Podcasts about Vue

durée : 00:03:41 - Tout va trop vite - par : Jean-Philippe Balasse - En cette journée carburant sur France Inter, zoom sur la flambée des prix vue des réseaux sociaux. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:03:41 - InterNational - par : Jean-Philippe Balasse - En cette journée carburant sur France Inter, zoom sur la flambée des prix vue des réseaux sociaux. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:03:41 - Le 5/7 - par : Jean-Philippe Balasse - En cette journée carburant sur France Inter, zoom sur la flambée des prix vue des réseaux sociaux. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

À première vue, rien de particulièrement extraordinaire dans Intérieur d'une cuisine. Cette peinture de Martin Drölling, réalisée en 1815 et aujourd'hui conservée au musée du Louvre, représente une scène domestique paisible : une cuisine, quelques personnages, des ustensiles et des meubles.Pourtant, derrière cette scène parfaitement banale se cache une histoire macabre. Une partie des pigments bruns utilisés par Drölling pourrait avoir été fabriquée à partir de… cœurs humains momifiés. Et pas n'importe lesquels : ceux de membres de la famille royale française.Pour comprendre cette histoire, il faut revenir à la Révolution française. À partir de 1793, plusieurs tombeaux et reliquaires royaux sont profanés. Or, selon une tradition ancienne, les corps des souverains pouvaient être enterrés séparément de leur cœur. Certains de ces cœurs embaumés furent donc récupérés et circulèrent ensuite entre différentes mains.À cette époque, les peintres connaissent déjà un étrange pigment appelé « brun de momie » ou « mummie ». Pendant plusieurs siècles, des marchands européens ont effectivement vendu une matière brunâtre contenant parfois des restes de momies égyptiennes réduites en poudre. Mélangée à l'huile, elle permettait notamment d'obtenir certains bruns et glacis.Le peintre Martin Drölling aurait ainsi acquis plusieurs cœurs momifiés provenant de membres de la famille royale, notamment ceux d'Anne d'Autriche et de Marie-Thérèse d'Autriche. Il les aurait utilisés pour fabriquer une matière picturale employée dans certaines de ses œuvres.C'est ici qu'Intérieur d'une cuisine devient particulièrement intrigant. Selon une tradition historique, certains de ses tons bruns contiendraient cette fameuse « mummie » royale.Mais il faut être précis : cela reste, pour ce tableau particulier, une hypothèse. Le Louvre confirme que l'œuvre est une huile sur toile peinte par Drölling en 1815, mais sa notice ne mentionne pas la présence de restes humains. Et aucune analyse scientifique du tableau n'a encore permis de confirmer que ses pigments contiennent effectivement des tissus provenant de cœurs royaux.En revanche, l'histoire n'est pas totalement fantaisiste. Des chercheurs ont étudié une autre peinture, Vue de Caen, d'Alexandre Pau de Saint-Martin, un autre artiste associé à cette pratique. Des analyses ont permis d'y retrouver des éléments compatibles avec l'utilisation de tissus cardiaques.Voilà donc ce qui rend Intérieur d'une cuisine si singulier : derrière une scène quotidienne presque insignifiante pourrait littéralement se cacher une partie de la monarchie française.Un tableau dans lequel l'Histoire ne serait pas seulement représentée : elle serait présente dans la peinture elle-même. Hébergé par Acast. Visitez acast.com/privacy pour plus d'informations.

Hervé Gattegno nous plonge dans les coulisses de l'actualité du jour en passant au crible les principaux journaux.D'abord, une enquête du Monde sur la rumeur du retour des gilets jaunes. Deux journalistes ont scruté les réseaux sociaux pour comprendre l'origine des appels à manifester le 17 octobre contre la hausse des prix des carburants.Ensuite, Le Parisien s'intéresse à la visite du pape en France. Des victimes d'abus sexuels dans l'Église vont le rencontrer à Lourdes le 27 septembre.Dans le domaine judiciaire, une décision rendue en juin secoue le monde des magistrats. Le collège de déontologie de l'ordre judiciaire considère pour la première fois que l'appartenance à la franc-maçonnerie est incompatible avec les obligations déontologiques des juges. Enfin, Point de Vue raconte la rentrée du prince George à Eton. Ses parents, William et Kate, ont négocié un accord avec les médias : quelques photographes autorisés pour immortaliser ce moment, en échange d'une préservation stricte de la vie privée du jeune prince. George portera l'uniforme traditionnel d'Eton et sera pensionnaire, sans accès aux smartphones.Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.

Is streaming really killing off the big screen? Greg Foot examines the future of cinemas and considers wonder products and trends that might rise to success.Greg is joined by experts and entrepreneurs to discuss what is shaping today's everyday technology and what it may look like tomorrow, before the futurist Tom Cheesewright offers his predictions on what cinemas might be like within five, ten and fifty years.In this episode:- Tim Richards - the founder and CEO of Vue cinemas - Patrick von Sychowski - known in the trade as "the Cinema Guru" he's a writer and blogger who works for CinemaNext, the largest installer and upgrader of cinemas across Europe. - Alexis Scalice - Vice President for Business Development & Entertainment at Cosm which is pioneering 'shared reality' cinematic experiences in large entertainment domesAmong the questions up for consideration: Why are there more cinemas screenings of films from the past? Why are plastic popcorn buckets becoming collectors' items? And could the cinema of the future be a giant, panoramic dome?Produced by Jon Douglas. Dough is an Audio North production for BBC Radio 4 and BBC Sounds.

Is streaming really killing off the big screen? Greg Foot examines the future of cinemas and considers wonder products and trends that might rise to success.Greg is joined by experts and entrepreneurs to discuss what is shaping today's everyday technology and what it may look like tomorrow, before the futurist Tom Cheesewright offers his predictions on what cinemas might be like within five, ten and fifty years.In this episode:- Tim Richards - the founder and CEO of Vue cinemas - Patrick von Sychowski - known in the trade as "the Cinema Guru" he's a writer and blogger who works for CinemaNext, the largest installer and upgrader of cinemas across Europe. - Alexis Scalice - Vice President for Business Development & Entertainment at Cosm which is pioneering 'shared reality' cinematic experiences in large entertainment domesAmong the questions up for consideration: Why are there more cinemas screenings of films from the past? Why are plastic popcorn buckets becoming collectors' items? And could the cinema of the future be a giant, panoramic dome?Produced by Jon Douglas. Dough is an Audio North production for BBC Radio 4 and BBC Sounds.

Het werd al verwacht, en toch valt het op, want het was alweer drie jaar geleden: de Amerikaanse centrale bank heeft een renteverhoging aangekondigd. Fed-baas Kevin Warsh trekt zich daarmee niks aan van president Trump, die juist wil dat de rente omlaag gaat. Toch blijft hij vooralsnog mild voor Warsh. Deze aflevering bespreken we dat je met die renteverhoging aan moet als belegger. We kijken wat het doet met de Amerikaanse economie. En we bespreken hoe Kevin Warsh invulling gaat geven aan de rest van zijn termijn. Hebben we het ook nog over het Chinese Huawei, want dat werkt aan een AI-chip die kan concurreren met Nvidia. Dat die chip eraan komt, dat wisten we wel. Maar het bedrijf haalt de release naar voren en wil 'm nu al begin volgend jaar op de markt brengen. Wat dat voor Nvidia kan betekenen, zoeken we voor je uit. Hoor je ook nog over: Een nieuwe handelsoorlog?? Trump dreigt ermee Unilever, dat verkoopt wéér een merk Een nieuwe investering van Softbank in OpenAI ExxonMobil, dat ziet nu ook concrete kansen in Venezuela Te gast: Erik Mauritz van Trade Republic BNR Beurs is een journalistiek onafhankelijke productie, mede mogelijk gemaakt door Saxo. Over de makers: Jelle Maasbach is presentator van BNR Beurs en freelance financieel journalist. Zijn favoriete aandeel om over te praten is Disney, maar daar lijkt hij de enige in te zijn. Sinds de eerste uitzending van BNR Beurs is 'ie er bij. Maxim van Mil is presentator van BNR Beurs en journalist bij BNR, waar hij zich focust op de financiële markten en ontwikkelingen in de tech-wereld. Je krijgt hem het meest enthousiast als hij kan praten over ASML, of oer-Hollandse bedrijven zoals Ahold of ABN Amro. Jorik Simonides is presentator van BNR Beurs, economieredacteur en verslaggever bij BNR. Hij wordt er vooral blij van als het een keer níet over AI gaat. Je hoort hem ook in de BNR-podcast Moerdijk: dorp van de rekening. Milou Brand is presentator van BNR Beurs, freelance podcastmaker en columnist bij het Financieele Dagblad. Jochem Visser is presentator van BNR Beurs, maakt Beursnerd XL en is redacteur bij de podcast Onder Curatoren. Vraag hem naar obscure zaken op financiële markten en hij vertelt je waarom het eigenlijk nóg leuker is dan je al dacht. Donner Bakker is presentator van BNR Beurs en economieredacteur bij BNR. Hij volgt macro-economische ontwikkelingen met bovengemiddelde interesse, net als de beursgenoteerde datingapps. Je hoort hem ook in de BNR-podcast All In The Game. Over de podcast: Met BNR Beurs ga je altijd voorbereid de nieuwe beursdag in. We praten je in een kleine 25 minuten bij over alle laatste ontwikkelingen op de handelsvloer. We blijven niet alleen bij de AEX of Wall Street, maar vertellen je ook waar nog meer kansen liggen. En we houden het niet bij de cijfers, maar zoeken ook iedere dag voor je naar duiding van scherpe gasten en experts. Of je nu een ervaren belegger bent of net begint met je eerste stappen op de beurs, de podcast biedt waardevolle inzichten voor je beleggingsstrategie. Door de focus op zowel de korte termijn als de lange termijn, helpt BNR Beurs luisteraars om de ruis van de markt te scheiden van de essentie.See omnystudio.com/listener for privacy information.

Het werd al verwacht, en toch valt het op, want het was alweer drie jaar geleden: de Amerikaanse centrale bank heeft een renteverhoging aangekondigd. Fed-baas Kevin Warsh trekt zich daarmee niks aan van president Trump, die juist wil dat de rente omlaag gaat. Toch blijft hij vooralsnog mild voor Warsh. Deze aflevering bespreken we dat je met die renteverhoging aan moet als belegger. We kijken wat het doet met de Amerikaanse economie. En we bespreken hoe Kevin Warsh invulling gaat geven aan de rest van zijn termijn. Hebben we het ook nog over het Chinese Huawei, want dat werkt aan een AI-chip die kan concurreren met Nvidia. Dat die chip eraan komt, dat wisten we wel. Maar het bedrijf haalt de release naar voren en wil 'm nu al begin volgend jaar op de markt brengen. Wat dat voor Nvidia kan betekenen, zoeken we voor je uit. Hoor je ook nog over: Een nieuwe handelsoorlog?? Trump dreigt ermee Unilever, dat verkoopt wéér een merk Een nieuwe investering van Softbank in OpenAI ExxonMobil, dat ziet nu ook concrete kansen in Venezuela Te gast: Erik Mauritz van Trade Republic BNR Beurs is een journalistiek onafhankelijke productie, mede mogelijk gemaakt door Saxo. Over de makers: Jelle Maasbach is presentator van BNR Beurs en freelance financieel journalist. Zijn favoriete aandeel om over te praten is Disney, maar daar lijkt hij de enige in te zijn. Sinds de eerste uitzending van BNR Beurs is 'ie er bij. Maxim van Mil is presentator van BNR Beurs en journalist bij BNR, waar hij zich focust op de financiële markten en ontwikkelingen in de tech-wereld. Je krijgt hem het meest enthousiast als hij kan praten over ASML, of oer-Hollandse bedrijven zoals Ahold of ABN Amro. Jorik Simonides is presentator van BNR Beurs, economieredacteur en verslaggever bij BNR. Hij wordt er vooral blij van als het een keer níet over AI gaat. Je hoort hem ook in de BNR-podcast Moerdijk: dorp van de rekening. Milou Brand is presentator van BNR Beurs, freelance podcastmaker en columnist bij het Financieele Dagblad. Jochem Visser is presentator van BNR Beurs, maakt Beursnerd XL en is redacteur bij de podcast Onder Curatoren. Vraag hem naar obscure zaken op financiële markten en hij vertelt je waarom het eigenlijk nóg leuker is dan je al dacht. Donner Bakker is presentator van BNR Beurs en economieredacteur bij BNR. Hij volgt macro-economische ontwikkelingen met bovengemiddelde interesse, net als de beursgenoteerde datingapps. Je hoort hem ook in de BNR-podcast All In The Game. Over de podcast: Met BNR Beurs ga je altijd voorbereid de nieuwe beursdag in. We praten je in een kleine 25 minuten bij over alle laatste ontwikkelingen op de handelsvloer. We blijven niet alleen bij de AEX of Wall Street, maar vertellen je ook waar nog meer kansen liggen. En we houden het niet bij de cijfers, maar zoeken ook iedere dag voor je naar duiding van scherpe gasten en experts. Of je nu een ervaren belegger bent of net begint met je eerste stappen op de beurs, de podcast biedt waardevolle inzichten voor je beleggingsstrategie. Door de focus op zowel de korte termijn als de lange termijn, helpt BNR Beurs luisteraars om de ruis van de markt te scheiden van de essentie.See omnystudio.com/listener for privacy information.

