Jose Brox 🏳️‍🌈🏳️‍⚧️

Qué desazón, qué rabia y qué impotencia: @EsRebelCientif @SciRebelPT @esXrebellion

In 2022, software engineer Shawn Ligocki discovered a six-rule Turing machine. Its runtime has more digits than the number of atoms in the universe. quantamagazine.org/busy-beaver-hu…

An expert is a person who has made all the mistakes that can be made in a very narrow field. – Niels Bohr

Qualia columns go where curiosity leads. This week, plummet down another rabbit hole with Natalie Wolchover. This one gets into the mysterious nature of the very building blocks of reality. quantamagazine.org/how-many-eleme…

Each decade the world is losing over 7 per cent of its freshwater storage capacity to sediment build-up, according to an analysis of over half a million reservoirs #Echobox=1782067542" target="_blank" rel="nofollow noopener">newscientist.com/article/252952…

⚠️ AVISO ESPECIAL | Ola de calor. ➡️ Temperaturas muy altas para la época tanto de día como de noche en la mayor parte de la Península y Baleares hasta el miércoles. ➡️ El jueves bajarán las temperaturas, aunque todavía con calor intenso. + info 👉 aemet.es/es/eltiempo/pr…

No habrá posibilidad de paz y estabilidad duraderas en la región si Palestina vuelve a quedar relegada al olvido Israel ha perdido el relato, sí, pero el apoyo de EEUU, Reino Unido y la UE le sigue permitiendo ganar territorio palestino ✍️↘️ eldiario.es/internacional/…

A new, more capable version of Mythos has emerged from training. I don't know whether it will be called Mythos 5.1 or Mythos 6, or if Anthropic will keep it internal to accelerate further development - but it has arrived. Stopping models like Fable 5 or Mythos 5 from being served to the public does nothing to slow down development. In fact, it probably speeds it up slightly by freeing up resources. There are also no rules preventing the labs from continuing to advance capabilities while any current model is under embargo - or from keeping progress quiet until they choose to release it. None of them can afford to pause or slow down. We need only look at how capable GLM-5.2 is as proof of this. To protect their business models, the frontier labs must continually train increasingly capable systems to stay ahead of open source, and each other. The current continues to rage beneath the ice, and we continue to race toward our destination.

Von Neumann probe. Send one spacecraft to a nearby star. When it arrives it mines raw material builds two copies of itself and sends those on to two more stars. Those build copies too. The fleet doubles at every stop. Traveling below the speed of light a single self replicating machine could reach every star in the galaxy in a few million years. The galaxy is thousands of times older than that. So one civilization at any point in cosmic history should have filled the sky with these probes already. We see none - the Fermi paradox.

An AI engineer has claimed to have decoded Linear A: If correct, this would be the greatest achievement in archaeology/linguistics in 74 years—if not much longer! Linear A and Linear B were mysterious writing systems used in Ancient Crete/Greece, from 3000+ years ago. Linear B was finally cracked in 1952 by Michael Ventris, as the greatest achievement of 20th century archaeology/linguistics.

El Ejército israelí mata al cámara Ahmed Wishah de Al Jazeera en un ataque en Gaza eldiario.es/internacional/…

He really was right on time, wasn’t he

Francia: Cientos de servicios de trenes suspendidos, carreteras derretidas, hogares, residencias y escuelas (ciudades enteras) construidas para un clima que ya no existe. Se nos va la vida en la ADAPTACIÓN al nuevo mundo.

"Israel ha perdido el relato, sí, pero el apoyo de EEUU y la UE le permiten seguir ganando terreno, con más anexión ilegal de territorio y apartheid" ANÁLISIS | Estados e Israel: quién maneja a quién ✍️↘️ eldiario.es/internacional/…

