Podcasts about Franken

Godfrey is joined by Dante Nero, Akeem Woods, and Vishnu Vaka to talk about Target getting caught selling a Halloween "Hide and Eek" costume that is literally blackface with the same hat and grin as the old Al Jolson Jim Crow imagery, why we need to stay boycotting Target and use Blapp instead to support Black-owned businesses, the 70+ Black people hanged across the South and quietly ruled suicides, the special needs Black kid who got choked up by two white men in front of Walmart, Carmelo Anthony's case heading to appellate court after all the new evidence came out about the racist who came for him, RIP Dolly Parton being the wrong 80-year-old to go while Ted Nugent is still walking around, Bernadette Stanis from Good Times still looking fine at 72, Cambridge professor Jason Arday being pushed to his death by the same jealous white nerd who just got fired from Gettysburg, Officer Tatum being a full Franken-coon with a wife who looks like Demi Moore left on a radiator, a Black man named Thomas Miller designing both the 7-Up and Motorola logos, and Steve-O getting called out for going racist on the Mexican deportation talk. Learn more about your ad choices. Visit megaphone.fm/adchoices

Today on The Gist, guest host Scott introduces a 2022 conversation between Mike Pesca and former US Senator Al Franken, who discusses his return to stand-up comedy and looks back at his tenure in the Senate. Franken examines how his background as a comedy writer and performer informed his sharp, forensic approach to committee hearings, including his memorable question during Jeff Sessions' confirmation that ultimately led to the Attorney General's recusal, and his viral exchange with Education Secretary nominee Betsy DeVos over growth versus proficiency. He and Mike also discuss the cynical comedic worldview of Lindsey Graham, the breakdown of cross-aisle Senate comity, and Franken's proposed reforms to fix the modern filibuster. Stop online threats before they become real-world attacks. Visit ironwall.com/GIST and request a free Risk Assessment to see exactly how exposed your executives are. Produced by Corey Wara Do you have questions or comments, or just want to say hello? Email us at ⁠⁠⁠⁠thegist@mikepesca.com For full Pesca content and updates, check out our website at https://www.mikepesca.com/⁠ For ad-free content or to become a Pesca Plus subscriber, check out ⁠⁠⁠⁠https://subscribe.mikepesca.com/ Follow us on Social Media:⁠⁠⁠⁠ YouTube https://www.youtube.com/channel/UC4_bh0wHgk2YfpKf4rg40_g⁠⁠⁠⁠ Instagram https://www.instagram.com/pescagist/ X https://x.com/pescami TikTok https://www.tiktok.com/@pescagist To advertise on the show, contact ⁠⁠⁠⁠sales@amplitudemediapartners.com Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Mano and Oscar have worked the world, and now they're both BACK! Just in time to meet a new group of UK queens, and... keep talking about Down Under vs. The World. Head over to Patreon.com/DragHerPodcast for full video AND a weekly bonus episode. Mano's on Instagram ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠@manoagapion⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, Oscar's ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠@ozzymo⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, and Good Get's⁠⁠ ⁠@goodgetproductions⁠. We've got merch at ⁠goodget.xyz/store⁠. Drag Her! is hosted and executive produced by Mano Agapion and Oscar Montoya. Our executive producers for Good Get are Erica Getto and Myrriah Gossett. Drag Her! is a Good Get Production. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Die Standseilbahn Braunwald, die das autofreie Braunwald mit Linthal verbindet, hat Geldprobleme. Deshalb springt der Kanton nun ein und stellt der Bahn kurzfristig 600'000 Franken zur Verfügung. Damit kann sie weiterfahren. Bis Mitte nächsten Jahres könnte es noch mehr Geld geben. Weitere Themen: · Die Thurgauer Regierung verteidigt die Thurmed-Pläne. Diese möchte sich am finanziell angeschlagenen Spital Wetzikon beteiligen. · Die Thurgauer SVP-Nationalrätin Diana Gutjahr wird bei den nächsten Wahlen im Herbst 2027 nicht mehr antreten. Im Interview spricht sie über ihre Beweggründe. · Seit über 20 Jahren lottert die ehemalige Raststätte am Walensee vor sich hin. Jetzt gibt es Bewegung: Ein Liechtensteiner Investor hat das Zepter übernommen.

Hildebrandt, Kerstin www.deutschlandfunkkultur.de, Studio 9

Wie wurden wir eigentlich zu Deutschen?Dieser Frage gehen Krautzone-Autor Friedrich Versago und Chefredakteur Florian Müller in der neuen Folge unseres Geschichtspodcasts nach. Gar nicht so leicht zu beantworten! Von Merowingern und Franken bis zu den Karolingern – bei diesem Dickicht aus Stämmen, Gebieten, Daten und Schlachten kommt selbst eingefleischten Geschichtsfans schnell der Kopf ins Drehen.Um Licht ins Dunkel zu bringen, arbeiten sich Florian und Friedrich heute durch ein gewaltiges Stück deutscher Geschichte. Doch eines steht jetzt schon fest: Deutschland ist weit älter als 1871.Unterstützt unser Format mit einer Spende:https://www.paypal.com/donate/?hosted_button_id=CGC6WAGA6TKLJIhr wollt noch mehr KRAUTZONE?Für kurze Zeit gibt es für alle Neuabonnenten eine exklusive Aboprämie! Wähle im Warenkorb einfach zwischen Beutel, Quartett oder Tasse und erhalte die Prämie direkt zu deinem Abostart.