Épisode 1499 : Quand une plateforme affiche « 1 million de vues », qu'est-ce qu'elle nous dit vraiment ?Pas forcément qu'un million de personnes ont regardé. Encore moins qu'elles ont aimé, compris, retenu ou partagé le contenu.Le paradoxe de la vue, c'est que c'est à la fois une donnée technique, une preuve de popularité, un argument business et une validation sociale.Alors on va essayer de répondre à cette question : ça vaut quoi, une vue ?YouTube : la règle vient de changer sous nos yeuxDepuis le 24 août 2026, YouTube compte une vue dès la première image de la vidéo, sans durée minimale.La règle s'applique à tous les formats en même temps : Shorts, vidéos longues, lives.Avant ce changement, la règle n'a jamais été publiée avec un chiffre précis. On parlait d'un seuil de 30 secondes mais cette limite semble venir des règles publicitaires, pour qualifier les vues facturées sur les formats in-stream skippables.L'ancienne règle, plus stricte, n'a pas disparu : elle est devenue la métrique « vues engagées », celle qui sert désormais à la monétisation et à l'éligibilité au programme partenaire.Sur Instagram et Facebook, Meta a tout fusionné dans un seulDepuis 2025, Meta a supprimé deux métriques historiques, « impressions » et « plays », pour les remplacer par un unique compteur : « vues ».Il s'applique à tout, Reels, vidéos, stories, carrousels, photos et lives confondus.La règle est simple : pour une vidéo, la vue est comptée dès qu'elle démarre, quelle que soit la durée regardée ensuite. Pour une photo ou un carrousel, elle est comptée dès que le contenu apparaît à l'écran.Chaque relecture recompte une vue, ce qui gonfle mécaniquement les totaux par rapport à l'ancien système des lectures.-TikTok : dans le même camp que MetaTikTok compte une vue après environ 1 seconde de lecture, sans pourcentage minimum ni obligation d'aller jusqu'au bout.Pour les vidéos de 3 minutes ou plus, le seuil monte à environ 3 secondes, pour éviter qu'un simple défilement gonfle artificiellement les compteurs sur les formats longs.TikTok exclut explicitement les vues du créateur sur sa propre vidéo, une règle qu'on n'a pas trouvée documentée aussi clairement côté Meta.Côté publicité, les relectures dans une même diffusion ne s'additionnent pas.-LinkedIn : la plateforme la plus opaque, et la moins stable dans le tempsLe chiffre documenté depuis plusieurs années dans la plupart des guides marketing est un seuil de 2 à 3 secondes de visionnage pour qu'une vue vidéo soit comptabilisée.LinkedIn est la seule des six plateformes à ne communiquer aucune définition claire et à jour de sa propre métrique phare.-
Vue is the new impression
Une impression, historiquement, ça voulait dire une seule chose : le contenu s'est affiché à l'écran.Aucune garantie d'attention, aucune preuve qu'un humain ait regardé quoi que ce soit. Une vue, elle, était censée signifier davantage : quelqu'un a réellement engagé le visionnage.…Retrouvez toutes les notes de l'épisode sur www.lesuperdaily.com ! Le Super Daily est le podcast quotidien sur les réseaux sociaux. Il est fabriqué avec une pluie d'amour par les équipes de Supernatifs. Nous sommes une agence social media basée à Lyon : https://supernatifs.com. Ensemble, nous aidons les entreprises à créer des relations durables et rentables avec leurs audiences. Ensemble, nous inventons, produisons et diffusons des contenus qui engagent vos collaborateurs, vos prospects et vos consommateurs.…Retrouvez toutes les notes de l'épisode sur www.lesuperdaily.com ! Le Super Daily est le podcast quotidien sur les réseaux sociaux. Il est fabriqué avec une pluie d'amour par les équipes de Supernatifs. Nous sommes une agence social media basée à Lyon : https://supernatifs.com. Ensemble, nous aidons les entreprises à créer des relations durables et rentables avec leurs audiences. Ensemble, nous inventons, produisons et diffusons des contenus qui engagent vos collaborateurs, vos prospects et vos consommateurs. Hébergé par Acast. Visitez acast.com/privacy pour plus d'informations.

durée : 00:04:09 - Le 5/7 - par : Laetitia Gayet - C'est projet qui est né avant le Covid. Pedrosa rêvait de parler de l'arrivée des Portugais sur les côtes nippones au XVIème siècle. L'auteur de L'Âge d'Or (prix fnac-France Inter en 2018) voulait le faire avec un auteur de manga singulier, Taiyo Matsumoto. Le résultat est un objet de lecture, rare. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:04:09 - Bulles de BD - par : Laetitia Gayet - C'est projet qui est né avant le Covid. Pedrosa rêvait de parler de l'arrivée des Portugais sur les côtes nippones au XVIème siècle. L'auteur de L'Âge d'Or (prix fnac-France Inter en 2018) voulait le faire avec un auteur de manga singulier, Taiyo Matsumoto. Le résultat est un objet de lecture, rare. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:04:09 - Bulles de BD - par : Laetitia Gayet - C'est projet qui est né avant le Covid. Pedrosa rêvait de parler de l'arrivée des Portugais sur les côtes nippones au XVIème siècle. L'auteur de L'Âge d'Or (prix fnac-France Inter en 2018) voulait le faire avec un auteur de manga singulier, Taiyo Matsumoto. Le résultat est un objet de lecture, rare. Vous aimez ce podcast ? Pour écouter tous les autres épisodes sans limite, rendez-vous sur Radio France.

Wie hat dir die Folge gefallen?Gut

Cory O'Hara welcomes Bryan and Kendall Howard to The Vue for a Q&A about their lives, testimonies, and calling to ministry. Get to know the Howards as they begin serving Bellevue's college ministry, and hear Bryan reflect on the mercy of Jesus in 1 Timothy 1:15–16.

durée : 00:20:12 - Le journal de 18h00 - L'inaltérable "Il va falloir se retrousser les manches" emprunté par l'actuel gouvernement, aux prises avec une équation budgétaire pleine de ronces. Annabelle Grelier remontera pour nous la piste des retraites quand Mélanie Kuszelewicz nommera les obstacles qui s'offrent à l'exécutif. - équipe : La Rédaction de France Culture, Jean-François Braun, Caroline Bennetot Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:20:12 - Les journaux de France Culture - L'inaltérable "Il va falloir se retrousser les manches" emprunté par l'actuel gouvernement, aux prises avec une équation budgétaire pleine de ronces. Annabelle Grelier remontera pour nous la piste des retraites quand Mélanie Kuszelewicz nommera les obstacles qui s'offrent à l'exécutif. - équipe : La Rédaction de France Culture, Jean-François Braun, Caroline Bennetot Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:06:22 - Jean-Michel Avias, vice-président du conseil départemental de la Drôme en charge des mobilités, était l'invité d'ici Drôme Ardèche ce lundi. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 01:15:11 Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:06:00 - La grande matinale - L'ancien Premier ministre est mort le 31 août à 97 ans. De sa naissance en Turquie à sa candidature à l'élection présidentielle de 1995, les journaux retracent sa longue carrière politique. - équipe : Marc Voinchet Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

An dëser Summerpaus.-Episod ass de "Monster"-Sommelier an Nr.1-"TikTok"-Hustler Philippe a.k.a. Argent Difficile zu Gaascht. Mir geheien den Hometrainer an d'Waasser, hänken eis un d'Paart a schwätze mat him iwwer dutch memes, de Kreativitéit-Endboss a seng Vue op d'Marie Curie.   Donieft geet et och ëm Dram-Sponsorings, onerwaarte Film-Favoritten  a Gossip aus der Gaming-bubble. Mat dobäi si Bieren, Otteren a Wëllef an och ugeknabbert Miwwelen, en aunt assault am "Atelier" an d'ënnerschätzt Hobby Frisbee.  Investigativ ewéi ëmmer gëtt journalistesch nees alles erausgeholl: Huet d'oppe Schwemm vision? Ass bei engem Veganer eigentlech all Beef fake? A wéi een Instagram life hack gëtt et, fir red flags ze erkennen?

durée : 00:13:21 - Le journal de 18h - Des centaines de voyageurs se retrouvent bloqués en gare Montparnasse, à Paris, à cause d'une panne de signalisation. Xavier Bertrand est-il vraiment candidat à la présidentielle ? Le député LFI Sébastien Delogu placé en garde à vue après une altercation avec trois policiers. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:52:10 - Le Cours de l'histoire - Quels sont les arguments historiques avancés par le pouvoir politique russe sur les liens entre l'Ukraine et la Russie ? Aux côtés d'historiennes et historiens, décryptage des discours des dirigeants russes pour comprendre comment ils lisent l'histoire ukrainienne à l'aune de leur projet politique. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

Six ans après leur rupture fracassante avec la famille royale et leur installation en Californie, le prince Harry et son épouse Meghan vont revenir vivre au Royaume-Uni, ont annoncé mercredi 19 août plusieurs médias britanniques. Après des années d'accusations et de déchirements, ce retour de Harry dans son pays avec sa famille est une véritable surprise. Écoutez l'analyse de Jérôme Carron, grand reporter chez "Point de Vue".Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.

 El cine documental es cine, pero no tiene tanta presencia en cartelera regular. ¿A qué se debe eso? Aunque parece que la respuesta es obvia, son más las causas que las que aparentan, así que en este episodio, lamentamos decirte que el asesino no es el mayordomo. Eso y más en esta entrega del mejor whodunit de cine: quién mató. Lecturas de InterésDiCaprio se queja de los cambios de la industria y la audiencia huevona que mejor se espera al streaming.“It's changing at a lightning speed,” DiCaprio said of the industry. “We're looking at a huge transition. First, documentaries disappeared from cinemas. Now, dramas only get finite time and people wait to see it on streamers. I don't know.”La recuperación apenas llega y las exhibidoras no se arriesgan, cediendo espacios antes reservados a documentales a cintas más redituables.“Even the big names in the documentary industry are now more focused on being shown on premium VOD,” Clanet said of the shift away from a theatrically-driven model. “It's the same everywhere. Distributors don't have the cashflow any more to release documentaries.”Falta de diversidad en cartelera.Mr Clifford, who has also worked for Cineworld and Vue, adds that audiences "haven't seen the wider range of films" with fewer documentaries, British films and foreign films released.El tamaño de las audiencias es pequeño.Los documentales “no son películas”.Viewers and critics tend to judge the subject matter of a doc rather than the movie itself, and criticism can feel mean-spirited because documentaries are usually viewed NOT as movies, but as important social goods made by well-intentioned people doing difficult work for very little money.Los documentales más taquilleros.Esta entrega fue traída gracias a:Productora Ejecutiva: Blanca LópezProductor Asociado: Luis Fernando Gallardo de León. ¡Bienvenido de nuevo!Co-Productor: Román RangelCo-Productor: Dale Carley. Welcome!Co-Productor: Dany Sadia. ¡Bienvenido de nuevo!Agradecimiento especial a nuestros Patreons: Adriana Fernández, Agustín Galván, Cris Mendoza, Fernando de Anda, Franky González, Jaime Rosales, Juan Espíritu, Zert, Álvaro Vázquez, Arturo Manrique, Fabiola Sándoval, Lau Berdejo, Marce, Miguel Huesca, Alejandro Alemán, Arturo Aguilar, Enrique Vázquez, Ernesto Diezmartínez, Mariana Padilla, Tania RG y Fernando Alonso.¡Gracias a nuestros suscriptores en Twitch ! Como el buen JIFF y CoyoteraxTú también puedes apoyar la creación de este y más programas y recibir crédito (para que aumentes currículum) y otros extras exclusivos en www.patreon.com/churrosypalomitas¿Quieren continuar la discusión? Tenemos nuestro canal de Discord de Charlas y Palomitas, con distintos temas, unos solo para productores del show y otros para toda la banda.

durée : 00:42:47 - Les émissions culturelles de France Culture - par : Aliette Hovine - Le 14 août 1956 disparaissait Bertolt Brecht, reconnu comme l'un des plus grands dramaturges du XXᵉ siècle. Olivier Neveux, auteur de l'essai "Brecht et les mauvais Temps Nouveaux", revient sur la carrière et les héritages de cette légende du théâtre. - équipe : Françoise Le Floch, Anne-Toscane Viudes, Théa Corler, Alice Puydebois, Louan Chailleux - invités : Olivier Neveux Professeur d'histoire et d'esthétique du théâtre à l'Ecole Normale Supérieure de Lyon. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:10:18 - Le journal de 18h - Des millions d'Européens et de passionnés du monde entier contempleront dans la soirée le ciel lors d'une rare éclipse solaire totale qui plongera très brièvement le nord de l'Espagne dans l'obscurité en plein jour. À la une aussi, un incendie mortel dans les Deux-Sèvres. - équipe : Quentin Perez de Tudela Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:17:11 - Le journal de 19h - Le retour à la maison pour près de 2 000 résidents de Lège-Cap-Ferret en Gironde touché par un mégafeu ces derniers jours, des maisons ont brûlé. La une, c'est aussi la garde à vue du cuisinier Jean Imbert ce lundi 3 août, suspecté de violences conjugales. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

A huge week for open models: Inkling (the first big US open-weight model since Gemma 4), Qwen 3.8, and Kimi 3 all dropped. Plus Vue 3.6 RC + Vapor Mode, the slow death of Stack Overflow, a decoy font that blinds AI, and the usual grab bag of fun links. Show Notes 00:00 Intro 00:44 Welcome to Syntax 03:01 Vue.js 3.6 RC Released Vue.js 3.6-RC.1 09:50 Brought to you by Sentry 10:35 Rift - new worktree alternative Rift Explaining Rift tweet 17:46 Every new browser feature MDN Plus Browser features updates CSS 'text-fit' property 23:09 Kimi K3 Released Kimi K3 release tweet Kimi K3 blog post Openrouter effective pricing 29:33 What are Open Weight Models 32:46 Kimi K3 Architecture Kimi K3's built up Cutting off access to Kimi 33:59 Qwen 3.8 Released Qwen 3.8 launch announcement 36:42 Inkling from Thinking Machines Thinking Machines release Inkling Inkling release post from Mira Murati Thinking Machines news blog post 40:45 Mole Cleanup App Mole 43:21 Moshi Terminal App Moshi app 46:05 RIP Stack Overflow Query for new questions on Stack Overflow Stack Overflow for Agents StackExchange 54:10 Scott's Foot Pedal 54:59 Scott's Feet 56:13 Decoy Font 59:16 Jurassic Park Computers 01:03:59 Ride the Tokyo Trains LIVE Hit us up on Socials! Syntax: X Instagram Tiktok LinkedIn Threads Wes: X Instagram Tiktok LinkedIn Threads Scott: X Instagram Tiktok LinkedIn Threads Randy: X Instagram YouTube Threads

durée : 00:02:29 - par : Sarah Lemoine - Le compte personnel de formation pourrait subir un nouveau tour de vis l'an prochain. Alors que les règles d'utilisation du CPF ont déjà été durcies ces dernières années, le ministre du Travail souhaite aller encore plus loin. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

