Félix Morales

Sometimes the best way solve your own problems is to help someone else

[Agenda Paraguay @agparaguay]

🚨 Nakayama insiste en eliminar seguro médico privado de funcionarios públicos

El senador @edu_nakayama planteó nuevamente la necesidad de eliminar el seguro médico privado de los funcionarios públicos y trasladar esos recursos a Salud Pública.

Sostuvo que la medida permitiría direccionar más fondos al sistema público de salud y cuestionó la falta de voluntad política para concretar la aprobación de la propuesta.

[Miguel Ángel Durán @midudev]

¿Tu IA siempre crea la misma UI? ¡Prueba esto!

Una colección de archivos DESIGN.md con el sistema visual de marcas reales:
Apple, Airbnb, Tesla, Shopify, Stripe, Notion…

Colores, tipografía, espacios, componentes:
→ https://t.co/66H3JKE1fd

[Robert Singer @SingerSports]

Si sos asunceno y el 4 de octubre no votas por este chad sos un hijo de mil putas luqueño que descendió en su centenario

[Probability and Statistics @probnstat]

Mixture of Experts (MoE): A Statistical View

A Mixture of Experts (MoE) model combines several specialized predictors (“experts”) through a gating function. For input x, the conditional distribution can be written as

p(y|x) = Σₖ₌₁ᴷ πₖ(x) pₖ(y|x),

where πₖ(x) ≥ 0 and Σₖ πₖ(x) = 1. The gating network therefore learns which expert should contribute, and by how much, depending on the input.

In the classical statistical setting, MoE provides a flexible way to model heterogeneity and multimodality: different regions of the covariate space can be governed by different regression or classification relationships. It can be viewed as a conditional form of mixture modeling, closely related to latent-variable models and EM-type estimation.

Modern neural MoEs replace conventional experts with neural networks and learn the gating mechanism jointly. A sparse MoE activates only a small subset of experts for each input:

y = Σₖ∈S(x) πₖ(x) fₖ(x), |S(x)| ≪ K.

This creates an important scaling advantage: a model can contain billions of parameters while using only a fraction of them for any individual input.

MoE architectures are increasingly important in ML and AI, particularly for large language models, multimodal systems, recommendation, computer vision and multitask learning. Experts can specialize in different languages, domains, tasks or patterns, while the gate dynamically routes information.

From a statistical perspective, MoE offers a powerful principle: learn the population heterogeneity instead of forcing one global predictor to explain everything.

[Miguel Ángel Durán @midudev]

Una joya de repositorio. Te da todos los pasos para proteger y asegurar tu servidor Linux.

Si tienes un VPS, esto es la clave:
→ https://t.co/yJD2GnXFPj

[Fazt @FaztTech]

Hoy se cruzó un umbral raro en IA:
Ya puedes generar video más rápido de lo que lo ves.

Usando Minimax H3 Max, un usuario en X armó un livestream infinito, bajo la tematica de Rick y Morty cambiando de canal y la tele generando shows interdimensionales en vivo.

Y es que ahora puedes generar 15 segundos de video en 9 segundos. Antes tardaba 2–5 minutos.

Eso habilita livestreams que no se acaban nunca.

Y otro usuario (@levelsio ) también ha creado su livestream donde pueden probarlo por ustedes mismos, usando este enlace: https://t.co/Hq7addbBbk

[Manz 🇮🇨⚡👾 @Manz]

Poco se habla del BRUTAL trabajo del equipo de OpenCode, no sólo con el agente, sino también con los planes OpenCode Go y OpenCode Zen y la filosofía que siguen.

¡Larga vida a OpenCode! 💖

[Jesús Melga @CJ_Melga]

Poco se habla de la agenda protestante evangélica que se viene activando a nivel político en el país. Esto tiene detrás una carga que va mucho más allá del culto. Hay implicancias que, desde la cosmovisión de estas denominaciones, afectan directamente lo económico y lo social de las sociedades.

