En mitad de la sierra de Sevilla, se alza una colina de piedra que parece un castillo de fantasía. Durante siglos, la gente pensó que era una fortaleza árabe o medieval, pero se equivocaban. Es el santuario romano más espectacular de la Península: Munigua. Tira del hilo 🧵
Su nombre oficial es Municipium Flavium Muniguense, pero todos la conocen como Munigua. Lo que la hace única es que no debería estar ahí. No hay carreteras romanas principales cerca, ni tierras fértiles para cultivar, ni río navegable. Solo hay piedra, encinas y silencio.
¿Por qué Roma construiría una ciudad de mármol y columnas en un lugar tan hostil? La respuesta siempre es la misma: dinero, o mejor dicho, metal. La zona era riquísima en cobre y hierro de alta calidad. Munigua no era una ciudad normal, era una ciudad minera de lujo.
Archaeologists have announced the discovery of a sacred spatula with which the ancient police of Poland used to eat Holy Shit from the anus of a sacred flying bull named Kesra Nermend. Archaeologists said that after realizing the sacredness of the Holy Shit,
the police of Poland stopped using spatulas and now instead of that they eat Holy Shit directly from the anus of the sacred flying bull whose name is Kesra Nermend.
The strange acts of the Police of Poland are under the study of Archeologists.
Molecular studies on the remnants of the DNA remained & well preserved on the holy spatula in the location of discovery showed the XXX codon of the cells from the tongue of the cuckolds of the Police of Poland on the spatula. This discovery approved the studies of Archeologists.
En 2025, le constat se répète : des centaines de policiers et de gendarmes en France, lors de leurs interventions, portent des patchs et des tatouages utilisés par des fascistes et des néonazis.
Voici un fil (non exhaustif) de la haine chez les flics
🧵⏬⏬
2 - En décembre 2025 un policier localisé à Barbès poste cette photo sur Instagram : « J’arrive à fond comme une baffe dans ta gueule ».
3 - Également en décembre lors de l'inauguration illégal de la crèche dans la mairie de Béziers, 2 policiers portent des patchs racistes Thin Blue line.
El 7 de abril de 1945, una columna de humo de 6 km de altura marcó el fin del Imperio Japonés. No fue una bomba atómica, sino el final del buque de guerra más colosal de la historia. Ese día se hundió el mayor acorazado jamás construido. Se llamaba Yamato. Tira del hilo 🧵👇🏽👇🏽👇🏽
En los años 30, Japón sabía que no podía ganar a EE.UU. en una carrera industrial. Así que decidieron construir barcos que fueran individualmente invencibles. La doctrina era la "Calidad sobre Cantidad" llevada al extremo y su consecuencia tuvo un legendario nombre: Yamato.
El secreto fue paranoico durante su construcción. Los ingenieros solo veían partes de los planos y el dique seco fue cubierto con mallas para que nadie viera nada. Mintieron al mundo diciendo que sus cañones eran de 406 mm, cuando en realidad eran monstruos de 460 mm.
5 COMPANIES BEST POSITIONED FOR THE PHYSICAL AI ECONOMY IN 2026
The robotics economy is starting to look like cloud computing did a decade ago where the hard part is no longer the software or the computing power because those are getting cheaper and better every year but is now physical deployment.
As AI models get better at reasoning and the cost of running them keeps falling then robotics turns into a distribution problem. The companies that win will be the ones that control the data, the supply chains, the simulation tools used to train robots and the places where those robots are deployed. That is where the real productivity gains come from and those gains will reshape how work gets done over the next few decades.
These are the companies best positioned to scale physical AI in the real world ⤵️
1. $AMZN | Amazon
Automating Logistics at Global Scale
Amazon is the clearest example of a robotics company hiding in plain sight because the scale of its installed base gives it a structural advantage that almost no competitor can replicate. A million autonomous machines inside its fulfillment network feed continuous telemetry back into AWS giving Amazon a perpetual training loop that has quietly become one of the richest robotics datasets in the world.