Bun quitte Zig pour Rust en 11 jours à coups de Claude Code, pour 165 000$ payés par Anthropic : la réaction du créateur de Zig ne se fait pas attendre. TypeScript 7 débarque, réécrit en Go, 8 à 12x plus rapide. Entre les deux, Vidocq réimplémente Jakarta EE en souverain, le COBOL met un uppercut aux microservices, et un CTO demande à son équipe combien de temps il lui faudrait pour revenir à sa vélocité antérieure sans Claude Code. De quoi réfléchir avant le prochain rewrite. Enregistré le 17 juillet 2026 Téléchargement de l'épisode LesCastCodeurs-Episode-342.mp3 ou en vidéo sur YouTube. News Langages Est-ce qu'on peut aussi utiliser des double, des longs, ou autre pour gérer les montants monétaires en Java ? https://blog.frankel.ch/bigdecimal-vs-double/ double (IEEE 754) Usage : Calculs scientifiques, métriques, statistiques. Avantages : Très performant (matériel), idéal pour l'approximatif. Risques : Erreurs d'accumulation, égalité (==) trompeuse, NaN / -0.0. Bonnes pratiques : Utiliser une tolérance (epsilon ou ULP) pour comparer ; utiliser des algorithmes de sommation compensée (Kahan/Neumaier) pour la précision. BigDecimal Usage : Finance, comptabilité, fiscalité (précision décimale stricte). Avantages : Contrôle total des arrondis et de l'échelle. Risques : Lent (allocations), immutabilité (risque de mauvaise réaffectation), confusion equals() vs compareTo(). Bonnes pratiques : Initialiser via String ou valueOf() ; utiliser compareTo pour l'égalité. Point fixe (long) Usage : Trading, systèmes haute performance, paiements. Avantages : Très rapide, déterministe, zéro allocation. Risques : Gestion manuelle de l'échelle et des débordements (Math.addExact). Points de vigilance en production Sérialisation (JSON) : Préférer les String pour BigDecimal pour éviter la perte d'échelle. Atomicité : double n'est pas atomique ; utiliser volatile ou DoubleAdder (pour les compteurs). Tests : Toujours définir un delta ou Offset pour les tests de flottants. Bibliothèques recommandées Moneta (JSR 354) : Standard bancaire complet. decimal4j : Optimisé pour le point fixe haute performance. Apache Commons Numbers : Outils robustes pour la précision et les sommations. Typescript 7 est de sortie devblogs.microsoft.com/typescript/announcing-typescript-7-0 Performance majeure : Portage natif en Go offrant des gains de vitesse de 8x à 12x et une consommation mémoire réduite. Architecture optimisée : Utilisation du multithreading (mémoire partagée) et parallélisation native (analyse, vérification de types,émission). Nouvelles options de contrôle : Introduction des flags –checkers, –builders (parallélisation) et –singleThreaded (mode mono-cœur). Nouvel observateur de fichiers : Passage à une solution basée sur @parcel/watcher pour une meilleure réactivité et stabilité du mode –watch. Compatibilité et transition : Compatible avec les bases de code TypeScript 6.0. Utilisation du package @typescript/typescript6 recommandée pour maintenir des outils dépendants de l'ancienne API. Changements de configuration : Durcissement des défauts (ex: strict activé par défaut) et suppression de nombreuses options obsolètes (target: es5, baseUrl, etc.). Amélioration de l'expérience éditeur : Serveur de langage (LSP) plus stable avec une réduction de 80 % des erreurs et 60 % des crashs. Limitations actuelles : Support incomplet pour les frameworks utilisant des plugins de langage (Vue, Svelte, Astro, Angular) en attendant une API stable. "Java, the documentary" est sur YouTube, retraçant l'histoire du langage youtube.com/watch?v=… La vidéo n'était pas encore disponible à l'heure de l'enregistrement. Sortie officielle le 17 juillet. Avec des interviews de James Gosling, Brian Goetz, Venkat Subramaniam, et bien d'autres. Librairies What's New in 8.0 - Hibernate docs.hibernate.org/orm/8.0/whats-new L'intégration de Jakarta Persistence 4.0 apporte des nouveautés majeures comme EntityAgent (qui standardise la StatelessSession), les mappings de result set en SQL natif, et de nouvelles options de configuration de session et de requêtes (Session Creation Options, Query Options). Le support de Jakarta Data 1.1 est ajouté pour les Hibernate Data Repositories, incluant l'intégration avec les requêtes statiques JPA4, les projections @Select, et les repositories asynchrones via Jakarta Concurrency ou Hibernate Reactive. L'introduction du Graph-based Flushing remplace l'ancienne approche basée sur des heuristiques par un modèle de dépendances utilisant les contraintes relationnelles, afin d'améliorer la fiabilité des tris, la gestion des batchs et les performances globales (bien que l'ancienne méthode reste temporairement disponible). L'API ProcedureCall a été améliorée pour faciliter le casting des résultats (asResultSetOutput) et permettre la déclaration paresseuse (lazy) du mapping des ResultSet. Hibernate supporte désormais la sécurité au niveau de la ligne (Row-Level Security) de manière native pour les bases de données compatibles (PostgreSQL, Db2, SQL Server, CockroachDB) afin de gérer la visibilité en contexte multi-tenant. Une nouvelle méthode getReference() permet dorénavant de récupérer la référence d'une entité directement à partir de son natural id. Le mode Safe Mode Validator (hibernate.query.safe_mode_enabled=true) fait son apparition pour bloquer les opérations risquées comme sql(), function() ou column() dans les requêtes HQL et Criteria, ce qui est particulièrement utile pour les applications exposées aux LLMs. La gestion des associations bidirectionnelles lors de la phase de flush peut maintenant être prise en charge automatiquement par Hibernate (hibernate.bidirectionality_management=true), synchronisant la référence côté inverse de l'association. Le Subselect Fetching est considérablement amélioré, supportant dorénavant les associations "to-one" pour le bulk select fetching (au lieu de se limiter aux collections) et devenant une option de premier ordre via FetchMethod.BY_SUBQUERY. Un des papas de Cucumber et Gherkin lance Var, une alternative pour le test et le BDD var.oselvar.com Lancement de Vár : Nouvel outil de test créé pour pallier les défauts de Cucumber. Limites de Cucumber : Syntaxe Gherkin trop rigide, intégration difficile avec les exécuteurs de tests et support éditeur limité. Usage avec l'IA : Conçu spécifiquement pour vérifier que les agents IA respectent les intentions et spécifications de l'utilisateur. Fonctionnement : Utilisation du Markdown plutôt que du Gherkin ; sert à la fois de guide et d'outil de vérification. Développement assisté : Code et documentation générés en grande partie par Claude sous supervision humaine. Appel aux retours : Projet ouvert aux tests et aux critiques de la communauté. Web Une nouvelle méthode HTTP : QUERY https://kreya.app/blog/new-http-query-method-explained/ Méthode HTTP QUERY (RFC 10008) pour les recherches complexes. Problème : GET (limité par l'URL) vs POST (sémantique inadaptée). Avantages : Permet un corps de requête, sûr, idempotent et cacheable. Limites : Support infrastructurel faible, non partageable par lien, cache complexe. Usage : À réserver aux requêtes complexes si l'environnement le permet. Comment je fais du design en tant que dev backend eventuallycoding.com/p/comment-je-fais-du-design-en-tant-que-dev-backend Hugo Lassiège retrace l'évolution de son workflow de création d'interfaces en tant que développeur backend, depuis ses débuts avec Bootstrap jusqu'à l'ère de l'intelligence artificielle. L'article explique comment la structuration des éléments visuels a progressé grâce à l'Atomic Design, l'émergence des design systems et l'adoption des design tokens via un framework comme Tailwind. L'auteur détaille son processus actuel qui s'appuie fortement sur Claude Design pour générer et itérer sur des maquettes à partir d'un brief, d'un screenshot ou d'un design system de référence. Il aborde également le risque de slopification et de standardisation extrême apporté par ces outils, rappelant que si l'IA simplifie la technique, il reste crucial d'injecter de l'identité et de l'originalité pour éviter un web trop aseptisé. Data et Intelligence Artificielle De l'utilisation de SKILL.md et de "loop engineering" pour augmenter sa productivité glaforge.dev/posts/…/of-skills-and-loops-with-ai-assistance Les skills permettent d'encoder une procédure de manière répétable et automatisable Le loop engineering enlève l'humain de la boucle afin que l'agent atteigne un objectif donné de façon plus autonome Pour écrire des Codelabs (sorte de tutoriel guidé pas à pas) Guillaume a transformé une séance de création de codelab avec son agent préféré (Antigravity) en skill réutilisable pour l'écriture de ses prochains codelabs Il a également utilisé l'approche de "loop engineering" à la mode en ce moment pour que son agent IA compile, exécute, teste les instructions et le code de son codelab, pour qu'il soit complètement fonctionnel Gain estimé : passer de 2 jours de travail à moins de 2 heures ! Redeploying Claude Fable 5 anthropic.com/news/redeploying-fable-5 Anthropic a annoncé le rétablissement de l'accès à ses modèles Claude Fable 5 et Mythos 5, qui avaient été suspendus suite à des restrictions d'exportation imposées par le gouvernement américain le 12 juin 2026. Cette suspension faisait suite à un rapport d'Amazon démontrant une méthode pour contourner les garde-fous de Fable 5, lui permettant d'identifier et d'exploiter une vulnérabilité logicielle (un jailbreak). Pour y remédier, Anthropic a renforcé ses mécanismes de sécurité en déployant un nouveau classifieur capable de bloquer cette technique spécifique dans plus de 99 % des cas, acceptant en contrepartie une augmentation des faux positifs sur des requêtes bénignes. Face à l'absence de consensus sur l'évaluation des jailbreaks, Anthropic s'associe à Amazon, Microsoft, Google et d'autres partenaires pour développer un standard industriel évaluant la sévérité de ces failles selon quatre critères : gain de capacité, étendue du gain, facilité d'arsenalisation et découvrabilité. L'entreprise s'engage également à approfondir sa collaboration avec le gouvernement américain, notamment via des évaluations pré-déploiement, un partage rapide d'informations sur les failles, et des ressources dédiées à la recherche conjointe sur la sécurité de l'IA. Outillage La réécriture de Bun en Rust et la réaction du créateur de Zig bun.com/blog/bun-in-rust et andrewkelley.me/post/my-thoughts-bun-rust-rewrite.html Bun, le runtime JavaScript et TypeScript écrit à l'origine en Zig, a été entièrement réécrit en Rust pour des raisons de stabilité et de gestion de la mémoire. Cette migration massive d'un demi-million de lignes de code a été bouclée en seulement 11 jours grâce à l'utilisation intensive de Claude Code fonctionnant en parallèle, pour un coût d'API estimé à 165 000 dollars financé par Anthropic. Andrew Kelley, le créateur de Zig, a réagi publiquement en qualifiant l'ancienne base de code de Bun de "slop" remplie de hacks et de fuites mémoire accumulées par une course aux fonctionnalités. Kelley exprime son soulagement face à ce départ, expliquant que les plantages incessants de Bun devenaient un passif réputationnel toxique pour le langage Zig et sa fondation. Le rachat de Bun par Anthropic fin 2025 avait déjà mis fin aux donations financières de Bun envers la Zig Software Foundation, facilitant cette séparation. La nouvelle version Rust de Bun passe désormais la quasi-totalité des tests, réduit la taille du binaire et est déjà déployée de manière transparente en production dans Claude Code. Nouveautés de Git 2.55 github.blog/open-source/git/highlights-from-git-2-55 Support natif de FSMonitor sous Linux via inotify pour accélérer les commandes comme git status sur les grands dépôts Intégration de la compaction incrémentale MIDX (multi-pack index) dans git repack pour optimiser la réécriture des métadonnées Amélioration drastique des performances de génération des bitmaps et des pseudo-merge bitmaps lors des tâches de maintenance Nouvelle commande expérimentale git history fixup pour intégrer facilement des modifications locales dans un commit antérieur Possibilité d'exécuter des hooks configurés en parallèle pour optimiser le temps de build et de validation Utilisation d'un autostash automatique lors d'un git checkout -m en cas de conflit de fusion pour éviter de bloquer l'espace de travail Nouvelle commande git format-rev permettant de formater rapidement des commits reçus via l'entrée standard (stdin) Support du push simultané vers un groupe de remotes configuré Protection contre l'exécution de séquences de contrôle de terminal malveillantes via les flux de progression distants Vidocq, une réimplémentation souveraine et sans dépendance de Jakarta EE et Microprofile vidocq.dev/posts/vidocq-a-sovereign-jakarta-ee-and-microprofile-runtime Lancement de Vidocq : Runtime Java open source complet, compatible Jakarta EE Core Profile et Souveraineté numérique : Projet européen hébergé sur Codeberg, sous licences EUPL 1.2, EPL 2 et GPL 2.0. Standardisation totale : Implémentation fidèle des spécifications (CDI, REST, JSON, etc.), validée par 5 650 tests TCK officiels. Sécurité radicale : Zéro dépendance externe et aucune bibliothèque tierce. Aucune manipulation de bytecode à l'exécution (« magie » générée à la compilation via JDK 25). Compatible JPMS, AOT, GraalVM et Leyden CDS. Disponibilité : Projet en phase alpha, code et documentation accessibles sur vidocq.dev. Article complémentaire qui revient sur la genèse de Vidocq, en utilisant l'IA et les TCKs pour driver l'aspect spec-driven development vidocq.dev/posts/the-story-of-vidocq Le "selfware" : Guillaume s'est fait plais' en vibe-codant son propre éditeur de texte glaforge.dev/posts/…/selfware-building-my-own-text-editor-without-knowing-swift Concept de « Selfware » : création de logiciels conçus exclusivement pour soi-même, sans monétisation ni contraintes liées aux utilisateurs tiers. Le rôle de l'IA : les agents de programmation (comme Antigravity) suppriment la barrière technique de l'apprentissage des langages (Swift, APIs) pour les non-développeurs. Développement minimaliste : privilégier la performance et l'utilité directe (démarrage instantané, interface native) au détriment des fonctionnalités complexes (plugins, télémétrie, gestion de comptes). Absence de pression : libération des contraintes liées à la compatibilité, à la maintenance logicielle et aux retours utilisateurs ; le logiciel n'a besoin d'être « assez bon » que pour ses propres besoins. Incitation à l'autonomie : encourager la création d'outils sur mesure pour résoudre les frictions quotidiennes plutôt que de subir les limitations des logiciels commerciaux. Architecture Le cobol a donné un uppercut au microservices https://freedium-mirror.cfd/@maahisoft20/your-microservices-lost-to-cobol-let-that-sink-in-8ce2e236d007 Retour d'expérience sur la migration d'un système COBOL vers des microservices cloud-native qui s'est soldée par un retour en arrière après avoir constaté que le traitement batch initial était plus rapide, moins cher et plus fiable Là où le batch COBOL traitait 2.4 millions d'enregistrements en 11 minutes, le système distribué modernisé à base de message queues, retries et Kubernetes prenait 47 minutes et tombait sous la charge COBOL brille par ses caractéristiques conçues spécifiquement pour la finance comme le calcul décimal précis sans floating point errors et l'absence totale d'overhead réseau, de conteneurs ou de cold starts Rappel que distribuer un système multiplie les points de défaillance silencieux et complexifie la gestion de la cohérence transactionnelle par rapport à une exécution locale séquentielle Une invitation à se demander si les projets de décomposition en microservices apportent réellement un gain de performance de bout en bout pour l'utilisateur final ou s'ils optimisent seulement le diagramme d'architecture Méthodologies Ma meilleure question d'entretien Spring beaufume.fr/articles/spring-interview Florian beaufumé partage sa question d'entretien favorite pour évaluer des développeurs Spring de niveau intermédiaire à avancé : "Que pouvez-vous me dire sur le paramètre spring.jpa.open-in-view ?". Ce paramètre détermine l'activation du pattern Open Session In View (OSIV) qui, lorsqu'il est à true (la valeur par défaut dans Spring Boot), maintient l'un EntityManager JPA ouvert durant toute la requête HTTP. Si l'OSIV facilite le développement en évitant les fameuses LazyInitializationException lors de la sérialisation des entités en JSON, il pose d'importants problèmes de performance en provoquant des requêtes SQL non maîtrisées (comme le problème du N+1 select) en dehors de la couche service. Maintenir l'OSIV actif augmente également le temps de rétention des connexions au sein du pool de la base de données, limitant la scalabilité de l'application. La recommandation est de désactiver ce comportement en le positionnant à false, et de gérer explicitement le chargement des données requises au sein des transactions (via des DTOs, des requêtes JOIN FETCH ou des Entity Graphs) pour garder le contrôle sur les accès à la base de données. 10 points à retenir du rapport AI Engineering 2026 : The Acceleration Whiplash faros.ai/blog/ai-acceleration-whiplash-takeaways L'IA a franchi un cap et est devenue l'auteur principal du code : le taux d'acceptation du code généré est passé de 20% à 60% dans les équipes étudiées par Faros AI. La vélocité métier est bien réelle, avec une augmentation de 66% des epics livrées et une hausse de 33,7% du throughput des tâches par développeur. Ce volume cache un code churn massif (+861%), ce qui signifie qu'une quantité énorme de code est supprimée ou remplacée peu après avoir été ajoutée. La qualité en aval se dégrade fortement : les bugs par développeur ont augmenté de 54% et le nombre d'incidents par pull request a explosé de 242,7%. Le processus de code review est complètement saturé, entraînant un temps médian de relecture multiplié par cinq et une augmentation de 31,3% des PRs mergées sans aucune revue. Le système repose de plus en plus sur les développeurs seniors qui subissent une "senior engineer tax", devant relire un volume insoutenable de code à l'apparence correcte mais structurellement fragile. Contrairement à certaines hypothèses récentes de DORA, une forte maturité DevOps ne protège pas les entreprises contre cette détérioration ; le "Acceleration Whiplash" frappe de la même manière les équipes très performantes. En résumé, les outils d'IA inondent les pipelines de livraison avec un volume de code pensé pour un rythme machine, alors que les systèmes de vérification reposent toujours sur un rythme de validation humain. Loi, société et organisation Le coût d'une equipe d'engineering qui ne sait plus ce qu'elle fait dans un contexte d'augmentation de coût des coding agents https://freedium-mirror.cfd/@developer_programmer/i-spent-47-000-on-claude-code-in-90-[…]-asked-me-one-question-and-i-couldnt-answer-it-af3b203f81bb Une équipe de 8 ingénieurs a vu sa vélocité de développement exploser en utilisant Claude Code de manière intensive, jusqu'à recevoir une facture d'API salée de 47 213 $ pour seulement trois mois d'utilisation. Face à cette dépense, la question piège du CTO n'était pas sur le montant, mais sur la dépendance : "Si nous arrêtions Claude Code demain, combien de temps faudrait-il pour que notre vélocité revienne à son niveau initial ?". L'auteur s'est rendu compte qu'il était incapable de répondre car son équipe, en particulier les profils juniors, avait commencé à perdre l'habitude de concevoir et d'implémenter des fonctionnalités complexes sans l'aide permanente d'un agent. Le deuxième risque stratégique soulevé est celui de la dépendance tarifaire et du vendor lock-in : si l'outil devient une infrastructure indispensable au quotidien, l'entreprise perd tout pouvoir de négociation face aux augmentations de prix de l'éditeur d'IA. Pour éviter que l'IA ne devienne une béquille qui atrophie les compétences de l'équipe, l'article suggère de poser des limites budgétaires strictes, d'organiser régulièrement des sprints sans IA ("AI-free sprints") et de concevoir des processus de développement portables. Retour de Nicolas Delsaux sur jqwik qui donne une perspective plus complète concernant jqwik, il me semble que vous oubliez (comme tous les gens qui parlent de LLM dans "l'industrie") que l'auteur n'a pas fait ça juste pour faire chier le monde, mais parce que ces outils ont des externalités incroyablement négatives, ce dont l'auteur s'explique dans son blog (blog.johanneslink.net/2026/06/09/the-jqwik-anti-ai-affair) Vous oubliez également de signaler que le ticket (github.com/jqwik-team/jqwik/issues/708) par lequel un utilisateur se plaint de cette fonctionnalité a été écrit par un agent. N'oubliez pas non plus que l'enthousiasme pour ces technologies n'est en fait pas universel, et que ces technologies sont loin d'être inévitables (les gains de vitesse ne sont, d'après circle CI - circleci.com/resources/2026-state-of-software-delivery, pas des gains de productivité ) OkHttp, Okio, Retrofit et SQLDelight rejoignent Commonhaus ! commonhaus.org/activity/315.html La fondation Commonhaus, via une publication de Andres Almiray, annonce l'arrivée de quatre projets majeurs de l'écosystème Java et Kotlin : OkHttp, Okio, Retrofit et SQLDelight. Ces projets, initialement créés chez Square (devenu Block), sont désormais regroupés et gérés sous la bannière lysine.dev au sein de la fondation. Jesse Wilson et Jake Wharton, créateurs et mainteneurs historiques de ces outils, rejoignent Commonhaus en tant que leaders de lysine.dev. Suite à leur départ de Block, ils expliquent avoir choisi Commonhaus pour offrir à leur immense communauté d'utilisateurs un cadre de gouvernance pérenne, stable et digne de confiance. Conférences La liste des conférences provenant de Developers Conferences Agenda/List par Aurélie Vache et contributeurs : 28-30 août 2026 : State of the Map - Champs-sur-Marne (France) 4 septembre 2026 : JUG Summer Camp 2026 - La Rochelle (France) 10-11 septembre 2026 : Nantes Craft - Nantes (France) 17 septembre 2026 : dotAI - Paris (France) 17-18 septembre 2026 : API Platform Conference 2026 - Lille (France) 18 septembre 2026 : WordCamp Bretagne - Rennes (France) 18 septembre 2026 : dotJS - Paris (France) 18 septembre 2026 : WordCamp Bretagne - Rennes (France) 22 septembre 2026 : Salon Data 2026 - Nantes (France) 22-23 septembre 2026 : Agile en Seine & IA 2026 - Paris (France) 24 septembre 2026 : aMP Day Montpellier 2026 - Montpellier (France) 24 septembre 2026 : OWASP AppSec Days France 2026 - Paris (France) 24 septembre 2026 : PlatformCon Paris - Paris (France) 24 septembre 2026 : React Native Connection 2026 - Paris (France) 24-26 septembre 2026 : Paris Web 2026 - Paris (France) 25 septembre 2026 : SAP Inside Track Paris 2026 - Paris (France) 28-29 septembre 2026 : 4th Tech Summit on AI & Robotics - Paris (France) & Online 1 octobre 2026 : WAX 2026 - Marseille (France) 1-2 octobre 2026 : Volcamp - Clermont-Ferrand (France) 2 octobre 2026 : DevFest Perros-Guirec 2026 - Perros-Guirec (France) 5-9 octobre 2026 : Devoxx Belgium - Antwerp (Belgium) 8-9 octobre 2026 : Forum PHP 2026 - Marne-la-Vallée (France) 12 octobre 2026 : Dev With AI - Paris (France) 22-23 octobre 2026 : Agile Tour Bordeaux 2026 - Bordeaux (France) 26 octobre 2026 : Agile Tour Montpellier - Montpellier (France) 27-29 octobre 2026 : Directions EMEA 2026 - Paris (France) 29-30 octobre 2026 : Campus Agile Grenoble - Grenoble (France) 29-30 octobre 2026 : BDX I/O 2026 - Bordeaux (France) 29-30 octobre 2026 : Agile Tour Nantais 2026 - Nantes (France) 29 octobre 2026-1 novembre 2026 : Pycon FR - Biarritz (France) 30 octobre 2026 : Cloud Nord 2026 - Lille (France) 4-5 novembre 2026 : Devoxx Morocco - Casablanca (Morocco) 14-15 novembre 2026 : Capitole du Libre - Toulouse (France) 19 novembre 2026 : DevFest Toulouse 2026 - Toulouse (France) 19 novembre 2026 : Agile Laval 2026 - Laval (France) 19 novembre 2026 : OVHcloud Summit - Paris (France) 19 novembre 2026 : Codeurs en Seine - Rouen (France) 27 novembre 2026 : DevFest Paris 2026 - Paris (France) 1-3 décembre 2026 : Apidays Paris - Paris (France) 2-3 décembre 2026 : Cloud Native AI Summit Europe - Paris (France) 4 décembre 2026 : DevFest Lyon 2026 - Lyon (France) 4 décembre 2026 : DevFest Dijon 2026 - Dijon (France) 9-10 décembre 2026 : OpenSource Expérience - Paris (France) 9-10 décembre 2026 : DevOps REX - Paris (France) 10 décembre 2026 : KCD Provence - Aix-en-Provence (France) 10 décembre 2026 : DevCon 28 : sécurité | post-quantique | hacking édition 2027 - Paris (France) 14-16 janvier 2027 : SnowCamp 2027 - Grenoble (France) 7-9 avril 2027 : Devoxx France 2027 - Paris (France) 3 juin 2027 : Cloud Native Days France 2027 - Paris (France) Nous contacter Pour réagir à cet épisode, venez discuter sur le groupe Google https://groups.google.com/group/lescastcodeurs Contactez-nous via X/twitter https://twitter.com/lescastcodeurs ou Bluesky https://bsky.app/profile/lescastcodeurs.com Faire un crowdcast ou une crowdquestion Soutenez Les Cast Codeurs sur Patreon https://www.patreon.com/LesCastCodeurs Tous les épisodes et toutes les infos sur https://lescastcodeurs.com/