45 Jahre auf der Bühne, unzählige Auftritte und Kultfiguren, die längst zum fränkischen Sprachschatz gehören: Bernd Regenauer ist zu Gast im Podcast „Mit.Menschen“. Viele kennen ihn als Erfinder der „Metzgerei Boggnsagg“, als „Harald Nützel“ oder als Liedermacher mit scharfem Blick für die (fränkische) Seele. Hinter dem erfolgreichen Kabarettisten steckt eine ebenso spannende wie bewegende Lebensgeschichte. Im Gespräch mit Redakteurin Anette Röckl erzählt der Franke von seiner Kindheit in Wolkersdorf, seiner Schulzeit in Nürnberg und seinem Weg vom Offset-Drucker auf die großen Bühnen Frankens. Dabei wird es mal lustig, mal nachdenklich. Offen spricht der 70-Jährige über die Schattenseiten des Erfolgs, über eine Depression nach seinem ersten großen Durchbruch und darüber, wie er sich aus dieser Krise zurückgekämpft hat. Außerdem geht es um das Älterwerden, um Gelassenheit und um die Frage, was im Leben wirklich zählt. Eine persönliche, ehrliche und unterhaltsame Reise durch 45 Jahre Bühnenleben und ein Stück fränkische Zeitgeschichte.

Im Oktober 1929 kommt's an der Wall Street in New York zum grossen Crash. Bald folgt eine wirtschaftliche Krise, die grosse Depression, die weltweit spürbar ist – auch in der Schweiz, verspätet zwar aber härter als andere Länder. Anfangs kann die durch die starke Nachfrage nach Wohnungen gut ausgelastete Bauwirtschaft die Auswirkungen der Weltwirtschaftskrise auf die Schweiz abdämpfen. Zwei Jahre nach dem Crash in New York, stehen dann aber auch in Bern, die Zeichen auf Krise. Vor allem die Schweizer Exportindustrie aber auch der Finanzplatz leiden, die Arbeitslosigkeit steigt und Armut macht sich breit. Zwar ist in der Schweiz nie ein so grosser Teil der Bevölkerung ohne Stelle, wie in anderen Ländern, aber die Arbeitslosigkeit hält sich hartnäckig. Während Länder wie die USA die Nachfrage ankurbeln und die Krise vergleichsweise schnell überwinden, reagiert die politische Schweiz zuerst mit rigiden Sparmassnahmen und sie wertet den Franken lange nicht ab. Erst allmählich setzt auch hierzulande der Staat auf Arbeitsbeschaffungsmassnahmen. 1936 gibt die Nationalbank mit der Abwertung des Frankens der Exportindustrie Schub. In dieser Episode von SRF Geschichte treffen wir den Historiker Bernard Degen im Amt für Wirtschaft und Arbeit an der Utengasse in Basel, einem Gebäude, dass 1932 eröffnet wurde, um der wachsenden Zahl von Stellensuchenden Unterstützung zu bieten. Dort fragen wir, warum es die Schweiz später aber härter trifft als andere Länder, was die Weltwirtschaftskrise in der Schweiz auslöst und was bis heute davon sicht- und spürbar ist. ____________________ Hast du Feedback, Fragen oder Wünsche? Wir freuen uns auf deine Nachricht via geschichte@srf.ch – und wenn du deinen Freund:innen von uns erzählst. ____________________ In dieser Episode zu hören: - Bernard Degen, Historiker, Universität Basel. - Nicole Hostettler, Leiterin Amt für Wirtschaft und Arbeit Basel-Stadt ____________________ Literatur: - Halbeisen, Patrick, u. a., Herausgeber. Wirtschaftsgeschichte der Schweiz im 20. Jahrhundert. Schwabe Verlag Basel, 2017. - Pressler, Florian. Die erste Weltwirtschaftskrise : eine kleine Geschichte der großen Depression. 2. Auflage, Unveränderter Nachdruck, C.H. Beck, 2019. ____________________ Recherche, Produktion und Moderation: Klaus Ammann ____________________ Hier lernt ihr die Schweizer Geschichte so richtig kennen – mit all ihren Eigenarten, Erfolgen, Fails, Persönlichkeiten und Dramen. Im Podcast «Geschichte» (ehemals «Zeitblende») von SRF Wissen tauchen wir in die Schweizer Vergangenheit ein – und möchten verstehen, wie sie unsere Gegenwart prägt. Habt ihr Themenvorschläge oder Feedback? Meldet euch bei geschichte@srf.ch.

Marlen Koch ist leidenschaftliche Bäuerin und lebt gemeinsam mit ihrem Mann Stephan auf dem Bauernhof «Obermettlen» in der Nähe von Root (LU). Die Natur, ihre Tiere und das Musizieren erfüllen sie und prägen ihren Alltag.Doch plötzlich verändert sich ihr Leben grundlegend: Aufgrund einer Entzündung im Rückenmark wird Marlen zur Tetraplegikerin. Bewegungen und Tätigkeiten, die zuvor selbstverständlich waren, sind von einem Tag auf den anderen nicht mehr möglich.In dieser Folge von Querschnitt sprechen Marlen und ihr Mann Stephan über Angst, Resilienz und die Strategien, die ihnen helfen, ihr Leben auf dem Hof Stück für Stück zurückzugewinnen. Dabei geben sie bewegende Einblicke in ihren Alltag und zeigen, wie sie trotz eines schweren Schicksalsschlags nicht aufgeben und gemeinsam nach vorne schauen.Moderation: Elena HirtDie Gönnervereinigung der Schweizer Paraplegiker-Stiftung (SPS) zählt 2 Millionen Mitglieder, die mit ihrem Mitgliederbeitrag das Leistungsnetz der Schweizer Paraplegiker-Gruppe ermöglichen und querschnittgelähmte Menschen unterstützen. Jeden zweiten Tag führt ein Unfall zu einer Querschnittlähmung. Mitglieder, die nach einem Unfall querschnittgelähmt und lebenslang auf den Rollstuhl angewiesen sind, erhalten 250 000 Franken – rasch und unbürokratisch. Weitere Infos: paraplegie.ch

Kirschenmichel, Kirschenjockel, Kerscheplotzer oder Kirschenplotzer – die Namen für den traditionellen Ofenkuchen aus der südwestdeutschen Küche sind vielfältig. Kirschenplotzer serviert man vor allem in Südbayern, Südhessen, Franken, der Pfalz und in Baden-Württemberg – und dort im badischen Rheintal und im Schwarzwald. Anders als bei vielen modernen Rezepten im Internet stützt sich der Rezeptvorschlag von Thomas Vilgis auf die Tradition. Allerdings unternimmt auch er geringfügige Abwandlungen und bevorzugt eine süß-saure Variante mit Sauerteigbrot, Schattenmorellen und Kirschwasser. Eine leicht beschwipste Huldigung eines Schwaben an das badische Rheintal und die angrenzenden Schwarzwaldhöhen. Die es natürlich auch als alkoholfreie Variante gibt.