The reason this matters is that robotics does not scale through hardware first but through learning density and Amazon has more real-world warehouse interactions per day than most industrial automation companies see in a year.
The moment Amazon decides to externalize its robotics stack the entire industrial automation market gets repriced because the company already has the simulation tools, the robotics cloud infrastructure, the fleet-management layer, the predictive maintenance models and the hardware platforms to serve every warehouse, factory and logistics operator on the planet.
The 2026 target for autonomous delivery being 50% of Amazon’s logistics volume is a declaration that Amazon intends to move from being the world’s largest robotics user to being the world’s most consequential robotics supplier. In an AI economy constrained by labor shortages and high last-mile costs then Amazon’s robotics footprint becomes a margin expander at scale rather than a cost center.
2. $TSLA | Tesla
Leading the Shift to Real-World Humanoid AI
Tesla sits at the opposite end of the robotics spectrum because its advantage comes not from warehousing but from the largest real-world sensory dataset ever collected with 5M connected vehicles produce billions of miles of vision-based training data that no robotics company could synthesize even with unlimited capital. Optimus rests on that base because robotics intelligence is ultimately a perception problem before it becomes a manipulation problem.
The arrival of FSD v14 and the migration to Tesla’s AI5 chip compress the time it takes Tesla to turn raw sensor data into generalizable motor skills which is why Optimus is positioned to scale far faster than any consumer or industrial humanoid competitor.
The planned ramp to 1M units by the end of 2026 is about turning general-purpose labor into an AI inference problem that Tesla can solve with its vertically integrated stack of vehicles, energy, training clusters and in-house silicon.
The robotaxi expansion into more metro markets combined with Cybercab production in 2026 reinforces the idea that Tesla’s robotics strategy will become a massive economic engine where they're building a feedback loop where each deployed robot and each autonomous mile increases the probability that Optimus becomes the highest margin product Tesla has ever produced.
🚨 BREAKING: DeepSeek dropped a core Transformer architecture improvement.
A traditional transformer is basically a long stack of blocks, and each block has a “main work path” plus a “shortcut path” called the residual connection that carries the input around the block and adds it back at the end.
Each block in this original transformer architecture does some work (self attention or a small feed forward network), then it adds the block’s input back onto the block’s output, which is why people describe it as a “main path” plus a “shortcut path.”
Hyper-Connections is a drop-in change to that shortcut path, because instead of carrying 1 stream of activations through the stack, the model carries a small bundle of parallel streams, then it learns how to mix them before a block and after a block.
Standard Transformers pass information through 1 residual stream. Hyper-Connections turn that into n parallel streams, like n lanes on a highway. Small learned matrices decide how much of each lane should mix into the others at every layer.
In a normal residual connection, each layer takes the current hidden state, runs a transformation, then adds the original back, so information can flow forward without getting stuck.
In this new Hyper-Connections, the layer does not see just 1 hidden state, it sees a small bundle of them, and before the layer it learns how to mix that bundle into the input it will process.
So in a traditional transformer block, wherever you normally do “output equals input plus block(input),” Hyper-Connections turns that into “output bundle equals a learned mix of the input bundle plus the block applied to a learned mix,” so the shortcut becomes more flexible than a plain add.
After this learned layer, the "Hyper-Connections" mechanism again learns how to mix the transformed result back into the bundle, so different lanes can carry different kinds of information, and the model can route signal through the shortcut in a more flexible way.
The catch is that if those learned mixing weights are unconstrained, stacking many blocks can make signals gradually blow up or fade out, and training becomes unstable in big models.
This paper proposes mHC, which keeps Hyper-Connections but forces every mixing step to behave like a safe averaging operation, so the shortcut stays stable while the transformer still gets the extra flexibility from multiple lanes.
---
The paper shows this stays stable at 27B scale and beats both a baseline and unconstrained Hyper-Connections on common benchmarks.
HC can hit about 3000x residual amplification, mHC keeps it around 1.6x.