durée : 00:29:33 - L'Entretien archéologique - par : Antoine Beauchamp - Entre hauts fonds, tempêtes et brumes, les eaux de Saint-Pierre-et-Miquelon ont englouti plus de 1600 navires depuis 1790. Que nous apprend l'archéologie sous-marine sur ces drames qui refont surface dans une exposition au Musée Les Pêcheries de Fécamp ? - équipe : Sandrine Chapron, Eve Etienne, Bruno Sensini, Olivier Bétard - invités : Cécile Sauvage Archéologue au Drassm et responsable du projet de fouille de Saint-Vaast-la-Hougue, Nadège Sébille Chargée des collections du Musée les Pêcheries de Fécamp Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:02:32 - Debout la Terre - par : Anne-Claire Poirier - Pour sa dernière chronique, Anne-Claire Poirier (Vert) a choisi un sujet qui l'énerve au plus haut point : l'obsolescence programmée. Une pratique encadrée par la loi mais complexe à démontrer pour les tribunaux. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:02:05 - Les Matins de France Culture - par : Astrid de Villaines - À mesure que l'élection présidentielle approche, vous entendrez de plus en plus "Espérons que l'élection présidentielle aborde ce sujet", ou pire, "Espérons que la présidentielle règle ce sujet". Comme s'il fallait remettre à plus tard tous les débats. - équipe : Félicie Faugère Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

En histoire, en général, plus c'est ancien, moins on sait de choses. Mais ça ne veut pas dire qu'on ne sait rien non plus ! Et c'est particulièrement le cas quand on parle d'architecture et de constructions en Égypte ancienne : certains disent qu'on ne sait pas du tout comment les Égyptiens ont construit des monuments comme les pyramides ; que leurs outils étaient trop rudimentaires pour bâtir de tels ouvrages… Alors qu'en fait, même s'il y a bien sûr des zones d'ombres, on est loin de ne rien connaître de tout ça ! Alors pour y voir un peu plus clair, j'ai eu le plaisir de recevoir dans un nouvel entretien historique Franck Monnier, un égyptologue spécialiste de l'architecture égyptienne ! Franck a écrit de nombreux articles et ouvrages sur le sujet, et il a aussi d'autres casquettes : il fait des restitutions 3D de sites archéologiques, et il est très actif pour lutter contre les idées reçues et les fake news. Et ça tombe bien, parce que quand on parle d'architecture égyptienne, il y en a beaucoup ! ➤ Pour en savoir plus et suivre le travail de Franck : ➜ Découvrez sa page Facebook : https://www.facebook.com/profile.php?id=100005396808455 ➜ Retrouvez ses articles sur Academia : https://cnrs.academia.edu/FranckMonnier ➜ Explorez le palais de Malqatta : https://antiqueprofil.wixsite.com/palais3d

durée : 00:37:25 - Le 18/20 : un jour dans le monde - par : Fabienne Sintes - Il a neigé ces derniers jours au Svalbard, entre la Norvège et le pôle Nord. Un peu de frais en pleine vague de chaleur. Mais ce décor arctique raconte aussi l'un des grands bouleversements du climat : la fonte des glaciers s'accélère, jusque dans les régions les plus froides de la planète. - équipe : Philippe Lefébure, Nathalie Poitevin, Thomas Lenglain, Mathias Dubois, Oriane Laromiguière - invités : Heïdi Sevestre Glaciologue française Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:37:25 - InterNational - par : Fabienne Sintes - Il a neigé ces derniers jours au Svalbard, entre la Norvège et le pôle Nord. Un peu de frais en pleine vague de chaleur. Mais ce décor arctique raconte aussi l'un des grands bouleversements du climat : la fonte des glaciers s'accélère, jusque dans les régions les plus froides de la planète. - équipe : Philippe Lefébure, Nathalie Poitevin, Thomas Lenglain, Mathias Dubois, Oriane Laromiguière - invités : Heïdi Sevestre Glaciologue française Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:53:19 - Les informés de franceinfo - Tous les soirs, les informés de franceinfo débattent de l'actualité autour de Victor Matet. Vous aimez ce podcast ? Pour écouter tous les autres épisodes sans limite, rendez-vous sur Radio France.