Verlorener Koffer oder Ambulanzjet aus Südostasien – nicht alle Reiserisiken sind gleich. Experte Harry Büsser erklärt, worauf man beim Versichern von Reisen und Freizeit wirklich achten muss – und warum die Kreditkarte trügerisch ist. Das Wichtigste zur Reise- und Freizeitversicherung: Vom Schlimmsten her denken: Nicht der verlorene Koffer, sondern der medizinische Notfall ist das eigentliche Risiko. Ein Spitalaufenthalt in den USA, ein Rücktransport in die Schweiz oder ein schwerer Unfall im Ausland können schnell zehntausende Franken kosten. Diese Risiken muss man nicht versichern: · Verlorenes Gepäck oder Täschchen · Flugverspätungen · Kaputte Sonnenbrille oder verlorenes Ladegerät Das ist ärgerlich, aber finanziell verkraftbar. Annullationskosten: nur bei teuren Reisen: Eine Annullationskostenversicherung lohnt sich erst, wenn der Reisepreis im Verhältnis zum eigenen Einkommen wirklich ins Gewicht fällt – etwa bei einer Kreuzfahrt. Für einen Flug braucht es das normalerweise nicht. Mietwagen: ruhig Blut am Schalter: · Haftpflichtdeckung mindestens 1 Million Franken – in den USA lieber mehr · CDW und LDW decken oft nicht alles: Häufig bleibt ein Selbstbehalt von 2000 Franken. Wer das selbst zahlen kann, muss es nicht extra versichern. · Wer keine Zeit für Diskussionen bei der Rückgabe hat: Zusatzversicherung aus Bequemlichkeit durchaus sinnvoll. Faustregel auf Reisen: · Vermeiden: Teure Uhr zuhause lassen, keine Wertsachen im Mietwagen · Vermindern: Mietwagen filmen vor Abfahrt, Passkopie separat aufbewahren, vorhandene Deckungen prüfen · Selbst tragen: Verspäteter Koffer, verlorenes Ladegerät · Versichern: Medizinische Notfälle, Rücktransport, Haftpflicht Kreditkarte ist kein Rundumsorglospaket! Die mit einer Kreditkarte verbundenen Versicherungen sind eher ein Sicherheitsnetz mit dicken Löchern. Schutz gilt oft nur bei vollständiger Bezahlung mit dieser Karte. Dazu kommen Deckungsobergrenzen und viele Ausschlüsse – unbedingt Bedingungen durchlesen. Merksatz: Versichern, was wirklich wehtut – nicht was bloss nervt.

Gegen das Vorhaben, die riesige Holzfigur oberhalb von Andermatt als «Max der Uristier» aufzustellen gibt es viel Widerstand. Die Gemeinde Andermatt bestätigt, dass 22 Einsprachen gegen das Baugesuch eingegangen sind. Eine private Petition fordert ausserdem mehr Transparenz. Weiter in der Sendung: · Der Luzerner Filmemacher Edwin Beeler wird von der Gemeinde Emmen mit dem Kulturpreis 2026 geehrt. Der Preis ist mit 3'000 Franken dotiert. · Was hilft am besten, um den Kopf durchzulüften? Um diese Frage dreht sich die heutige Folge der Regionaljournal-Sommerserie zum Thema «Time-out».

Es ist ein grosser Brocken: Der neue Bahnhofplatz in Olten soll total 160 Millionen Franken kosten. Im September kommt der Kredit ins Stadt- und ins Kantonsparlament. Die Diskussion hätte schon früher stattfinden sollen. Die zuständige Kommission des Kantonsrates hatte aber genauere Zahlen verlangt. Weiter in der Sendung: · Aarau: Ein neues WC für 280'000 Franken auf dem Bahnhofplatz soll einen Beitrag leisten zu Ordnung, Sauberkeit und Sicherheit. · Die Regierung des Kantons Solothurn will bei der SoH Ausbildungsplätze schaffen für Hausärztinnen und Hausärzte. Das kostet 500'000 Franken pro Jahr. · Bremgarten: Die Stadt fordert die Einwohnerinnen und Einwohner zum Wassersparen auf und liefert dazu missverständliche Zahlen.

Polling has Peggy Flanagan and Angie Craig at a near dead heat heading into Tuesday. And it's to the fill the seat of Tina Smith, who succeeded Al Franken, who succeeded Norm Coleman, who lost to Franken in a long recount in 2008. Following me? Norm Coleman's attorney in that process was Fritz Knaak....and he joined Vineeta for a great conversation on Thursday on The WCCO Morning News.

Der Umsatz von Swisscom sinkt um 3 Prozent auf 7.2 Mrd. Franken – trotz höherer Abopreise. Konzernchef Christoph Aeschlimann begründet die Preiserhöhung mit einem besseren Leistungsangebot. Dafür steigt der Gewinn durch Synergien aus der übernommenen Vodafone Italia. SMI: -0.2%

Polling has Peggy Flanagan and Angie Craig at a near dead heat heading into Tuesday. And it's to the fill the seat of Tina Smith, who succeeded Al Franken, who succeeded Norm Coleman, who lost to Franken in a long recount in 2008. Following me? Norm Coleman's attorney in that process was Fritz Knaak....and he joined Vineeta for a great conversation on Thursday on The WCCO Morning News.