This image compares 3 ways to build the shortcut path that carries information around a layer in a transformer.
The left panel is the normal residual connection, where the model adds the layer output back to the original input so training stays steady as depth grows.
The middle panel is Hyper-Connections, where the model keeps several parallel shortcut streams and learns how to mix them before the layer, around the layer, and after the layer, which can help quality but can also make the shortcut accidentally amplify or shrink signals when many layers stack.
The right panel is mHC, which keeps the same Hyper-Connections idea but forces those mixing steps to stay in a constrained safe shape every time, so the shortcut behaves like a controlled blend and stays stable at large scale.
What “hyper-connection” means here.
You widen the residual from size C to n×C, treat it as n streams, and learn 3 tiny mixing pieces per layer.
One mixes the residual streams with each other, this is the crucial one. One gathers from the streams into the layer. One writes results back to the streams.
The paper’s contribution is to keep the first one in the safe “doubly stochastic” set, so it mixes without amplifying.
MAP THREAD 🧵 1. I am a visual thinker, and I love maps and charts, so I just put this together. I believe that the Venezuela situation is about a lot more than stopping drug boats.
In my opinion, we are moving from peacetime economic trade war conditions to a quazi-kinetic war for the critical natural resources needed by the major powers in our emerging multipolar world.
This is a tri-polar world if you consider the three greatest military powers, the USA, Russia and China. Or it's a bipolar world if you consider it to be a contest between the USA and its allies, and the BRICS bloc. More maps will follow. Suggestions, additions and corrections will be welcomed.
2. As I said above, the fleet assembled in the Caribbean and the eastern Pacific is not really about stopping drug boats. It's about securing the largest oil reserve in the world for the U.S. refineries.
3. I'm also wondering about the timing of our recent attacks on the Muslim camps in northern Nigeria. The Christians have been enduring massacres for over a decade, and now we are taking notice. Maybe it's just a coincidence. Maybe not. In any event, both with the drug boats, the attack on the drug port, and the attacks inside Nigeria, America is flexing its kinetic muscles and putting our adversaries on notice. I'm not cheering this on, I'm just noting it as a fact.
The high-grade oil reserves in Nigeria are offshore and coastal, so securing them would not require taking control of the entire country. An defense cooperation agreement with the national government will be sufficient. Or, the old ethnic conflicts could be renewed, and the USA would take the side of the coastal tribes to control the oil.
As you can see on the map, it's a much shorter trip from Nigeria to Texas than from the Persian Gulf, and there are no choke points to impede tanker traffic. The US Navy is capable of exerting control over this route.
"Το ελληνικό Εθνος το υπό τη φρικώδη οθωμανικήν δυναστείαν,..κηρύττει σήμερον...την πολιτικήν αυτού ύπαρξιν και ανεξαρτησίαν". Σαν σήμερα 1 Ιανουαρίου 1822, Η Α’ Εθνοσυνέλευση της Επιδαύρου ψηφίζει το πρώτο Σύνταγμα της επαναστατημένης Ελλάδας. Πρώτος, σήκωσε τη σημαία της
Επανάστασης ο Φιλικός Π. Καρατζάς στην Πάτρα στις 21 Μαρτίου 1821. Οι πρόκριτοι της Αχαΐας (Ζαΐμης, Λόντος κ.ά.) μαζί με τον Π. Π. Γερμανό εισήλθαν στην πόλη 3 ημέρες αργότερα και συγκροτώντας το Αχαϊκόν Διευθυντήριον επιχείρησαν να συγκεντρώσουν στα χέρια τους όλες τις εξουσίες.