We're announcing AIEWF speakers this week! Take the AI Engineering Survey!Today's guest Ethan first joined us for the LS Paper Club as the lead on NVIDIA Cosmos World Model, but then joined xAI and built Grok Imagine in 3 months:He comes back on Latent Space with some nuclear hot takes: that Video Models primarily get their intelligence from LLMs, not from training on video data, and that the next frontier for truly interactive, realtime, long-horizon world models is to work on LLMs (perhaps Interaction Models as well…)Put it this way: In the near term, the next Sora won't be a better video model, but a video agent.Generative Media may more closely follow the evolution of AI coding which went from focusing on one-shot output performance and cost, to multiturn reasoning and planning models for agents and systems that can plan, edit, test, debug, and submit PRs.At a certain point, coding models got so good that the only significant next step to improve performance was handling the orchestration of these models.Now as the performance of video models increases significantly across realism, consistency, & prompt adherence while becoming more cost efficient, the next evolution of video generation may also be systems that can plan, generate, edit, critique, and iterate across an entire creative task. In this episode, Ethan joins swyx and Vibhu to unpack what it actually takes to build frontier image and video systems: data, VAEs, diffusion transformers, audio-video alignment, inference speedups, and the hidden cost of storing and moving massive video datasets. From building NVIDIA's Cosmos world model to joining xAI as Grok Imagine was being built from zero to one, Ethan He has been at the center of some of the most important work in video generation, multimodal models, and real-time world models.We go deep on Grok Imagine, how a small xAI team shipped its first multimodal video model in three months, why iteration speed matters more than almost anything in model development, and why many of the biggest gains come from fixing tiny bugs in data and training pipelines. Flipbook: The future of VideomaxxingVideo agents are almost a sure bet to be the trend in the coming year. We end with a glance at what's beyond video agents:Flipbook caused a minor sensation this year when it was released, but most treat it as a fun demo. Ethan takes it very seriously — with the speed and cost of inference coming down every year, the future of custom video JIT UI is closer than you think. We talked about why videogen models may become the front end of AI, how generative UI could replace traditional HTML/CSS, why world models need to be real-time, interactive, and long-horizon, and why the future of video generation may depend more on language models and agents than on diffusion alone.We discuss:* Why fast iteration mattered more than meetings* Why small training bugs can drive huge model quality gains* Why coding models may make compute the bottleneck again* How image and video models are trained with synthetic captions* The role of VAEs and latent space in frontier video models* Why image models are the foundation for video models* The tradeoff between temporal compression and real-time interactivity* Flipbook, Neural OS, and the future of generative UI* Why future interfaces may go from user intent to pixels* The hidden cost of training video models: storage, egress, and GPU hours* How step distillation and consistency models (like OpenAI sCM) makes video inference orders of magnitude faster* Grok Imagine 0.9 and large-scale audio-video generation* Why audio-video alignment is harder than text-video alignment* Ethan's definition of world models* Reference-to-video, video extension, and long-context video generation* Why xAI's research communication undersells Grok Imagine* How xAI culture shaped the speed of development* AI watermarking, SynthID, and detecting generated media* Why prompt rewriting matters for video models* Grok Imagine Agent and the rise of video agents* Why language models may unlock better video generation* Robotics, physical AI, and embodied world models* Why Ethan left xAI and shifted focus toward LLMs* Self-managed context, memory, and the next frontier for language modelsEthan He* LinkedIn: https://www.linkedin.com/in/ethanhe42* X: https://x.com/EthanHe_42Timestamps00:00:00 Introduction00:01:25 From NVIDIA Cosmos to xAI00:03:24 Building Grok Imagine from Zero to One00:10:07 How Image and Video Models Are Trained00:18:53 Video Compression, VAEs, and Real-Time Tradeoffs00:22:10 Generative UI, Flipbook, and Neural OS00:32:10 The Cost of Training Large Video Models00:37:04 Distillation, GANs, and Fast Video Inference00:41:21 Audio-Video Generation and Grok Imagine 0.900:48:34 What Makes a World Model?00:55:51 Reference Videos, Long Context, and Video Memory01:00:11 xAI Culture, Research, and First-Principles Building01:09:45 AI Safety, Watermarking, and Prompt Rewriting01:13:10 Video Agents and AI-Assisted Creation01:27:32 Why Language Models Unlock Better Video01:31:15 Robotics, Physical AI, and Embodied World Models01:32:38 Why Ethan Left xAI01:34:16 Self-Managed Context and the Future of LLMs01:38:43 Ethan's Career Path and Closing ThoughtsTranscriptIntroduction: Ethan He, Latent Space, and the Path to xAISwyx [00:00:00]: We're here in the studio with Ethan He, most recently of xAI. Welcome.Ethan [00:00:10]: Thank you. Glad being here.Swyx [00:00:11]: We're also here with Vibhu. you were first coming to us or joining the latent space world because you were working on Kosmos at NVIDIA, and you did a paper. We loved it. you presented it as well, so thank you for doing that.Ethan [00:00:23]: I've actually, I also presented the MoEs twice at latent space.Swyx [00:00:29]: How did you actually hear about us? Did we reach out to you? Is that how it worked?Ethan [00:00:33]: No, actually, I-- the community. Like I realized, oh, there is this online community that people talk about AI and also learn from each other through papers every week through the Paperclip. It's very nice.Ethan [00:00:49]: I learned a lot.Swyx [00:00:49]: I think three years stop. We haven't stopped even on Christmas and New Years. many weeks I want to stop but it keeps going.Vibhu [00:00:58]: No, that was good. I think you had posted that you worked on a paper, and I was “Oh, very cool. We have Paperclip. Present then.”Vibhu [00:01:04]: But I might have reached out to you after.Swyx [00:01:05]: you-- because it's an amateur club, right?Swyx [00:01:08]: so it's very unusual and but we have sometimes paper authors come by and actually explain the paper. Today we just did, the poolside paper, which was apparently very good.Vibhu [00:01:18]: Came out yesterday.Vibhu [00:01:19]: pretty interesting, right? Fully open. They talk about everything, systems. So it's a good one. We'll, we'll recommend people to read it.Swyx [00:01:25]: Bring us up to speed on your transition to xAI, ‘cause I actually don't even know when you joined. just like tell the, tell the story about the sort of transition.From NVIDIA Cosmos to xAI: Scaling Video and World ModelsEthan [00:01:34]: Before xAI, I was working on Kosmos world model as in-- at NVIDIA. So Kosmos is, it's a giant video foundation models that can-- that aims to simulate the world and for-- it serves as a foundation of-- for all of the roboticists to build on top of. There, once I built the Kosmos one, I realized as this thing also has a scaling law similar to language model, we need to scale up the video models further. that's, that's why I realized I need to move to somewhere with much more compute resources. That's how ISwyx [00:02:13]: Than NVIDIA?Vibhu [00:02:14]: The GPU rich came themselves.Vibhu [00:02:19]: And timeline-wise, when was Kosmo? It was pretty early, right? It was open world model, open paper, everything.Ethan [00:02:25]: It was end of twenty-four.Vibhu [00:02:28]: End of twenty-four.Ethan [00:02:30]: Then at mid twenty-five, I moved to xAI. At that time-- I joined about the time when xAI was about to build video models and in multi-model models. There were no infra, no data, and no model, and it just-- as a few engineers, we built it in three months and released the first model, Grok Imagine zero point nine.Ethan [00:02:55]: And since then, I keep working on video models and move more from training and to post-training of the video models. For example, like a reference to videos, kind of like the cameo feature and, video extensions. And, before I left, I worked on a world model, leading a small team to focus on the real-time long horizon video generation.Building Grok Imagine From Scratch in Three MonthsSwyx [00:03:24]: Can you give like a rough roadmap of okay, you're on a brand-new team. Grok previously was only text, or they partnered with BFL for their image gen stuff. What do you-- what are the building blocks, right? You have compute, data you can procure somewhere. Like just what are like the sequence of things that people should think about when you're setting up a new team?Vibhu [00:03:43]: actually even deeper, not just data you can procure. You guys had to go through getting the data too, right? So you shipped it pretty fast, but yeahSwyx [00:03:51]: three months is likeVibhu [00:03:52]: From everythingSwyx [00:03:52]: actually like very surprisingly fast.Ethan [00:03:55]: One thing I say like thanks to my experience at NVIDIA, ‘cause first time when we were building Kosmos together, we built it, for about a year. So this is like the second time I do it. Roughly have an idea, what to do. I say the most important thing is the talent. Everyone were very strong and clever, very close with each other towards a common goal. So that speed up things a lot. So you reduce the communication bandwidth among people, and everyone can work towards the same goal. It's, it's like every day there's not that much meetings on the calendar, like maybe like a, like a sync a day, and after that it's, it's just all building. It was pretty fun at that time.Ethan [00:04:47]: And another thing is that xAI has very strong foundations of like data inference, model inference, and the supporting there can help the model develop a lot. When I look at, training models, I don't so actually the top important thing is like how many, how many iterations can you do, per day? and the more iteration can you do, you can, you can train the model much faster. So if you have very strong infra and you have a lot of compute, you can, you can train these models in very short period of time. That can give you a much larger buffer to, for errors, and it also gives you the opportunity to spot more bugs.Iteration Speed, Compute, and Debugging Model PipelinesSwyx [00:05:46]: What is an iteration? Is it like a few hundred steps or what are youEthan [00:05:50]: Let's say just the train-training the model, like from acquire new data and maybe design new algorithms and train a new model, maybe at smaller scale orSwyx [00:06:01]: So cycle time for like any hyperparam that you're searching.Ethan [00:06:04]: Cycle time and tune to like eval this model. Is this model better than my previous iteration?Ethan [00:06:11]: SoSwyx [00:06:11]: So it's like before you, someone had already set this up that you can iterate very quickly.Ethan [00:06:15]: I think the foundation there is extremely good forDeveloping and research models.Ethan [00:06:23]: And often I find is it-- this is kind of boring, but like a lot of the improvements does not come from new algorithms. It comes from finding small bugs here and there in the data pipeline, in the, in the model training pipeline. Those give, those give the biggest boost to the model quality.Vibhu [00:06:46]: It's interesting, right? So you say it's like small team, less communication bandwidth, but also a lot of quality is like find little bugs. It seems counterintuitive, right? You have a lot of people, you can iron out more of those, but it's interesting to see the other side, right?Swyx [00:07:00]: I also wonder, have you-- do you try using LLMs to look for bugs? I don't know.Ethan [00:07:05]: I remember at that time it was mid two thousand and twenty-five, so it's the coding model wasn't quite there yet. I remem- I remember like December two thousand and twenty-five, it was extremely good. Yeah, I've been, I've been using it at that time. It's, it's helpful. sometimes it produce codes that are kind of difficult to maintain, even though like the first time it built something extremely fast. But it gave the, like a spaghetti code, thousands of lines that I couldn't maintain, and the LLM itself couldn't figure out what's, what's wrong and how to improve on top of it. But now I find it much better. Yeah, I want to bring up another point here is now coding models are much more efficient and can help us implement stuff much faster. Compute might become a bottleneck again because previously, like if you want to train a new model, say you want to generate new synthetic data and then or write a new algorithm, it might take a few weeks. And during that period of time, you don't-- you might not have experiments to run. But now you can build that thing within a few hours, then you can immediately train a model.Ethan [00:08:24]: Now you have to have enough compute to try all of the ideas. So compute might be the bottleneck of iterating speed again.Swyx [00:08:36]: yeah, I actually, honestly, I think it's like kind of a stressful job because you're “Well, I should be trying everything, and if I'm not, then I'm not doing my job well.”Vibhu [00:08:48]: there's also the stress of you're eating thousands of GPUs per hour, which is very expensive and, compute can go to other researchers.Swyx [00:08:56]: You got the daddy Elon toVibhu [00:08:57]: You got daddy Elon.Ethan [00:08:59]: It wasVibhu [00:09:00]: But there's still finite amount of compute, like you want to use it, you want to use it well, you want more of it.Ethan [00:09:06]: That was quite stressful indeed. Yeah, I think one thing is the-- with coding models now, like a lot of these jobs can be automated, which is much better. A second, it's a, it's a marathon, so you got to maintain good health and, a regular schedule.Vibhu [00:09:28]: It's, it's hard to hear that when you shift from zero to nothing in two months.Swyx [00:09:32]: and, I think obviously the culture at xAI is very famously, people work very hard. one thing I did want to dive into, in our-- in the notes that you, that you sent ahead of time, you had specific comments about the cost of Video Gen training. presumably this is on the Colossus-1, right? the two hundred megawatt cluster. Any whatever you want to just share on that.Vibhu [00:09:54]: I think there's, there's three things we're talking about, right? So there's Video Gen, there's also the Image Gen model that you put out. Do you want to like complete the, okay, so zero to one, you have a few months. Just what are the stages of create Image Gen model?Swyx [00:10:06]: Oh, yeah, maybe I got distracted.How Image and Video Models Are Trained: Synthetic Captions, Tokenizers, and VAEsVibhu [00:10:07]: Sorry. and then, from there's Video Gen, there's Audio Gen. Would love to get into those next. But what is that first few months like? So small team, a lot of bugs, iterations, but what does it look like? Do we take something off the shelf? Do we just get data compute? What's, what's the few months like? How do you go to state-art Image Gen model? How do you just start?Ethan [00:10:28]: I cannot comment specifically how xAI did, but it's, it's a quite standard process. I can draw some, examples from Cosmos. So mainly it's building a video model, you actually need to build a image model first. And building these two models, the data you need is a hundred percent synthetic pair of language and image or language to video. Because on the, on the internet, actually, the videos don't naturally associate with text. So you can say, oh, like on YouTube, you have the title and you have the description and the commentsSwyx [00:11:11]: TitleEthan [00:11:11]: of a video, but usually they're not relevant to the video itself. And say maybe like the video is a natural scene of mountains or something, and the title is, I'm so happy today.Ethan [00:11:26]: So they have they have no correlation at all. So the first step is to, you have to generate synthetic pair of language with the videos. So you gather videos from the internet, and you use a VLM to caption the videos. So that part, here's a question, like how do you, how do you gather VLM to begin with? So if there's noSwyx [00:11:55]: You, so you fuse the model, right? LikeEthan [00:11:57]: Say if there's no like VLM exists, like how do you generate the text to the beginning, right? It's, it's impossible.Swyx [00:12:04]: I see.Ethan [00:12:05]: In the beginning, it's like you ask human to describe the video as detailed as possible.For example, you ask them to describe everything, like all objects, all characters, and all interaction and dialogues in the, in the videos. So that's in the protocol of Cosmos labeling. We require the objective we give to the labelers was that you have to describe the video as detailed as possible, such that a blind person hears a blob of text can reconstruct what the video is like from their head.Swyx [00:12:43]: Video or image? You're talking about images.Ethan [00:12:44]: Video or image, either one of them.Vibhu [00:12:47]: This was pretty common when we went from clip and DALL-E, right?Vibhu [00:12:51]: It's all training on really detailed captioning of images. So same is applied to video, but insteadEthan [00:12:57]: same appliedVibhu [00:12:57]: of using multimodal model to pass in video images and write rich descriptions, you can alsoSwyx [00:13:04]: I think there's this traditional perspective of supervised, or, very highly human curated thing. I feel like there's a unlock with unsupervised, right? Where like you have enough to bootstrap that you can just throw common corpus on it or, whatever. like unsupervised vision and language pairing, right? Like where you just have, interspersed image and text and it just learns. To me, that is the VLM breakthrough that is different from the clip, different from the LM era.Ethan [00:13:36]: It's interesting to see that you kind of need both data.Ethan [00:13:41]: For example, for theSwyx [00:13:41]: You need it to bootstrap it up. YeahEthan [00:13:43]: for the generative model training, there's also usually like a small percentage of unlabeled data. So the model is instructed to generate a video without any text instruction. That can also help the model generalize. So after this stage of generative synthetic pair, so, one important common step is to train a compressor or a tokenizer of the image or videos. So because, if you train-- If you can technically, theoretically train image or video models on pure pixels, but the problem is that the, it's, it's a lot of tokens. So like one image, it's, a thousand by a thousand, it's like one million tokens, one million pixels. It's impossible to train transformer on that. So it's, you need to train a tokenizer, which can go from image to latent space and latent space back to image.Swyx [00:14:45]: That's why we named the podcast.Swyx [00:14:48]: But, basically, you're talking about vocabulary science.Ethan [00:14:50]: so vocab.Swyx [00:14:51]: And so, what is, what is imp-- like a million is impossible?Ethan [00:14:54]: In generative models, the vocab is continuous. It's a continuous space. We can think about like you map an image to a vector. It's a, it's a fixed length vector. It's sixteen or forty-eight, something like that. And then you map that vector back to the image space. And the mapping is, has-- The mapping is patch-based. So you say you haveEthan [00:15:22]: a sixteen by sixteen patch and you match, you map that patch of pixels into this latent space.Swyx [00:15:29]: We've covered thisVibhu [00:15:30]: This is like the vision transformersSwyx [00:15:32]: VAEs,Ethan [00:15:33]: VAEs.Vibhu [00:15:34]: You basically compress your input, you do your generation, you're reasoning all that generation in smaller dimension, and then you project back out.Swyx [00:15:43]: VAE is a form compression, but I think the for me, the patching thing is from VIT, right?Ethan [00:15:48]: You can make those.Swyx [00:15:49]: Literally the, yeah, the paper is titled like sixteen by sixteen is all you need. something like that. and then I think also, people make a lot of comparisons with this kind of patching with convolutions.Swyx [00:16:02]: Which is you're, you're kind of re- reconstructing the old paradigm with the new.Ethan [00:16:05]: Actually, in VAEs, there are, there are both convolution networks and transformers. You can actually do both.Ethan [00:16:14]: After this VAE, so what you've got is you've got latent space tokens and you've got the language tokens. So now the training of the diffusion transformer, usually generative models use diffusion transformers. It is actually quite standard. It's, it's very similar to how you train a language transformer models. It's not that much difference. It's just the tokens, the visual tokens in, visual tokens out. The only difference is there's a denoising process. So you train the model to unmask some of the noise. So you add, you add random noise to the visual tokens, and then you train the model to remove those noise to generate the clean tokens. Any inference, the model can iteratively remove noise from a hundred percent noise.Swyx [00:17:12]: And then there's also, to speed things along on the tech tree of diffusion, there's CFG, and then there's, there's also, latent diffusion that, there's, there's someone in there. I think, somewhere along the line, obviously, like stability and all these other guys, pioneered a lot of this, architecture. I don't know if you want to get into that or just, or do the video side up to you.Bootstrapping Video from Image Models and Temporal CompressionEthan [00:17:37]: After you train such model, such image model, the reason it's a, it's a foundation for video models is that image models are cheaper to train, and they have much denser connection between language and text. So, sorry, language and images. For example, you train a billion, you train on a billion images, and there's a mapping from the text to the image. And the cost to train the same, like the, a billion, a billion text to a billion videos, that's much more expensive because videosNaturally have more tokens than images. Because the diffusion models, their understanding of, language purely come from this mapping. So if you don't have enough mapping, so if you only train on like a ten million videos or something, there-- you might not see enough language tokens in your training, so your model does not understand human intention enough. So that's why you really-- you train-- you first train this image diffusion models, and then you bootstrap the video model from there.Swyx [00:18:53]: One thing I did want to ask, because I-- actually, I think you're, you're the first per-- video model person I've ever talked to, I think. we've, we've like talked to Luma and all those folks. There's all these tricks in video compression where basically frame by frame there's not that much difference, so actually you don't have to regenerate or save the whole frame, right? but I think MP4 compression or something else like that.Swyx [00:19:16]: is it tempting to use that? Or as far as I can tell, everyone just treats it as, “No, we would just generate every frame.” Is that roughly the state-art?Ethan [00:19:27]: There are a few different approaches. Let's say first, like you want to just directly use MP4 compression and use that as the tokens for the transformers to train, right? So people actually have tried that, but the main challenge is the latent space for the MP4 tokens were not, were not very comprehensible for the models. It's, it's extremely hard to train on that. And there's aEthan [00:20:01]: So that's why they created VAEs, which creates more continuous, latent space, so the models can understand that latent space and learn from it much easier. Even within the VAEs, there are different difficulties of the latent space. So you can imagine something the simplest, the most naive VAE is like you have an image, and you just shuffle all of the images into a, into a vector. So you don't need to train any VAEs, right? But that latent space is extremely hard for models to train on top of. That's why there are some debate on like how do you compress the tokens. So you mentioned like you can compress frame by frame. Also, you can compress, the temporal dimension.Ethan [00:20:52]: The difference is if you compress the temporal dimension, you get a much higher compression rate. Because there's temporal redundancy between frames, because, this frame and the last frame, likely they are mostly similar, so there's only some small difference. for example, I think in 12.1 VAE, they have like a eight by eight by four compression rate. So the four temporal tokens are compressed into one tokens. That can save a lot of, save a lot of the context length. If you do it frame by frame, you have to do maybe like eight by eight by one. Your context length will be four times larger. That being said, the benefit of the frame-- per frame compression, we might come back to this later, is, real-timeness and interactivity. ‘Cause if you, if you strain the output of the model, frame by frame, you can-- the model can respond to any user request immediately. So if you have like a temporal four compression, four times compression, thenSwyx [00:22:06]: It might be laggyEthan [00:22:07]: there's a lag there in nature.Swyx [00:22:10]: So you're very pilled on this. let's just go ahead and bring it up ‘cause we have the visual prepared anyway. There's some frontier applications of real-time video gen. So Flipbook is one of the examples that went viral recently, right? What is Flipbook?Real-Time Generative UI: Flipbook, Neural OS, and Diffusion Front EndsEthan [00:22:23]: Flipbook is kind of like a web brow- web browser. You can see like it has the web bro- browser UI on top. The difference is all of the UIs are generated by generative image model in real time, and anything here are fake. But you can, you can explore inside this wor- this imaginary world. Say like we-- here we have engineering the Great Pyramid. Like the model generates this for us to understand how it works, and if we want to navigate around and understand further, we can click on some of the, some of the description here, and the model will generate a new page, new subpage describing the details we want to know about.Swyx [00:23:14]: So it's basically kind of we're playing a video, but it's pausing for our next interaction, and then it just plays the next thing based on our interaction.Swyx [00:23:23]: Which is kind of cool.Vibhu [00:23:25]: and you kind of decide your story. So this was, how do you make a pyramid? levering technique seemed interesting, right? It shows how do you take Okay, I want to know what is thisSwyx [00:23:35]: The demo, the demo tweet had more animation between frames.Vibhu [00:23:38]: I think it's just skipping,Swyx [00:23:39]: Oh, it's just skipping a lot of frames.Ethan [00:23:40]: they also have a video modeVibhu [00:23:42]: It takes a lot. There's a lot of peopleEthan [00:23:42]: but, a lot of people are using it.Ethan [00:23:45]: So it's not available.Vibhu [00:23:46]: There's a live video stream. We can try,Swyx [00:23:50]: So this is an example of the kind of future that you see at the extreme. We don't-- we're obviously not in it today.Swyx [00:23:56]: But in a world where inference is completely free this is better than generating code and text?Ethan [00:24:02]: So this is, this is a final state of where Viva will be at for word model, I think. Imagine internet doesn't exist, and then you type in google.com. Like what should, what should, what should a model show you?the model can imagine something, and this is what the model imagine. And these web pages, they completely do not exist. So I think as the inference costs come down, we are going to have generative UI for everything. If you think about how the coding model works, so they write code for a web page, and they render the code might be con- converted into binary, and the binary render the pixels on the screen. So we in machine learning, every time we have some breakthrough, obviously it's, it's more intuit. So why don't we have like user instruction to the pixel directly? So the generative UI will be user intention to the pixels directly. And say like even if I want email, let's say everyone have the same interface, but I want, I want it slightly different. I want the email to show to me like a TikTok, so I can swipe left and right for the emails. And or maybe you want something else. We can have completely different things. Or like I have I'm looking at, Instagram stories, and I don't like the Like button. I always may click it. And, generative UI resolved it. So it's going to be a revolutionary replacement of the interface. So in the future, we might have much more powerfulEthan [00:25:50]: LLMs and coding models running behind the scene. And in the, in the front-end, the diffusion model will actually be the front-end to show stuff to you. That's how I imagine it.Swyx [00:26:02]: Diffusion front-end, deterministic back-end.Swyx [00:26:04]: Something like that. I find that very expensive, but,Vibhu [00:26:08]: I find it interesting you called LLMs writing code on the back end deterministic, but okay.Swyx [00:26:14]: you write it onceVibhu [00:26:15]: Compare it toSwyx [00:26:16]: And then you execute.Ethan [00:26:17]: If you think about the cost, say, let's say H100 costs $1 per hour, and if you use this eight hours a day and thirty days, so, every month you're paying this two forty, you'll actually not wanna pay for that. That's even more expensive than Cloud Code Max. But if you think about the compute costs come down like two times every year, and I think the future will likely arrive like within few years.Vibhu [00:26:49]: It's everything, right? compute cost comes down, compute gets faster, model gets smarterEthan [00:26:54]: More efficientVibhu [00:26:54]: model gets smaller.Swyx [00:26:55]: I don't know why you say two times, ‘cause I think it's like 100 times. In language models, it is roughly one hundred to a thousand times every twelve to eighteen months, for the same given level of LMSys, ELO.Vibhu [00:27:08]: That's a net of everything, right? That's model performance alongside compute. So different than just compute costs come down. But, a very interesting future.Swyx [00:27:19]: So the web designers will have to shout out that accessibility is an issue, right? how do you deal with screen readers or whatever. But yes, this is higher bandwidth storytelling than anything you can possibly generate with code, right? So I think that's the rough idea.Ethan [00:27:34]: And I'd like to add a little bit that so human naturally have the maximum bandwidth when we are looking at things, look at videos, and we also have maximum output bandwidth when we are talking. So in the future, it might be something like we talk to AI models, and the AI model responds back with a generative UI. So that would be the maximum input and output bandwidth to interact with AI models before neural link happens.Vibhu [00:28:06]: And it's also very custom, right? Some people are very visual, some people are not as visual, right? They prefer the text. But the best thing about generative UI, right, it can also be text.Swyx [00:28:17]: There's another project that we wanted to highlight, which is the Neural OS. Kinda similar idea, but here you're literally operating, simulating an operating system with a video model.Swyx [00:28:27]: and you can play Doom, you can do Firefox. I find this like mildly less impressive, obviously, because it's an OS that I can run.Swyx [00:28:37]: But here everything is imagined.Vibhu [00:28:40]: I was, used to the Command+W to close the Firefox tab. It didn't crash. That's why I saidSwyx [00:28:45]: It's too immersive.Vibhu [00:28:46]: It's, it's too immersive for me.Swyx [00:28:47]: Too immersive.Vibhu [00:28:48]: I wanted to close the tab.Vibhu [00:28:49]: But yes, I can play generated diffusion.Swyx [00:28:51]: this is shockingly fast.Swyx [00:28:54]: Because I remember there was a demo about like maybe one to two years ago. Someone tried to do the first-person shooter with a image model. There was no consistency. It was very slow. But