Drehen Sie Ihre Wohnung auf den Kopf: Alles, was herausfällt, ist Hausrat. Comparis-Versicherungsexperte Harry Büsser erklärt, warum viele ihren Hausrat massiv unterschätzen und wo die grösste Gefahr lauert: die Unterversicherung. Das Wichtigste zur Hausratversicherung: Der Selbsttest: Könnten Sie Ihren gesamten Hausrat morgen neu kaufen, wenn er durch Brand oder Wasserschaden zerstört würde? Kleider, Schuhe, Möbel, Computer, Geschirr, Sportgeräte – die Beträge überraschen viele. Wer Nein sagt, braucht eine Hausratversicherung. Faustregel anwenden: · Vermeiden: Kerze nicht neben Vorhänge stellen, Velo am Bahnhof abschliessen · Vermindern: Rauchmelder montieren, Waschmaschine nicht unbeaufsichtigt laufen lassen · Selbst tragen: Kleinere Schäden je nach Einkommen selbst zahlen · Versichern: Was einen finanziell wirklich treffen würde Spezialversicherungen für Velo und Ski: Vieles ist bereits über die Hausratversicherung gedeckt – auch Diebstahl ausserhalb der Wohnung, meist bis 2000 – 5000 Franken. Nur bei sehr teuren Geräten lohnt sich ein Zusatz. Garantieverlängerungen: fast nie sinnvoll: Ob Mixer, Fernseher oder Laptop – in den meisten Fällen ist ein Defekt kein finanzieller Notfall. Ausnahme: wer im Schadensfall rasch auf Hilfe angewiesen ist und niemanden im Bekanntenkreis hat, der helfen kann. Achtung Unterversicherung: Das ist das häufigste Problem. Wer seinen z.B. Hausrat auf 30'000 Franken schätzt, besitzt in Wirklichkeit oft das Dreifache. Folge: Bei einem Schaden von 10'000 Franken zahlt die Versicherung nur 3000 Franken. Es lohnt sich, die Wohnung einmal realistisch durchzugehen und dies alle paar Jahre erneut zu tun. Merksatz: Die richtige Deckungssumme ist wichtiger als jede Handy- oder Garantieversicherung zusammen.

Die Fischervereine sind empört: Wegen einer technischen Panne im Laufwasserkraftwerk Rupperswil lagen Seitengewässer trocken und Fische starben. Nach Ansicht der Fischer müssen es viele sein. Die Betreiberin des Kraftwerks (SBB) sagt aber, man könne keine Zahlen nennen. Weiter in der Sendung: · Die Versteigerung des Autonummernschildes «SO 1» im Kanton Solothurn lief in der ersten Runde schief. Es gab Jux-Angebote, die den Preis auf über 1 Mio. Franken trieben. Die Motorfahrzeugkontrolle machte eine Anzeige gegen die Bieter. Doch die Staatsanwaltschaft hat keine Undersuchung eingeleitet. Sie sagt, Jux-Angebote seien in diesem Fall strafrechtlich nicht zu ahnden. · Circus Monti: Das Aargauer Unternehmen kann die Zahl seiner Besucherinnen und Besucher stabil halten und sogar leicht steigern. Es ist damit eine Ausnahmeerscheinung in der Zirkus-Landschaft der Schweiz. Diese schrumpft; wegen finanzieller Probleme haben diverse Zirkusse den Betrieb aufgegeben, Nock, Royal und gerade kürzlich auch Stey. · Sommerserie «Timeout»: Die Firma SIGA im Kanton Luzern stellt an jedem zehnten Arbeitstag die Maschinen ab. Die Belegschaft hat an diesem Tag Zeit, um Ideen umzusetzen, welche die Prodution effizienter machen können.

In der Aare bei Rupperswil sind Hunderte bis Tausende Fische verendet, sagt der nationale Fischereiverband. Grund war eine Störung im Laufwasserkraftwerk Rupperswil-Auenstein der SBB. Der Wasserpegel wurde stark abgesenkt und so der Fischaufstieg trockengelegt. Weiter in der Sendung: · Das Künstlerhaus Boswil schreibt erneut rote Zahlen. Der Verlust für das Jahr 2025 beläuft sich auf über 270'000 Franken.