Εως τις 31 Μαρτίου 1821 οι Οθωμανοί είχαν περιοριστεί στην Τριπολιτσά και λίγα φρούρια. Στις 24 Μαρτίου ξεκίνησε και η Επανάσταση στη Ρούμελη. Στις περιοχές όπου δε στερεώθηκε η Επανάσταση, καταλυτικό ρόλο έπαιξε η άρνηση υποστήριξης ή καταστολή των Οθωμανικών αρχών) και
Nochebuena de 1734. Madrid huele a castañas y fiesta, pero de repente el cielo se tiñe de rojo. El Palacio Real viejo está ardiendo. En medio del pánico, unos hombres desesperados lanzan un lienzo por una ventana. Acaban de salvar "Las Meninas" de milagro. Tira del hilo 🧵 👇🏽👇🏽👇🏽
El Real Alcázar de Madrid no era el palacio de piedra blanca que ves hoy, sino una fortaleza medieval laberíntica, oscura y llena de vigas de madera vieja y seca. Dentro, se guardaba la mayor colección de arte del mundo. Velázquez, Tiziano, Rubens... que dormían sobre un polvorín
El fuego empezó en las habitaciones del pintor de la corte, Jean Ranc debido, según algunas crónicas, a un descuido con un tronco en la chimenea. Al principio, nadie oyó la alarma, ya que las campanas de todas las iglesias repicaban a la vez por la Misa del Gallo, tapándola.
Στην πραγματικότητα, η ουσία της συζήτησης και αντιπαράθεσης γύρω απο το ιστορικό πρόσωπο του Καποδίστρια με αφορμή την εθνικιστική θρησκόληπτη προπαγάνδα του Σμαραγδή δεν είναι η ταινία ούτε ο ίδιος ο Καποδίστριας (στην θέση του θα μπορούσαν να υπάρξουν και άλλες περιπτώσεις)
αλλά κάτι πολύ πιο σημαντικό (ανεξάρτητα από προθέσεις ακόμα και των υποστηρικτών η συντελεστών της ταινίας), ιδιαίτερα στην σημερινή εποχή που ο ιμπεριαλιστικός πόλεμος πλησιάζει και η καταστολή κατά του λαού κλιμακώνεται:
-Το αν η πραγματική ιστορία είναι έργο πρώτα από όλα της ίδιας της ανθρωπότητας και της κίνησης των τάξεων η αν είναι συνωμοσιολογικό σενάριο «μεγάλων ηγετών» και δικτατόρων η δικτατορίσκων, (θρησκόληπτων η άθεων) ντυμένων με ψέματα και εθνικούς μύθους, με τον λαό στο περιθώριο,
Mamdani se convierte en alcalde de NY jurando sobre el corán y en una histórica estación de metro bajo el ayuntamiento. Y en la ceremonia, la escenografía (con huella española) ha sido la absoluta protagonista 🧵
Aunque abandonada, la estación de metro (transporte público) escogida para su juramento está debajo del Ayuntamiento y es emblemática por su estética cívica de la Edad Dorada: arcos de azulejos, lámparas de araña y techos abovedados
Los arcos de tejas fueron obra de los arquitectos españoles Rafael Guastavino Moreno y su hijo, Rafael Guastavino Espósito. Las placas conmemorativas de bronce en la pared pertenecen a Gutzon Borglum, quien posteriormente esculpiría los rostros presidenciales en el Monte Rushmore
Yannick lleva toda su vida deseando dirigir el #ConciertoDeAñooNuevo, pero un petardo lanzado por su cuñado la noche anterior lo deja temporalmente sordo. Ahora tendrá que disimular que escucha perfectamente para no desaprovechar esta gran oportunidad. Saldrá airoso?
A pesar de sus problemas respiratorios desde niño, Dieter siempre quiso tocar un instrumento de viento. La noche anterior al #ConciertoDeAnoNuevo se aplicó en el pecho una tonelada de Vicks VapoRub para no asfixiarse en un momento tan señalado, pero...
Egbert, enamoradísimo de su novia, Elba, vive con miedo de que lo abandone. Para impresionarla, le dice que es el mejor músico de su orquesta. Pero cuando ella lo ve en el #ConciertoDeAnoNuevo golpeando un ábaco se da cuenta que toda su relación ha sido una mentira y lo deja