here it looks like realistically it's-- this is Doom.Vibhu [00:29:07]: I think there's two sides to that, right? There's okay, what is running a game? The heavy part of it is actually the game engine, all the lighting, all that stuff, the graphics. This is just kind of video, right? Like we've solved consistency. This is still, it looks like a few years old image generation. There's some temporal consistency, but it's, it's kind of just images stitched together as frame video. But it's a good visual representation to pi- to picture the future you wanna see, right? that's, that's what I see in these more so.Ethan [00:29:38]: This reminds me of how the video models gets better and better. So Neural OS is kinda if you just look at it feels like it's just a crappy version of the, like the Windows we could have, right? And, but the difference is, so the model, this model is overfitted on the existing operating systems. It can generate nothing different than that. But it's actually also similar to video models. So when we are training these video model, image model, we train them on internet. There's no imaginary supernatural stuff on the internet. But once we train this model, you can prompt the model to generate something supernatural that have never existed in the data set. So if you train your Neural OS or neural computer on the standard screen recordings on the entire internet. The model can imagine completely new interface to interact with the computer.Swyx [00:30:43]: This is one of those things that is magical to me. usually generalizing out of distribution is bad, but somehow we have learned some kind of internal world model that you say, this plus, but it looks like rainbows and butterflies, it'll do it and it will kind of make sense.Swyx [00:31:03]: So yeah, that's kind of cool. Yeah, I don't know if there's any comment more on there. I do, I do wanted to, I did wanted to touch a little bit more on the model architecture stuff, which I think you were getting. It's, really fascinating. We don't get a chance to talk about this enough. So one of the papers that we covered, we've covered every annual, segment anything release. and I don't know if you follow-- you're a computer vision guy, so youEthan [00:31:26]: I knowSwyx [00:31:27]: . So they did memory attention, which is kind of interesting. And I always think, anything where you can, across the temporal dimension, keep some consistency, I think it's, very fascinating, and I don't know if Basically, does that-- the CV side bleeding into video gen side, I think is underexplored, right? we talk about it for labeling, but actually you can borrow the architecture itself.Ethan [00:31:50]: There's, there's also complete different approaches, right? you brought up the term world model, so we went from video model to world model. There is diffusion, but there's also other approaches that people are doing. So maybe we get into those after as well,?Swyx [00:32:03]: He has a whole definition of world models and stuff. I feel like we threw a lot at you. Whatever you want to comment on.Why Video Models Are Expensive: Storage, I/O, and Training ScaleEthan [00:32:10]: I think one thing that we should actually comment back on is okay, so we were talking about the steps to train image gen to video model. One thing we don't see as much of is okay, you brought up the delta in training data, right? SoEthan [00:32:24]: you won't have as much a video model might not generalize, but what is the cost of training a large video model? So we know for LLMs roughly, okay, even like the poolside thing that came out today, right? It's a Gemma level model trained on roughly forty trillion tokens at this many H200s over this much time, right? You can see what is the exact cost of that. So how many GPU hours over how much H200 costs? So how do we do the back-end math of, same thing for video models, image models. How do you, how do you kind of break that down? I can share some back-envelope calculation. So surprisingly, video models is-- the cost is very-- is comparable to language models and obviously the largest scale is language model, maybe like a medium scale to language models. I said just storing the videos alone, it costs a lot. You can, you can maybe look up on AWS or something.Ethan [00:33:20]: You really, say if you have a billion videos and let's say, let's just say like each video, like five megabyte, then you need five petabyte to just store those videos. And also remember we talk about you use a VAE to compress the videos, and you also need to store, typically you need to store those continuous feature, in-- also in your storage. That's also comparable size with the videos themselves. So just storing these videos and the features is tens of petabytes alone. And,Swyx [00:33:58]: I just, I just looked up the calculation. Five petabytes on S3 Standard is one hundred K per month.Ethan [00:34:05]: AndSwyx [00:34:05]: It's comparableEthan [00:34:05]: and you needSwyx [00:34:06]: AndEthan [00:34:06]: And then like tens of petabytes, two hundred K. And even more expensive is you have the ingress and egress.Swyx [00:34:13]: Oh, yeah.Ethan [00:34:14]: Like you-- through the internet. You have to just to download those videos, I believe it's, it's more expensive on AWS than just storing those videos.Swyx [00:34:25]: Storing, yeah.Ethan [00:34:25]: And each training runs, you probably need to pull them once. If you train multiple times, it's, it's even more than that. So it's like just storing the network, those costs is just, it would be a few, a few millions per month to just storing everything, not to mention the GPU cost.Ethan [00:34:45]: AndSwyx [00:34:45]: my side tangent, the compute rental, like GPU rental is very efficient. There's one side, okay, you can be XAI and build your data center. Should we not just build our, storage compute as well? LikeEthan [00:34:57]: Of courseSwyx [00:34:57]: cloud cost compared to just,Ethan [00:34:59]: You save so muchSwyx [00:35:00]: store. Yeah, exactly.Swyx [00:35:01]: Especially with like egress and stuff. So.Ethan [00:35:04]: That's a good idea, but it also comes to-- there are some of its own challenges.Swyx [00:35:09]: Of course, of course.Ethan [00:35:10]: like people who build the GPU data centers, they might not expect this much, storage. And yeah, people build storage, typically they just build it somewhere with just CPUs.Swyx [00:35:23]: I just looked it up. Five-- AWS only charges for egress, not ingress. Tier five for five petabytes is two hundred and thirty K.Ethan [00:35:32]: Even more expensive than the storage.Swyx [00:35:34]: But storing is per month, right? You check in, then you cannot check out. so it's so cool. It's okay. So there's that side.Ethan [00:35:41]: So the TLDR, my backhand mathSwyx [00:35:42]: Data is larger than you think. Yes.Ethan [00:35:44]: my backhand math of GPU hours times GPU cost is also very much, I'm missing some storage.Swyx [00:35:49]: You're also-- you're basically like also more IO bound than normal training.Swyx [00:35:55]: Yes. ‘Cause like data loading, so caching everything, it becomes super important.Ethan [00:36:00]: So in Cosmos, we did a lot of optimizations to make it not IO bound. So, speaking of the training, actually training the model, the GPU cost, if you look up like the open source model, how big these video models are, I think like LTX has nineteen B parameters. That's a dense model. And people are also exploring, MoEs, so it might be twenty B active and, like a hun- hundreds B, total. So that's, that's even-- that's similar size as medium-sized LLM models. And if you, if you look at number of tokens-Uh, we disclose that in Cosmos. It's also like tens of trillions of tokens on the visual tokens. So putting this together, the cost of, training these video models, it's actually comparable with LLMs. Not to mention, the infra is slightly different from LLM, so it might be less efficient to train these models.Inference Speedups: Step Distillation, Consistency Models, and GANsSwyx [00:37:04]: Do you get the benefits of traditional diffusion speed-up? So for, images, there's LCM, LoRAs for, fine-tuning. There's, there's a lot of stuff that's beenEthan [00:37:15]: Flow matching.Swyx [00:37:16]: there's flow matching. There's a lot of stuff that's been done. there's some overlap that applies to diffusion on the inference side and stuff or?Ethan [00:37:23]: so the difference-- the inference side is a completely different story.Ethan [00:37:28]: I think for the training side, it might be a little bit hard to reduce that cost. And for the inference side, the biggest gain is from the distillation of these models. You can-- It's called step distillation, slightly different from knowledge distillation in LLMs. So you-- Typically, for flow matching models, you need like 100 steps or something. Like a distortion model even need even more, like 1,000 steps to generate a good image or video. A step distillation is try to learn to generate fewer step from the model itself. It's kind of like now we-- you use the full model to generate in 100 steps, and then you take a model that only generate 10 steps and let that model to learn from the perfect one.Ethan [00:38:25]: why this workSwyx [00:38:27]: Strong to weak seemingly.Ethan [00:38:28]: It is. It's kind ofSwyx [00:38:29]: DistillationEthan [00:38:29]: kind of like strong to weak. the-- from the modeling perspective, the strong model, the teacher model is trying to model the image and videos of inter-internet, and that distribution is extremely complex. But the step distilled model is just trying to learn from the teacher. The teacher is a model, and the size is fixed, as the distribution is much simpler than the whole internet. That's the intuition I have why step distillation can work. So usually these models serve in productions, they only run in a few steps. In Cosmos, I believe we have, we have like four step and eight steps. If you do some simpler task, image-image translation, it can even run in fewer step, like one step in Cosmos Transfer.Swyx [00:39:22]: I think this is the same intuition that guides a lot of the consistency model work. I sent you a link for, SCM. I don't know if you covered that. To me, that was actually one of, the most impressive papers I've ever seen from OpenAI.Swyx [00:39:34]: That this is the unifying grand concept of consistency models. I don't know if you have any comments on this.Ethan [00:39:41]: So there are, there are a few different approaches,Swyx [00:39:46]: Oh, yeah. Here it is.Swyx [00:39:47]: Two steps versus twenty or 100 steps, whatever. It's already done.Ethan [00:39:52]: So there are, there are a few different approaches, for example, consistency model, and there are also Actually, we shouldn't forget GAN. So GAN, actually, that was, that was the OG ofSwyx [00:40:05]: OGEthan [00:40:05]: step distillation ‘cause it trained just one step to begin with. So actually, a lot of, uh-- For example, there's a distribution matching distillation which use, which uses GAN, as one of the laws for distillation. It-- GAN just tells you, “Hey, generate an image,” and thenEthan [00:40:31]: it has a discriminator to tell, is this image real or not? So the model, the model just need to learn one of the distribution, not the full distribution. Because in training, the model is asked to reconstruct the ground truth image from the internet, which is extremely hard. And in-- When you're training GAN, it's a step process. It's just a, “Hey, you generate image. Does this image look as real as the image from the internet?” Which is a much simpler task. And, yeah, combining a lot of these approaches together, people typically do that, like consistency model and distribution matching and GAN, and we can get these few step models.Audio-Video Generation and Time AlignmentSwyx [00:41:21]: Then there's one step I wanted to add, which is audio and video.Ethan [00:41:26]: So, Grok Imagine zero point nine, I believe it's, it's a first audio video transmodel deployed at a large scale. SoSwyx [00:41:39]: And that was your first model?Ethan [00:41:40]: that was, Grok Imagine's first model. It's, it's audio video, joint generation. I think the hard part is, the modality alignment, ‘cause before this transmodel, we have, we have text to video alignment. We have this, correspondence between text and video. Typically, most of the VLMs, they understand images and videos. Video's very rare, and they don't understand audio mostly. And if you look at the audio generation on the LLM side, you can talk to them perfectly fine, but if you ask them to sing a song or something, it typically is not very good. Also, they don't have, they don't have music either. The hard part is thatUh, actually audio has two component. It has like a discrete component, a continuous component. The discrete component is like the language.Ethan [00:42:44]: So when we speak, it's just, someSwyx [00:42:47]: It's an ASR issue, yeah.Ethan [00:42:49]: It's, it's text token with some characteristics, I would say.Ethan [00:42:54]: But musicSwyx [00:42:56]: I think the speech guys would disagree with this.Swyx [00:42:57]: Like disfluencies and then,Vibhu [00:43:00]: There's tones you can get angry.Ethan [00:43:01]: Well, I say largely.Ethan [00:43:03]: the mu- but the music is completely different. It's, it's very continuous, and you cannot model them like discrete tokens in language models. this is like the hard part for models is, not to mention we have to align text, video, and audio together.Ethan [00:43:26]: SoVibhu [00:43:26]: How?Ethan [00:43:28]: So significant-- some significant challenges are like-- So first, like we talk about as the VLMs, they cannot understand most of them cannot understand audio.Ethan [00:43:39]: So you have to have some way to do the synthetic data generation for audio. You have to caption the model, and that involve, that involve synthetic data and human data effort a lot. And not just surprisingly, most of the LLMs are very bad at recognizing, like the beat, tone, and the details of the of music. They can, they can give some general prediction of which song is this, but it's very hard to describe the details of the music. like we mentioned in image generation, like you have to describe image as detailed as possible so that someone blind can reconstruct that. So here is like someoneVibhu [00:44:32]: DeafEthan [00:44:32]: someone deaf can reconstruct how the music sounds like without actually listening to it. Maybe you can think of it need to have the-- or they call the script.Vibhu [00:44:49]: Subtitles, yeah.Ethan [00:44:49]: You gotta have all the details of the music, and the dialogue.Vibhu [00:44:55]: So is the challenge there typically stuff like music and audio, or is it just Like is there a baseline? Okay, there's enough data where we can understand, narration, conversation, but there's nuances in audio that's where you hit all the data issues or is it just from stage zero, you just do it all right?Ethan [00:45:15]: So one important thing is like the alignment. So the model, the model has to know like the video and audio, the, uh-- it has to have a time-based alignment, like at which time step the video and the audio token correspond to each other. But we actually don't have this kind of alignment for most of the other modalities. If you think about like text and image, text and video, they are loosely aligned. So you can, you can have a description of what's going on in the video, but you don't have to exactly, You typically don't have exact description, oh, at, time step one second like what happened?Vibhu [00:46:02]: It's veryEthan [00:46:03]: At time step two second what happenedVibhu [00:46:03]: coarse. Yeah.Swyx [00:46:05]: So what was the ideal time step? You have to oblate it, and then it's like four seconds or something.Ethan [00:46:09]: So that comes down to how you design the model to, for the model to be aware of as a time, as a time modality. So the model is like a time aware. And that's something pretty unique if you think about LLMs. So if you ask LLM to complete a task, say they, uh-- you ask them and they will say, “Oh, this task will probably take twelve hours to complete,” and they come back in one hour. Say “I've already spent two days on this and I've exhausted everything.”Ethan [00:46:47]: So the LLMs them-themselves, they don't have a sense of time there.Vibhu [00:46:53]: I actually don't think that's just them not having a sense of time. I think it's somewhat based, right?Vibhu [00:46:58]: Like you tell someone, “Okay, go work on this feature. Go implement this,” there's a general understanding you would have of how long that would take without LLMs working at LLM speed, right? So you think back like two years ago, if I tell you to like build me like a new front end for latent space, have a search bar, have all this, you'll estimate that it'll take a few days, right?Vibhu [00:47:19]: So you tell an LLM, “Go build this.” It'll take me a few days. But I think it's somewhat grounded as opposed to them not having the best-- Not saying that they have a great understanding, but I think that example is like you can see where it comes from, right? You're trained on all over the text.Swyx [00:47:35]: They're, they're trying to estimate what a human would say.Vibhu [00:47:37]: because that's what the, that's what the data kind of represents. It's not themEthan [00:47:41]: It came from the corpus on the internet. People have a estimate of how much time.Vibhu [00:47:45]: And not even just in direct like training samples, right? Just your world understanding of tokens of how long stuff takes, right? Go read a book. It'll take you a while, right?Vibhu [00:47:56]: Even if you do nothing but read a book, it takes a few days. So yeah, LLM, I read it took me a few hours.Vibhu [00:48:01]: It'll take me a few hours to go through this research. But this is a tangent.Swyx [00:48:05]: Somewhat, yeah.Swyx [00:48:06]: This is a train of thought I haven't really expressed until now is, which is basically like a full world model must also be recursive, meaning that the participant in the world model must also be aware that they have a world model. which is like this whole recursive thing down the, down the line. but yes, and that the world model can be wrong and that they need to update it and blah. Yeah. We've, argued this on the, newsletter as well, that there needs to be sort of recursive or adversarial world models.World Models: Real-Time, Long-Horizon, Interactive VideoVibhu [00:48:34]: just, to ask, how do you define world model?Swyx [00:48:38]: Oh, yeah, let's go there.Ethan [00:48:40]: SoVibhu [00:48:40]: So just for context, we talked about, video generation, and then there's a-- if you say there's a distinction between world models, what's your, what's your definition? How do you see the two?Ethan [00:48:53]: So disclaimer, I'm not going to debate, what is world model. Yeah. there are many definitions, so I'll just talk about my definition. Since I came from the multi-model, multi-model domain, so mainly talking from video. So world model is like real-time interactive long horizon videos. So there are three parts. so we-- let's talk about them one by one. So the so interaction, so we just, we just look at Facebook and neural computer. So the interaction part of it, so you, world model can allow you to interact with them through keyboard, mouse, and maybe also voice. So these all is-- all is a modality. You can, you can interact with the model, and the model should respond reasonably. Second part is real time. So once you, once, say, you move your mouse, if, say, the world model generate a game, how fast can the game respond? So if you're like professional CS: GO players- -my say, oh, you have to respond- He's beginner within sub ten milliseconds or- Yeah even less. So that's not most of the- No, sixty FPS. Let's go. Oh, three hundred FPS. Oh, five hundred FPS. Wait. okay, yeah. I didn't do the math, but yeah, okay. Uh- Yeah, three hundred FPS, that's a three millisecond. So you have to respond- Oh, s**t. Okay. YeahEthan [00:50:29]: within a millisecond. Most of the video models cannot do that. Yeah. And, but if you, say, if you have a video model that is, say, like a digital human, the response time might be more generous. Maybe typically, for real-time voice interaction, it's like two hundred millisecond. So that's, that's much more generous. But even two hundred millisecond is pretty, it is pretty tricky, ‘cause remember we mentionedEthan [00:51:01]: you have this, temporal compression coming from the VAE. So if you, if you don't compress the temporal dimension, your sequence length is going to explode. So if you want to have this real-time, real-timeness in your model, you have to do is one context problem. And the third part is long horizon, ‘cause we-- if you're not going to just play with, video games just, a few seconds, most video models only a few seconds. We're going to play with minutes, hours. The model have to be able to generate long-form content.Ethan [00:51:42]: So putting these three together, it's, real-time, long horizon interactive videos. I think the final state will be, for example, like a video, a video version of Playbook, where you can, you can interact with, a neural computer. You move your mouse, and you click on the generative interface, and it will reply to you through pixels- generating in real time. But getting there, it's, it's a very long way to get there. So one of the first step, at Grok Imagine, where I led a small world model team there, was to build video extension. So, video extension- it's the first step of interactivity. Yeah. It's, it's the first step. Yeah. So it's the first step- You have it here, video editing, yeah. Yeah. Yeah. So the first step is because, this unlocks long horizon videos. Typically, for most of the video generation models, you give it a prompt or an image as an initial frame. You generate video, that's it. That's just, one time, done. And some creators would try to, use the last frame as a first frame for the second video. It can-- sometimes it works, but if you do it a few times, it says the quality would decrease. And- It doesn't have that context- Yeah over the full video, so the temporal- Yeah, exactly. Yeah, ‘cause you only gave it the last frame, of course, right? Yeah. Exactly. And- it's actually a pretty fun hack. if you've seen like- Oh, no, he's saying something better. Yeah. And for example, like Vue, I remember Vue 3 has like a second context of the last video. It is slightly better than using the last frame, but it has the same problem-- similar problem that it, the quality would decrease. if you extend a few times to, one minute, the video quality would look much worse than the first video. Second, another problem is that the model doesn't have long-range knowledge of, what's happening before. Say, if they generate some dialogue, some, two people speaking, and their voice might change, over some time, especially if the second conditioning, it does not cover the previous context. So these are the core challenges. So the Grok Imagine video extension, it has historical context of all of the previous generated videos. It can, It has, it has the context of, who is speaking and what objects have appeared and everything, having that to generate the next video. So if we naively do this, you can imagine, just, put all of the previous history video tokens into the context. The context lens will easily explode. Especially for video models, that can be like a few, a few million context, I would imagine- context lens. Yes.Yeah.Swyx [00:54:58]: Let's run with that.Ethan [00:54:59]: for example, like in Cosmos, I think just five seconds of video is like a fifty K or sixty K number of tokens. So like if you do, if you do fifty second, that's a five hundred K tokens. If you do longer than that, easily explode. This long horizon, problem was the first step we're trying to solve world model. It turns out people, yeah, people love video extension. Like a lot, a lot of the creators love using video extension to create longer form videos. This is the part I liked that you have a, you have an intermediate step toward the final goal instead of just a straight shot to the final version very much.Swyx [00:55:48]: But I can see you have a strong vision of where we want to end up.Long Context, Redundancy, and Efficient Interactive VideoVibhu [00:55:51]: Does it seem like it's an efficiency issue? okay, we're at a few million tokens context,. If you draw the parallel to language models, we had very short context, two thousand, eight thousand, then, you scale it up one million, ten million. sure, there's effective context, but at the end of the day, it's just what's it worth? sure, there's a whole training data side. In video, it might be slightly easier ‘cause we have a hundred million token video, right? Just take a movie with the full context there. Like is this efficiency from an inference standpoint that like it's expensive, but we know how to solve it? Or like why is this not the approach? So like my broader point was on your second point of world models, you say it needs to be interactive and live, right? You should be able to play a game and see the interaction live. So one thing I see with research is a lot of what you actually serve is different than what you build, right? So we talked about distillation. You train big model, you distill it, you do quantization, speculative decoding. We do all this stuff to serve it efficiently. Should we not just have a solution, like a world model that can interact well, do inference optimization, serve it, distill it secondary, so make it real time after you solve it? So like a-- another parallel is say, continual learning, right? What we need is someone to solve it and show it works inefficiently. Give it a few years, people will make it efficient. Same thing with regular attention, right? It worked. Over a few years, people have different forms of attention, and we've scaled it to be efficient at log context,? So kind of two things there, right? One is it seems like it works. You've scaled it. Can we not just scale it a lot more efficiently over time? Do we need a separate approach if this works? And same thing with interaction, right? if we can get it done, like if we can solve some way that it works, we can solve making it more efficient from an inference standpoint later.Ethan [00:57:53]: that's actually a very good point. So in videos, there's actually a lot of redundancies. So we solve a lot of the pixel redundancy from VE, but there's more redundancy in long range and long horizon videos. Say, if a character appear in the first clip and then it disappeared, it only reappear at the end of the video, you probably don't need the-- the context, like in the middle of the generation. So you only need that character, where you need. So that's why, I helped build another feature. It's a reference video.Vibhu [00:58:36]: Is it here?Swyx [00:58:36]: is it the same model release or different one?Ethan [00:58:39]: It's a different one.Ethan [00:58:41]: You probably need to search onSwyx [00:58:43]: I'll find itEthan [00:58:43]: X reference to video.Ethan [00:58:46]: So reference video allow you to like upload up to seven images as condition and generate the video. Say, if like I want-- it can, it can be characters or objects or even scenes. Say like I want, I want condition on, Sean's selfie and holding a bladeSwyx [00:59:07]: We have a dogEthan [00:59:08]: or whatever.Swyx [00:59:08]: We put the dog in the thing.Ethan [00:59:09]: you can put them there and the video models will generate the video from and copies the context over. So that can solve a lot of the problems there, like the long context problem. It doesn't need to have a very long context, but it's-- I feel like it's an intermediate solution. The modelSwyx [00:59:29]: It's cheating.Ethan [00:59:30]: the model should be able to like selectively know, where should I draw the references. So say if I want to generate a movie, I generate it autoregressive, like a ten second at a time or something. And now this character appear, I can look back to where it first appear and, bring that back. Yeah, this one, I put the references. Yeah, that's, Optimus, Einstein myself, Annie.Vibhu [01:00:02]: Oddly enough, I used Grok Search to find it, and it pulled your LinkedIn post. But yeah we found it.Ethan [01:00:08]: Interesting.Vibhu [01:00:10]: ButxAI's Underrated Work, Culture, and WatermarkingSwyx [01:00:11]: this is a problem. This is not your fault, but like XAI doesn't communicate all this work that you do very well because they just have the model release and then that's it. But actually, these details are very good.Swyx [01:00:22]: As far as I understand, everything you just described is state-art, like no one else has done it.Vibhu [01:00:30]: A lot of-- yeah, I have a lot moreSwyx [01:00:32]: And then, and then you just put this blog post with the cookies. I'm this is not enough,?Swyx [01:00:37]: but I, obviously this is like the high level numbers that people want to know. But no, okay, soVibhu [01:00:42]: And I wonder, like part of that is also some labs don't share research into what happens. And ifSwyx [01:00:50]: No, but this is literally bragging about how good they are, right?Swyx [01:00:54]: Like, why would you not say that you are capable of extending with full context? this is not a secret sauce. This is like we did the work. yeah, I don't know.Ethan [01:01:02]: different labs have slightly different communication styles.Swyx [01:01:07]: Anyway, if anyone from XAI is listening we are always happy to help you tell your story. Yeah, okay, so you did references, and I think, I think kind of the point you're, you're making is it is sort of like a kludge, right? this is-- you can do seven, but what about 100?Swyx [01:01:23]: Right? Then you need a completely different thing.Ethan [01:01:26]: So I think it's-- this is, a mechanism to, select the context from the history, and you might not put the entire history into the context. for example, there's a paper called Frame Pack, which haveEthan [01:01:41]: a heuristic that the latest history, the last one second, I put the entire history, and the history before that, I would, compress it and makes the video smaller. So they follow this pattern, this build overall pattern that the maximum sequence length is fixed. So the further you are from the current frame, you have a smaller image. So this is just a heuristic. I think it can be more automatic. The model is aware like which history part of it can be select. So this part of the research is actually being actively, worked on by a lot of people. It's also quite interesting. I feel this is actually, this part of long context is a little bit ahead of the LLM part.Ethan [01:02:31]: So for example, like in LLMs, if you-- so contexts keep growing. Let's say if you call tool and the tool call history is extremely long, that's still in context, and keep growing, keep growing. Even if you switch the topic to something else, the whole context was there. There are some agentic harnesses that help you to, say, prune the tool results and, prune Like when you, when you query a file, only show like the top 200 lines or something. Those were very heuristic-driven.Swyx [01:03:08]: For listeners, we did a write-up on the cloud code, leak where there are eight different kinds of pruning, including like you prune the tool results and all that. So you can, you can read up on that kind of thing.Ethan [01:03:17]: I think, one breakthrough in continual learning might be like a way to automatically, manage its own context.Swyx [01:03:27]: These are all heuristics, and they will be replaced by machine learning.Ethan [01:03:30]: InterestinglyVibhu [01:03:32]: TheEthan [01:03:32]: the same thing is being researched in both LLMs and video models.Vibhu [01:03:36]: The interesting thing is also like in the paper you showed, it's actually happening at the model level, right? Compared to like language models, sure, we have base attention, but we'll do our own compression, we'll do our own pruning, which is separate from model error.Vibhu [01:03:49]: Eventually, it all just boils in, hopefully.Swyx [01:03:52]: I think this is a form of like attention, but like also know sort of reasoning attention. I feel like that's different than normal attention.Swyx [01:04:03]: Does that, does that make sense?Ethan [01:04:04]: It's, it's different in the sense that attention, not to mention, set sparse attention aside,