Su visión individualista atenta contra el sentido comunitario, con una lógica centrada fundamentalmente en el individuo y en su propia comunidad de creyentes. A esto se suma la doctrina de la prosperidad, que termina condenando a los más pobres al presentar el éxito material como una manifestación de bendición y, en muchos casos, la pobreza como consecuencia de una falta de fe o de compromiso religioso.

En lo político, una parte importante de estas denominaciones se alinea con el sionismo cristiano, corriente que funciona como un instrumento de influencia política y cultural favorable a los intereses del Estado de Israel y que ha logrado una creciente penetración en distintos países de la región. Basta observar lo que ocurre en Argentina, Colombia y el propio Paraguay.

Ni hablar de las implicancias teológicas: rebautismos completamente ilícitos cuando ya existe un bautismo válido, doctrinas contradictorias incluso entre las propias denominaciones protestantes y un marcado sentimentalismo que termina convirtiendo la experiencia de fe en algo esencialmente subjetivo, dependiente de la emoción y de la experiencia personal.

Lo increíble es cómo la sociedad todavía no dimensiona el impacto que puede generar este lobby cuando trasciende el ámbito estrictamente religioso y comienza a adquirir capacidad de influencia política, económica, social y cultural.

[Gui Cury @guicury]

🗣️ Você faz o que da vida?
🤠 Caço vampiro
🗣️ Mas eu nunca vi um vampiro!
😉 De nada!

[NPY Oficial @npyoficial]

🚨 #URGENTE | Renuncia masiva de médicos del Hospital General Pediátrico Niños de Acosta Ñu

El 86% del plantel decidió tomar esta medida ante el silencio del presidente de la República.

🔴 Seguí la transmisión en vivo: https://t.co/zXQhHR2kMj

#NPY #NosConecta

[elvis @omarsar0]

Impressive new paper from Google DeepMind.

(bookmark it)

It takes Co-Scientist out of simulation and into real-world experiments.

A summary of the results:

In computer science, it found an inference-time scaling architecture that beat six frontier models on HealthBench Hard and Professional under blinded physician review.

The system designed a safe precursor route for MXenes and drove a semi-automated chemical vapor deposition reactor, producing a lamellar 2D material with structural similarities to the Ti3C2Tx lattice.

It also tailored growth recipes to laboratory constraints in minutes, enabling single-attempt growth of monolayer MoS2, MoSe2, and WS2. In biology, it predicted E. coli swarming phenotypes across inducer gradients from sparse imaging data, matching unpublished real-world measurements.

30 domain experts wrote 450 reviews on end-to-end generated papers, and the reliability modules reduced hallucination and plagiarism.

Paper: https://t.co/M2NHPCJ5so

Chat with Paper: https://t.co/JlWJLDnaUv

[OpenRouter @OpenRouter]

GLM-5.3 from @Zai_org is now open-weight and live on OpenRouter.

Built for complex software engineering, long-horizon agents, and cybersecurity, with 1M context and configurable reasoning effort.

Try it: https://t.co/xZCGZAcvme

[Fahad Shah @FAHADSH97656634]

After the DuckLabs acquisition, is it time to bet on Apache DataFusion?

A growing ecosystem of data systems is already building on DataFusion, backed by the Apache Software Foundation's predictable open-source governance model.

As recently shared by @andrewlamb1111, Apache Foundation brings more than a license:
→ Multi-vendor participation that spreads investment and reduces dependency on a single company
→ Stable licensing with a low risk of sudden changes
→ Open and predictable communication
→ Long-term maintenance that is less tied to corporate strategy shifts or funding cycles

For organizations making long-term bets, that combination of governance, transparency, and ecosystem growth matters.

An open foundation that is trusted by a growing ecosystem for building infrastructure with long-term stability.

[Tom Yeh @ProfTomYeh]

Vector Database by hand ✍️ ~ 10 steps walkthrough below

Vector databases are the backbone of Retrieval Augmented Generation (RAG).

How do they actually work?

Goal: index three sentences, then answer a query by finding the nearest one, filling in every cell yourself.