Watch the full episode on YouTube:We first covered Baseten last year when DeepSeek mania was at peak hype. Now they have raised a monster $13B round and become one of the new cohort of AI Infra decacorns that are (with Nvidia, Intel, and the semis complex) chief beneficiaries of the Inference Inflection. We return to Baseten at the peak of the 2026 edition of Open Weights debate. Ali has published a viral breakdown of Kimi K3:And since you last saw him, Philip has spoken at AI Engineer and written the definitive book on Inference Engineering spotted all over SF:Three years ago, inference engineering barely existed as a category.Today, it is one of the most critical disciplines in AI. Inference engineering inherently tackles a different question than standard model training: “How do you turn those weights from training into a product that is fast, reliable, and affordable at scale?” Focusing on these creates an entirely new optimization problem.In one recent GLM-5.2 experiment, quantizing more of the model actually preserved its benchmark quality while increasing throughput by 20%, because the errors introduced in different layers could cancel each other out.Inference is no longer just the final step after training. It is becoming its own engineering discipline, with its own research problems, infrastructure, and increasingly specialized roles.In this episode, Baseten's Philip Kiely and Ali Taha join swyx and Vibhu to explain what actually happens after a new open model is released and what it takes to turn “we generated a token” into a fast, reliable, production-ready API.We go deep on cache-aware routing, disaggregated prefill and decode, quantization, speculative decoding, KV-cache movement, model parallelism, GPU kernels, and the race to make frontier models up to 10× faster. Philip and Ali explain why inference optimizations can still produce gains of 20%, 100%, or even 200%; how quantization errors can cancel one another out; why identical weights can behave differently across clusters; and how Baseten grafted a Kimi vision encoder onto GLM-5.2 without changing the underlying language model.The conversation then expands beyond LLMs into NVIDIA Dynamo, mega kernels, Rubin, AI-specific chips, local inference, video generation, diffusion versus autoregressive models, and the enormous compute barrier to generating coherent long-form video. Finally, we explore the convergence of training and inference, continual learning through persistent KV cache, and the emerging loop where models help optimize the infrastructure that runs them.We discuss:* What happens when a 200,000-token request enters an inference system* Cache-aware routing and reusing previously computed KV cache* Why prefill and decode are increasingly handled by different GPUs* When dedicated deployments become cheaper and more reliable than shared APIs* How speculative decoding uses a smaller model to accelerate a larger one* Tool calling, structured outputs, and what LLMs actually do* What it takes to support a new open model on day zero* Grafting Kimi's vision encoder onto GLM-5.2* Retrofitting inefficient model layers with components from other architectures* Why models sometimes collapse into repeating the same token* How hardware, kernels, and race conditions create nondeterministic failures* Preserving model fidelity while making inference faster* How quantization errors can cancel each other out* Why inference optimizations still deliver gains of 20%, 100%, and 200%* How optimized serving can make a model up to 10× faster* NVIDIA Dynamo, KV-aware routing, and distributed model serving* Speculative decoding the speculative decoder* Why local AI is about making models less dumb while data-center AI is about making them less slow* Tensor, expert, and pipeline parallelism across GPUs* Hardware-aware model design, auto-tuning, and the case against mega kernels* Rubin and why inference is becoming a systems problem* Whether modern GPUs are evolving into programmable AI ASICs* Why enormous models like Kimi K3 require GB300-class hardware* Why open-source video generation still trails Veo, Kling, and other closed models* The quadratic attention bottleneck behind long-form AI video* Autoregressive video, real-time generation, and compounding quality drift* Why future video systems may combine autoregressive and diffusion architectures* Training for inference and inference for training* Continuous post-training, deployment, evaluation, and improvement loops* How GLM-5.2 helped optimize the kernels serving GLM-5.2 itself* Why faster networking could unlock dramatically faster decoding* Continual learning, KV-cache compaction, and persistent model memoryShow Notes* How to build a day-0 API for Kimi K3* 22580: From GPT2 to Kimi3, ExplainedPhilip Kiely* LinkedIn: https://www.linkedin.com/in/philipkiely* X: https://x.com/philipkiely* Inference Engineering: https://www.baseten.co/inference-engineering/Ali Taha* LinkedIn: https://www.linkedin.com/in/aliestaha/* X: https://x.com/waterloointernTimestamps00:00:00 Introduction and the 200K-Token Prompt00:03:18 Dedicated Deployments, Speculative Decoding, and Tool Calling00:11:26 Launching Production-Ready Open Models00:19:06 Model Retrofits, Failure Modes, and Nondeterminism00:28:22 Quantization and Canceling Errors00:32:15 The Race to 10× Faster Inference00:40:48 Dynamo, Speculation, and Local vs. Data-Center AI00:50:18 Model Parallelism, Auto-Tuning, and Mega Kernels01:00:55 Rubin, GPUs vs. ASICs, and Custom AI Chips01:10:03 Giant Models and the Limits of GPU Memory01:12:42 AI Video, Quadratic Attention, and Autoregressive Generation01:21:47 Audio, Images, and Diffusion Models01:27:32 Training, Self-Optimizing Models, and Continual Learning01:40:06 Closing ThoughtsTranscriptIntroduction: Baseten, Waterloo Intern, and Inference EngineeringSwyx [00:00:00]: Okay, we're here in the studio with Philip, old friend from Inference Engineering, the book, as well as Baseten and everything that you've done, you and I have done before, as well as Ali. Welcome.Ali [00:00:15]: Pleasure to meet you.Swyx [00:00:15]: Waterloo intern.Ali [00:00:16]: Waterloo intern, always.Swyx [00:00:17]: When did you get “Waterloo intern” as a handle?Ali [00:00:19]: As a handle? Oh.Ali [00:00:20]: I think the rebranding happened mid-March. When I saw it was open, I was like, “I have to take it. Up for grabs.”Philip [00:00:26]: The problem is that Ali is really good at his job and is not gonna be an intern much longer.Philip [00:00:30]: So we have to figure out who's gonna get the handle.Ali [00:00:33]: Well, I'll pass the torch over to the next intern.Swyx [00:00:34]: Oh, okay. It can be, like, you just pass it to another Waterloo grad.Ali [00:00:37]: To another Waterloo intern. No, bruh.Philip [00:00:39]: Yeah.Ali [00:00:39]: Intern.Swyx [00:00:40]: Intern, yeah.Ali [00:00:40]: And no.Philip [00:00:41]: You gotta get an intern from Waterloo.Ali [00:00:42]: Yeah, I've gotta get an intern from Waterloo.Swyx [00:00:44]: Right.Ali [00:00:44]: But they have to follow the path.Swyx [00:00:45]: Oh, it could, but it could come from Baseten, so it's like whoever Baseten gets from Waterloo.Ali [00:00:48]: Right.Swyx [00:00:49]: Has the title of Waterloo.Ali [00:00:50]: It stays in the ecosystem.Philip [00:00:51]: Exactly.Ali [00:00:52]: Halfway through the internship, you either get it or you're out.Philip [00:00:55]: You should also do, like, a big graduation ceremony where you change the handle.Ali [00:00:59]: Just say it.Philip [00:00:59]: For everybody.Swyx [00:01:00]: You guys are good at ceremonies, clearly. We had a nice launch of the book, very successful. But before we get into all that, I wanna start off with a fun question for you. Okay, you're an expert inference engineer. What happens when I send a long query, say two hundred thousand tokens into Baseten's inference? What's the process of query through GPU model routing, balancing, all that? What is all the stuff that we don't think about?Long Context Requests, KV Cache, and Cache-Aware RoutingPhilip [00:01:26]: With a long query specifically, the first thing that I'm gonna ask is, “Have you sent me this query before, or at least part of it?” and I really hope you have, because it's gonna be a lot easier for me and a lot cheaper for you. So the first thing that we're gonna look at is some cache-aware routing, where we're going to see, we probably have a number of instances, a number of replicas up serving whatever model you're hitting. We want to send this one to something with, number one, available prefill workers, and number two, ideally some cached input already there so that we can skip prefill on at least part of these two hundred thousand tokens. If you're doing two hundred thousand tokens, it's probably coding or a multi-turn agent or something where you would expect to have that cached. If you don't, we're gonna have to send it to a prefill worker. We've at least on certain models disaggregated prefill and decode, so you're going to have one set of GPUs that's solely going to process the input, create the KV cache, and get you your first token, and then that's going to be passed over to a separate set of GPUs, which is going to run decode. We're going to iteratively make those tokens. We're probably going to have some speculator model in front of that. I'm going to assume that you're doing coding, and because of that, our speculator model, which assumes you're doing coding, is gonna have a high draft token acceptance rate. If I'm wrong and you're asking me to summarize every Harry Potter book, it's gonna be slower. And then we stream that output to you and account for it, charge you, a couple of pennies and say, “Hey, would you like to send another one?”Swyx [00:03:04]: Except Baseten doesn't charge by pennies.Philip [00:03:07]: Well, yeah, we charge. I'm assuming that we're talking about the public model APIs. If you are setting up a dedicated deployment, then yeah, it's not pennies.Public APIs vs. Dedicated DeploymentsSwyx [00:03:18]: Yeah, one of the key differentiators when I was talking with Baseten initially was that people who want very high volume just need to rent by the box, ‘cause then it's up to you to figure out how to saturate the box.Ali [00:03:31]: And more often than not, it's, like, way cheaper if you're pushing, like, millions of tokens per hour, if you just pay per hour instead of pay per token.Philip [00:03:37]: Yeah, they do. I think that we've increasingly seen a lot of demand for the pay per token APIs, just because everyone wants to try open models, and then once they find a use case that's really sticky, then they move over to dedicated.Swyx [00:03:51]: Is there a best practice on when it's time to swap over?Philip [00:03:54]: Couple reasons. Yeah, reliability, that's a big one, right?Ali [00:03:57]: Like, if they have a very specific use case, they want you to train something specifically for them, like they want their own spec dec, for instance, for their own traffic.Swyx [00:04:04]: Spec dec is speculative decoding.Speculative Decoding and Custom SpeculatorsAli [00:04:05]: Speculative decoding, yeah.Swyx [00:04:07]: You have to explain.Ali [00:04:07]: Sorry. Like, speculative decoding is like, if you have a huge model, right? And so the model is going to be generating one token at a time every single turn, every single forward pass. So we attach, like, this little, like, parasite, like this layer that goes on top of the model, and this model just has to predict. It does three very fast autoregressive forward passes, and it will predict, like, three certain tokens, and then you do one forward stage over the entire original model in order to see if those predictions were correct or not, and then you accept them or you reject them. Now, this draft model is traffic specific, so if you, like, Philip said, if you're summarizing Harry Potter books, I can train exclusively that draft model on Harry Potter books, and I can guarantee you that I'm gonna accept the three tokens every single time. And so with that case, I increase your decode speed. I wouldn't be able to provide this to you if you're a shared endpointSwyx [00:04:53]: YeahAli [00:04:53]: ‘cause I have no idea if you're doing Harry Potter, if you're doing coding, if you're doing English. We don't know. Also, there was a thing in the book that mentioned that if they really cared about a specific threshold, chapter four, I think. Do you remember that?Philip [00:05:06]: Yeah. The things that you can do is you can set a specific, like, batch sizing, a specific, like, parallelism strategy if you're trying to optimize for, like, throughput versus latency. You can. Maybe a NVFP4 quant doesn't pass your benchmarks and you wanna run a model at higher precision, you could do that. There's just a bunch of reasons why you might wanna have your own endpoint and the biggest one, of course, just being, like, you don't have to deal with someone else doing a hundred million tokens of benchmarking traffic at the endpoint when you happen to be trying to serve your users.Swyx [00:05:40]: Yeah. I think one thing that is. That is a classic journey. Like, it's people is asking the, what happens when you type Google into the browser. Tool calling, is that just, you're generating JSON or is there more complication beyond that?Tool Calling, JSON, and Structured OutputsAli [00:05:58]: Certain customers that we have, they have their own post-trained models, and so they demand a tool calling that's not just, like parse a file or go find the weather. It's something that's very specific and you have to do post-training on this. And if the post-training on the model is not good or if the quantization after the post-training to get the inference to be fast, the model will struggle reading the JSON file and reading the tool calling. But it doesn't require its own like sandbox. It's not like it's going to use that tool calling to like escape a sandbox or like it doesn't have to be contained. It can just be a normal dedicated deployment. The challenge with tool calling more and more seems to be that the companies want