Here is Debbie in her own words.With over 15 years experience in Frontend development I have worked as a Tech Lead and consultant for many important clients with various technologies and often with a strong focus on performance. I have lead teams both in house and remotely as well as giving workshops and training. I have many years of experience as a mentor for online learning platforms, Treehouse and OpenClassrooms and am a teacher at Vue School as well as Jamstack Explorers, and I am a writer for Ultimate Courses.I am a Platform Engineer – Applied AI at Zephyr Cloud, Google Developer Expert in web technologies, Nuxt Ambassador, and am a former Microsoft Most Valuable Professional in developer technologies, Media Developer Expert and GitHub Star Alumni.I have a special love for JavaScript frameworks especially Vue.js and Nuxt.js and am now focused on testing especially end to end testing with Playwright. I have a Frontend and FullStack Tech Degree and am Microsoft certified. I am an international speaker, and have spoken at many meet-ups and conferences worldwide on many continents including Antarctica.I am Irish but live in Mallorca, Spain and when I am not writing code and studying new technologies you can find me doing all sorts of sports from running, cycling and skiing, body combat and of course Taekwondo as I am a 4th degree black belt.You can find Debbie on the following sites:BlueskyBlogLinkedInGitHubYouTubeXPLEASE SUBSCRIBE TO THE PODCASTSpotifyApple PodcastsYouTube MusicAmazon MusicRSS FeedYou can check out more episodes of Coffee and Open Source on https://www.coffeeandopensource.comCoffee and Open Source is hosted by Isaac Levin