“¿Dónde vivía esta gente indignada –y con toda la razón– en el periodo 2015-2024? ¿Dónde estaban cuando se aprobó y se aplicó la Ley de Partidos? ¿Por qué no se rompían la camisa en Twitter cuando se conoció el caso Titiriteros, el caso Alberto Rodríguez, el caso Granadinas, el caso Monedero, el caso Niñera, el caso Dina, el caso Oltra, las cuentas falsas de Trías en la portada de El Mundo, los desmanes de la policía patriótica, la causa general y la condena a los procesistas catalanes, la reunión de Inda y Urreztieta con la fiscal general del Estado (y su mano derecha, Álvaro García-Ortiz) en casa de Baltasar Garzón el día que salía Villarejo de la cárcel ¡Qué escándalo, aquí se juega!, gritan ahora con el pecho henchido de furia quienes agacharon la cabeza como avestruces cuando el Estado Profundo y sus terminales judiciales, mediáticas, empresariales y policiales trataban de destruir por todos los medios al movimiento de izquierda más fuerte de Europa, aquel que competía con el PSOE por ganar las elecciones y gobernar el país Durante años, tanto la prensa progresistamimada por la banca y la gran empresa como el PSOE (el sanchista y el antisanchista) tomaron partido por las cloacas y eligieron ser cómplices de Ferreras, Inda, Olivera, García-Castellón y sus secuaces. Tal vez consideraron que era el mal menor, que aquellas acciones ilícitas de la judicatura y de los aparatos policiales y mediáticos corruptos eran solo un accidente del camino, una anomalía temporal, desagradable pero necesaria para mantener a flote el deteriorado sistema monárquico del 78” Un editorial muy valiente de @ctxt_es ctxt.es/es/20260601/Fi…

Decrecimiento; muchos beneficios en pocas palabras.

A counter to this is "yes but people *don't* ignore it". But not ignoring things on scans is our norm because, until recently, we only did scans if there was a clear need. If we move to a scan-more-often paradigm, the norms of what we do with that information will surely adjust.

@adsuara Solo en la Conplutense hay 53 cátedras extraordinarias como la de Begoña Gómez. El 75% está dirigido o codirigido por personal externo a la Universidad. Así que sí, claro que se puede codirigir cátedras sin ser mujer del presidente. Es lo más normal. eldiario.es/politica/begon…

“I can’t help but feel flummoxed. And apparently that’s the general reaction.” quantamagazine.org/how-many-eleme…

8 RAG architectures for AI Engineers: (explained with usage) 1) Naive RAG - Retrieves documents purely based on vector similarity between the query embedding and stored embeddings. - Works best for simple, fact-based queries where direct semantic matching suffices. 2) Multimodal RAG - Handles multiple data types (text, images, audio, etc.) by embedding and retrieving across modalities. - Ideal for cross-modal retrieval tasks like answering a text query with both text and image context. 3) HyDE (Hypothetical Document Embeddings) - Queries are not semantically similar to documents. - This technique generates a hypothetical answer document from the query before retrieval. - Uses this generated document’s embedding to find more relevant real documents. 4) Corrective RAG - Validates retrieved results by comparing them against trusted sources (e.g., web search). - Ensures up-to-date and accurate information, filtering or correcting retrieved content before passing to the LLM. 5) Graph RAG - Converts retrieved content into a knowledge graph to capture relationships and entities. - Enhances reasoning by providing structured context alongside raw text to the LLM. 6) Hybrid RAG - Combines dense vector retrieval with graph-based retrieval in a single pipeline. - Useful when the task requires both unstructured text and structured relational data for richer answers. 7) Adaptive RAG - Dynamically decides if a query requires a simple direct retrieval or a multi-step reasoning chain. - Breaks complex queries into smaller sub-queries for better coverage and accuracy. 8) Agentic RAG - Uses AI agents with planning, reasoning (ReAct, CoT), and memory to orchestrate retrieval from multiple sources. - Best suited for complex workflows that require tool use, external APIs, or combining multiple RAG techniques. Most architectures here involve some form of retrieval-time decision. But they all run on top of whatever was already indexed. If that indexing step outputs messy chunks, every architecture inherits them. Improving it is a separate problem from the 8 above. My co-founder wrote about a better unit for the indexing step. The technique: - cuts corpus size by 40x. - reduces tokens per query by 3x. - improves vector search relevance by 2.3x. And it doesn't alter the retrieval algorithm, the reranker, or the embedding model. Read it below.

Most creatures use daylight to keep time. But a newly discovered jellyfish species has a mysterious biological clock that tracks periods of 20 hours, instead of 24. It suggests there may be more unconventional clocks across the tree of life. quantamagazine.org/the-jellies-th…