certain tool calling which is a very sensitive thing to train. And because you're dealing with all of the JSON outputs, if it doesn't like close the end of the request in a very certain manner, you end up with a model that did the tool calling and like the thinking and so as a result of that, it didn't see the result and just hallucinated the result as it decoded. That seems to be the most challenging thing with tool calling, not really the sandboxes model.Philip [00:06:56]: Yeah, that's a challenge on the training side and then on the inference side, there's work that you can do to scope the possible output. So we published this at this point close to two years ago, the solution to this problem which is you make a state machine and you use that to constrain the output to a specific format. So this is the structured output problem. If you remember backSwyx [00:07:27]: Yeah, the specific grammar is,Philip [00:07:29]: Yeah, exactlySwyx [00:07:30]: GML had this thing.Philip [00:07:31]: Yeah. So it's like the old-school “make sure this is only JSON”, return only JSON orSwyx [00:07:38]: YeahPhilip [00:07:38]: Grandma's gonna die type of prompts.Swyx [00:07:39]: Is it BNF grammar? At some point OpenAI had released a thing that was like, yeah, if you want to constrain your output, write BNF grammar, back as NOR.Philip [00:07:47]: In our inference system, it's just a specified output format. And you get the guarantee that your output's gonna be structured along that format. And so applying that to tool calls can like help cut down on. You can still call the wrong tool or call no tool. It doesn't solve the certainty problem but it at least solves the output structuring problemSwyx [00:08:10]: YeahPhilip [00:08:10]: Within tool calls.Swyx [00:08:12]: And MCP is just another form of tool, right.Philip [00:08:14]: Yeah, exactly.Swyx [00:08:15]: As far as there's no special thing there.Philip [00:08:16]: The thing I'm always like explaining to people is the LLM is not capable of doing anything. It's only capable of making suggestions of what to do and then if those suggestions are formatted in a certain way and applied to a system that knows what to do with them, then an action occurs.Swyx [00:08:32]: Yeah. Part of the fun stuff is, this is solved outside of tool calling too. Like in an agent loop if the output is not correct or you're right, like reasoning, tool calling was done in the reasoning trace, just be like, “Oh, I don't know what to do. Let me just try again.” And it might get there after a few tries. And on your point of training, sometimes this is harder in smaller models, so you don't have the same exact quality outputAli [00:08:56]: Right.Swyx [00:08:57]: When you just swap from a big model, right?Ali [00:08:59]: Yeah. I will say that, before, I think we need to go back to inference engineering proper.Ali [00:09:04]: But, I had expected that something would replace JSON because it's hard to stream JSON ‘cause JSON must be complete and you must have open and close brackets and everything. So it's hard to parse something or validate something while it's being streamed. So people invented all sorts of things that are like, I forget the name of some of these alternatives, but it's something like TOML, something like YAML. But JSON seems to be dominant still.Philip [00:09:30]: The JSON outputs aren't that long, right? Like you could have a long-- ‘cause tool calls also contain the arguments in them and perhaps for a certain tool you might pass like a very long argument. But my impression of the median tool call is that it's a relatively small number of tokens, right? So I would expect that speculators are generally fairly good at something as formatted as JSON. And so you would have like a pretty fast decode step there and that the streaming wouldn't be as valuable, but maybe I'm wrong about that.Ali [00:10:02]: I think you're also bounded by the software or that the model is gonna integrate with if the software is built with JSON for the tool calls or if the company that you'- if your customer says that this is how our software works and our tools are interfaced with JSON, you can ask them to like, change their software and say like, “Yeah, this is gonna be better for the model.” but like with the right training shouldn't be that much of a difference. Also more profitable if it outputs more tokens probably.Swyx [00:10:25]: Depends on your business model.Swyx [00:10:27]: It really depends. But I will say that, as a writer with like experience a lot with generated output, I do try to move from text to JSON text which is very long JSON, right? Like there's paragraphs in every field because I'm trying to structure it, right?Philip [00:10:44]: Right.Swyx [00:10:44]: I want you to first make factual statements, then make opinions then make bullet point summaries, have dates, have entity references have your sources for references, all these things. Anyway, so these are things that like I think people who really experiment with structural output have to really care about. But, let's, let's recurse up the stack a little bit. Before we started recording, you mentioned something really cool, which is that there's a lot of engineering that-- inference engineering that goes on when a new model provider releases a new model, right? So let's call it GLM-5.2, Kimi K3. I had previously assumed, especially if it's like, well, GLM 5 to 5.1 to GLM-5.2, like that you've supported them before. Is it that much work?What It Takes to Support a New Open ModelAli [00:11:26]: It's a lot of work.Swyx [00:11:28]: Yeah. Okay. So like, a lot of people, all you guys, right whenever a new model launch like, people rush to say like, “Oh, Hugging Face supports this, Fireworks supports this, Spacetime supports this,” and I'm like, “Yeah, of course we support it.” But what goes into that? What goes intoPhilip [00:11:40]: I think it's more than just support it too, right? It benefits the consumer a lot. Like I think it was with Kimi K2.5 or GLM-5.2 the latest, there was an inference war, right? X provider is at 90 tokens a second. The next day we're at 150. The nextSwyx [00:11:55]: I kinda kicked that off with the GLM-5.2.Swyx [00:11:58]: I wrote a Twitter article about. It got like half a million views,Ali [00:12:02]: Based on being numberSwyx [00:12:03]: YeahAli [00:12:04]: Or it's for something else.Swyx [00:12:05]: Yeah. Which,Ali [00:12:06]: Oh my GodSwyx [00:12:07]: Which then got everyone really excited about, hey, how can we, bend tracks a little bit further and,Philip [00:12:14]: There's a difference between support the model, as in I can make a token out of this model, and support a model, as in I have a production-ready API from this model.Philip [00:12:26]: Getting to the point of I can make a token out of this model is not that hard because generally the, open source inference engines, vLLM, SGLang of the world oftentimes even receive weights ahead of time, maintainers do, or the people making the model merge PRs to ensure support. So you generally can, just get it working on