durée : 00:08:32 - Les journaux de France Culture - L'usage de l'eau, la question des pesticides, celle de l'élevage, le revenu des agriculteurs, autant d'enjeux qui risquent d'envenimer les débats lors de l'examen de la loi urgence pour la protection et la souveraineté agricoles. "Un texte de réconciliation", selon le gouvernement. - réalisation : Mathieu Laurent, Annie Brault, Martin Desclozeaux, Caroline Bennetot - invités : Pierre-Marie Aubert Agronome, sociologue et chercheur à l'IDDRI, Institut du développement durable et des relations internationales. Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:03:47 - Les Matins de France Culture - par : Lucile Commeaux - Cannes : à propos de rochers, d'autofiction, de Scarlett Johansson et de nostalgie Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:03:47 - Le Regard culturel - par : Lucile Commeaux - Cannes : à propos de rochers, d'autofiction, de Scarlett Johansson et de nostalgie

durée : 00:03:47 - Les émissions culturelles de France Culture - par : Lucile Commeaux - Cannes : à propos de rochers, d'autofiction, de Scarlett Johansson et de nostalgie Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:38:19 - Le 18/20 : un jour dans le monde - par : Fabienne Sintes - Douglas Kennedy regarde aujourd'hui son pays avec inquiétude. Entre retour de Trump, polarisation extrême ou durcissement du débat public, il décrit une Amérique traversée par un malaise profond. - réalisation : Philippe Lefébure, Nathalie Poitevin, Thomas Lenglain, Mathias Dubois - invités : Douglas Kennedy Ecrivain Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:38:19 - InterNational - par : Fabienne Sintes - Douglas Kennedy regarde aujourd'hui son pays avec inquiétude. Entre retour de Trump, polarisation extrême ou durcissement du débat public, il décrit une Amérique traversée par un malaise profond. - réalisation : Philippe Lefébure, Nathalie Poitevin, Thomas Lenglain, Mathias Dubois - invités : Douglas Kennedy Ecrivain Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France

durée : 00:37:40 - L'Invité(e) des Matins - par : Guillaume Erner, Yoann Duval - La dette explose, la croissance est révisée à la baisse, le ministre de l'Économie avoue naviguer "à vue". Ex-membre du gouvernement Bayrou, en poste à Bercy de décembre 2024 à octobre 2025, Éric Lombard décrypte une équation française devenue impossible. - réalisation : Félicie Faugère - invités : Éric Lombard Homme politique français

durée : 00:14:50 - Journal de 8 h - Un accord de paix avec l'Iran est très proche. Déclaration de Donald Trump alors que Téhéran menace de refermer le détroit d'Ormuz si le blocus américain des ports iraniens se poursuit.

durée : 00:03:37 - Le Pourquoi du comment : philo - par : Frédéric Worms - Leibniz soulignait que chaque être perçoit le monde à sa manière. La conciliation dépasse le simple arbitrage : elle est essentielle à la vie démocratique et au vivre-ensemble. Elle cherche à synthétiser les contraires plutôt qu'à les nier. - réalisation : Luc-Jean Reynaud