= 1. Given =

A dataset of three sentences, three words each. In practice it is millions of them.

= 2. Word embeddings =

Let us look up each word in an embedding table. Here the vocabulary is 22 words; in practice it is tens of thousands, and the vectors have thousands of dimensions rather than four.

= 3. Encoding =

We feed the sequence to an encoder, one linear layer and a ReLU, and get one feature vector per word. In practice the encoder is a transformer.

= 4. Mean pooling =

Let us average across the columns. Three word vectors collapse into one, which is what people mean by a text embedding or a sentence embedding.

= 5. Indexing =

We multiply by a projection matrix and the four dimensions become two. It is doing the job of a hash: a short representation that is faster to compare, and it is what gets saved in the vector storage.

= 6. Process "who are you" =

Let us repeat steps 2 to 5 on the second sentence.

= 7. Process "who am I" =

We do it a third time. The database is now indexed.

= 8. Query "am I you" =

Let us push the query through the very same pipeline: lookup, encoder, mean pooling, projection, and it lands as a 2D vector in the same space.

= 9. Dot products =

We transpose the query and multiply, which takes the dot product against every stored vector at once. The dot product is the estimate of similarity.

= 10. Nearest neighbour =

Let us scan for the largest: 60/9 beats 44/9 and 40/9, so the answer is "who am I". Scanning billions of vectors one at a time is what makes this the slow step in practice, which is why real databases use an approximate nearest neighbour index like HNSW.

The outputs:
Stored index vectors = [5/3, 2/3], [5/3, 0], [7/3, 2/3]
Query vector = [8/3, 2/3]
Dot products = 44/9, 40/9, 60/9
Nearest neighbour = "who am I"

The takeaway: a vector database is an embedding pipeline, a projection, and a dot product. Every step here is arithmetic you can do in pen, which is worth remembering when the word "database" makes it sound like something else.

💾 Save this post!

[freeCodeCamp.org @freeCodeCamp]

AI-generated tests can pass without actually protecting the behavior they’re supposed to verify.

In this tutorial, @dannwaneri teaches you how mutation checking can expose these vacuous tests.

You’ll also learn how to distinguish broken tests from genuinely untestable criteria and handle human verification more carefully.

https://t.co/TL1ht6gmtu

[FIUNA @IngenieriaUNA]

El TEI ad hoc del Consejo Directivo de la FIUNA informa a la comunidad universitaria que ha culminado el proceso de elección de Decano y Vicedecano de la Institución para el periodo 2026–2031.
Leer más: https://t.co/c4McZZj12m 🔗

[Rémi Ouazan @remi_or_]

same day new qwen and new glm drop
life feels good in open source

[Python Coding @clcoding]

Generalized Bhattacharyya and Chernoff Upper Bounds on Bayes Error Using Quasi-Arithmetic Means

The paper covers:

Bayesian classification and Bayes error
Bhattacharyya upper bounds
Chernoff information
Quasi-arithmetic means
Statistical divergences and affinity coefficients
Applications to Cauchy and multivariate t-distributions

👉 Download / Read the Free PDF: https://t.co/KCJDHKu82S

[Design Arena @DesignArena]

BREAKING: MiniMax H3 Max sets the new Pareto Frontier for video generation, nearly 50x faster than the base model.

This model is post-trained by @fal on @MiniMax_AI H3, and it's in a league of its own: no other Image to Video model on the arena delivers higher preference at a lower generation time.

Its Image-to-Video generation time is just 6.4 seconds, 18x faster than average, and its Text-to-Video generation time is just 4.7 seconds, 24x faster than the average.

Huge congratulations to the @fal team on this release!

[ODSC (Open Data Science Conference) AI @_odsc]

Your AI agent can be brilliant—and still fail if its context is outdated.

On Sept. 9, see how IBM, Confluent & Google Cloud use real-time data to build more reliable AI agents.

Live architecture + practical patterns.
Register: https://t.co/SWuPj6zG4u

#AIAgents #ODSCAI