw. the lecture is free. knowing whether your 80% survives on live capital is exactly the part the course skips.

[Mathelirium @mathelirium]

The Branch-Point Loom

Some equations do not draw curves but instead draw weave sheets.

This scene comes from the algebraic curve

y² = Δ(z,t),

where

Δ(z,t) = Πⱼ(z - βⱼ(t)).

The moving points βⱼ are branch points. These are special places where two sheets of the curve touch and swap identities. In the animation, those branch points appear as glowing mineral spindles moving across a dark surface.

The landscape is shaped by -log|Δ(z,t)|. Therefore, the branch points rise into sharp seams while the surrounding surface folds around them. The bright threads follow the horizontal trajectories of the quadratic differential Δ(z,t)dz².

#Mathematics #ComplexAnalysis #AlgebraicGeometry #MathArt #MathematicalArt #Geometry #PhysicsVisualization #STEM #Art

[leopardracer @leopardracer]

THE MOMENT YOUR SECOND BRAIN CLICKS INTO PLACE LOOKS EXACTLY LIKE THIS

scattered dots that mean nothing on their own and then the connections turn on and suddenly there’s a structure that wasn’t there before

this is what karpathy was describing, knowledge isn’t in the notes themselves, it’s in what forms between them when someone finally draws the lines

the LLM wiki does this automatically every time you add a source and your vault has never done this once

bookmark & like this so you remember to actually set it up

[Kalshi @Kalshi]

BREAKING: Workers are receiving the smallest share of US income on record

[Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) @teortaxesTex]

The kids will be scary indeed. Imagine +4SD high schoolers who grew up watching DeepSeek/Moonshot/Unitree ascendance and their age peers becoming legends, using frontier models, who learned "wait I'm actually able to build Big Things already"

[leopardracer @leopardracer]

THIS IS WHAT 3 YEARS OF OBSIDIAN NOTES SHOULD LOOK LIKE BY NOW

what you see here is thousands of connections forming a structure that didn’t exist before

this is exactly what karpathy built with the llm wiki pattern, every source rewires the whole thing and new structure emerges automatically instead of you manually linking notes for years

knowledge that connects itself looks nothing like a productivity system

bookmark this and send it to whoever still thinks obsidian is just for taking notes

[Techartist @techartist_]

This is what 30 mins of vibe coding can get you these days: a beautiful landing page with a pelagic canvas and morphing jellyfish made of particles.

[Financial Dystopia @financedystop]

This guy owns a roofing company and walked away from an $8,000 job over a $200 difference. He said there were already some early red flags with the customer, and once the guy started haggling over $200 on a project that size, he realized the job probably wasn’t worth the headache.

[How To Prompt @HowToPrompt__]

NVIDIA just open-sourced a model that takes broken, blurry 3D scans and rebuilds them clean from any angle..

It's called ArtiFixer. It uses video diffusion to generate the camera angles you never captured, then reconstructs the scene from the generated frames.

→ 70x faster than anything
→ Finishes in 1 to 4 steps
→ Beats SOTA by 3dB
→ Works from just text prompts

100% Open Source.

[Quanta Magazine @QuantaMagazine]

Decades ago, Hungarian mathematician Paul Erdős used randomness to illuminate the vast and weird world of networks. Today, mathematicians are making his technique even more powerful.
https://t.co/sccNL8nG1l

[angel @angeldot_]

https://t.co/Vo9KvUDmsK

[angel @angeldot_]

GOOGLE HA LIBERADO EN SILENCIO UNA IA QUE PREDICE PATRONES

Ventas. Precios de mercado. Tráfico web.
Demanda energética. Volatilidad cripto.

Se llama TimesFM:

→ Entrenada con 100B de datos reales
→ Forecasting zero-shot, sin fine-tuning
→ Corre en local.

Probablemente el lanzamiento más loco que ha hecho Google en los últimos años, y nadie está hablando de ello.

100% Gratis y Open Source.

Enlace abajo👇

[Dilum Sanjaya @DilumSanjaya]

Haven't seen any demos of using vibe coding to prototype gadgets like this

So I made one. Everything here is generated with code

Parts of the code and logic can even be reused on a Raspberry Pi to build the real thing

Some tips ↓

[Dilum Sanjaya @DilumSanjaya]

Did some experiments with local models

In this example, all requests are handled by Gemma 4. It generates a new circuit as JSON based on the prompt

Edited out the waiting times from the clip. It usually takes around 5 - 10s

More details ↓

[sarcaxtix @sarcaxtiSX]

https://t.co/mSHCOiQ2J5

[sarcaxtix @sarcaxtiSX]

DÜNYANIN EN GARİP REPOSUNU BULDUM.

githubda bulduğum bu repo, aristo, feynman, kahneman, torvalds gibi 18 farklı yapay zeka kişiliğiyle senin en zor sorunu alıp kendi aralarında tartışıyor ve tek bir ortak karar çıkarıyorlar. artık en ufak bir kararı bile tek bir aiya sormak tembellik gibi geliyor. çoklu perspektif demeyi bırak, artık çoklu kafa dönemi. repo tamamen açık kaynak. kaydetmeyi unutmayın yorumlara koydum.

[Vai Viswanathan @vai_viswanathan]

A world model controlling a drone flying outdoors at 100Hz! 🚁

- Cuts position-tracking error 26-38% and attitude error up to 54% versus predictive baselines

- Trained in simulation and deployed zero-shot on hardware.

- predicts future dynamics in latent space instead of forecasting raw states, dodging the compounding-error drift that wrecks autoregressive models. A physics-inspired “prober” maps frozen latents back to real position, velocity, and attitude for an onboard MPPI controller.

- Why it matters: The first example of JEPA world models surviving the messiness of outdoor deployment.

The whole model is ~9K parameters, small enough to run on an embedded flight computer 🧠

[deep Manifold @BetaTomorrow]

Title: Reasoning as Attractor Dynamics: Latent Memory Retrieval via Gibbs-Weighted Energy Minimization
Author: Kanishk Awadhiya

#DeepManifoldInterpretation
** The central limitation of the Hopfield formulation is not its definition of equilibrium. Equilibrium as an unchanged state is mathematically standard. Its foundational hypothesis is that memory and neural computation should be represented by convergence under a fixed operator toward static, stable states. This may describe restricted associative retrieval, but it should not be generalized into a theory of dynamic neural inference or reasoning. **

From the Deep Manifold view, the relevant object is not necessarily one static equilibrium. It may instead be a prompt-conditioned, dynamically generated fixed-point class reached through multiple stochastic iterated-integral pathways. The operator, boundary and local geometry co-evolve, so “the neural state no longer changes” is too restrictive—and perhaps the wrong definition of successful inference.

For the paper we are discussing, the problem becomes even greater: it borrows Hopfield’s equilibrium interpretation without even possessing Hopfield’s recurrent state dynamics. It has neither a fixed neural-state operator nor demonstrated convergence. The supposed “attractor” is created by externally weighting completed trajectories.