the standard open source stack without too much pain in most cases. The challenge is, every inference company is gonna have own proprietary stack. Some open source components, some in-house stuff. And for any arbitrary model, there's going to be some new stuff. Sometimes you get lucky, like K, two five to two six was, like, pretty similar.Quantization, Speculators, and Production ReadinessAli [00:13:16]: Yeah. It was pure continued post-trainingPhilip [00:13:18]: YeahAli [00:13:18]: If I remember correctly.Philip [00:13:19]: Even in those cases, there's still stuff you have to do. You have to redo the quantization work. You're taking the model from. Generally, these models are not released in NVFP4, and we want them to be in NVFP4 for maximum Blackwell compatibility. So we have to perform that quantization, and, calibrate the quantization to make sure that we're not causing any regression in the model's intelligence. And then we also have to train the speculator, as we've talked about. Generally, we have. We have ZDR, zero data retention on our model APIs, so we don't know exactly the traffic that people are sending us, but we know what's popular. We know that coding use cases are popular. We know that agents, agentic use cases are popular. So we can get public data sets that are representative of that traffic and train general speculators. Now, with speculators today, you need to train the speculator using the base model itself because you're getting hidden states out of the model from running inference on these specific prompts, and that is the training data you use to create the speculator. So there's that process which you need the real model weights for. And then there's of course just the process of, standing up all the infrastructure behind it, loading all this stuff, testing it. And then when there's a new model with a newer architecture, I think that, like, the DeepSeek models tend to be the most challenging as they have, like, the most novel architectural stuff going on, model after model. But every new model has something. Kimi K2 had. Oh, sorry, GLM-5.2 hadAli [00:14:53]: Sparse attention.Philip [00:14:54]: Yeah,Ali [00:14:54]: YeahPhilip [00:14:54]: the DSA.Ali [00:14:55]: Right. Which is brought from DeepSeek.Philip [00:14:57]: Yeah. AndAli [00:14:59]: So you can copy-paste then?Philip [00:15:01]: It kindAli [00:15:01]: I don't know how this works.Philip [00:15:02]: So, like we had to, like, build support for that into our runtime. And you're right, like it is really interesting the way that all of these open source labs borrow from each other. For example, like GLM-5.2 doesn't have vision. So something that, Haley, a guy on our team, if we could take a look at this, he, like, grafted the Kimi vision encoder onto GLM-5.2.Retrofitting Vision into GLM-5.2Ali [00:15:27]: We'll be training the projector.Philip [00:15:28]: Exactly. So if you think about, like, the encoder, there's the encoder, which is the part that looks at the image and turns it into latent information, and then there's the projector which likeAli [00:15:38]: You can say latent space. It's okay.Philip [00:15:41]: And then there's the projector that maps it onto, the model itself, and then there's the model weights. You don't wanna mess with the model weights because you run a chance of making the model dumber at something else for the purpose of giving it vision. So instead, Haley started with just a projector, which is only a handful of millions of parameters.Ali [00:16:02]: That would be, yeah.Philip [00:16:02]: Yeah.Ali [00:16:03]: Can you show the training one?Ali [00:16:04]: Like the way it groksPhilip [00:16:05]: YeahAli [00:16:06]: Very interesting.Philip [00:16:06]: And maybeAli [00:16:07]: That right therePhilip [00:16:07]: Maybe Ali, you should take it from here. You've got a betterAli [00:16:10]: Ooh, double the sandPhilip [00:16:11]: Understanding of this than I do.Ali [00:16:11]: Yeah. You can see, like, he. The way he trained this is really cool. At the beginning, he was training it using just like, “Here's a picture of a mountain. Can you describe what's in this mountain?” And that caused it just like the first, learning walls. Like here you can see this all we're trying to teach it is to translate the encoded. Like it's already taken the encoder from Kimi K. It's taken the image. It'Philip [00:16:31]: Yeah. FrozenAli [00:16:31]: FrozenPhilip [00:16:32]: With adapter.Ali [00:16:32]: Exactly.Philip [00:16:33]: Yeah.Ali [00:16:33]: So the brain is frozen and the eyes are frozen. It's just we're tryingPhilip [00:16:37]: AlignAli [00:16:38]: Interconnect between the eye and the brain, right? So the projector. And so you take the tokens and then he's like, “Oh, can you describe what's in this image?” And he's like, “Oh, it's a mountain,” or it's a person or it's a human, whatever the case is. But that didn't cause complete understanding. So he changed it such that every image was associated with a data set of questions. Like, does this image have a white male? Does this image have birds in the top corner? Does this image have a scientist in it? All of that stuff. And it would have to answer questions correctly. And using not just training on describing an image, but being able to answer question, another question, answer over time. Like you can see the grokking, which is like genuinely insane, that retrofitting vision into a large LLM can learn to that extent. And even for images that it doesn't perform well on, for instance, if you ask it a picture of like Stephen Hawking, “Who is this?” Maybe it doesn't get it, but it will say something like, “This is Albert Einstein.” Like it still understandsPhilip [00:17:25]: Close enoughAli [00:17:26]: That this is a scientist who is a man who has, some significant achievements, all that stuff. So that's like really cool.Philip [00:17:32]: Yeah. So, we've covered Hao Tian before, who the author of the LLaVA paper that did this, a while ago. And I think that's very foundational work for anyone who hasn't done vision work before.Ali [00:17:41]: Same with the CLIP and MetaCLIP, where you go from just captioning to building out questionsPhilip [00:17:47]: RightAli [00:17:47]: Off the image and how much better you can get performance.Philip [00:17:50]: Right. Right. Right. Yeah. But what's, what's so exciting about this is if you look at a model like this. Now, this is a little bit more of a research project. It's not. It got to 56% on MMLU Pro, I think. So not quite frontier. But if you're running this model, you haven't suffered any loss on your GLM-5.2 quality. If you don't have an image, it'll just behave exactly the way it used to. And ultimatelyAli [00:18:14]: Which in the inference code you literally do not include the other part, right?Philip [00:18:18]: Yeah. You would just skip the encoder if you don't have an image input.Ali [00:18:22]: Okay.Philip [00:18:22]: Just confirming.Philip [00:18:23]: YeahAli [00:18:23]: Does it affect a lot on the overall inference side? Like you're not adding much, you're adding a very small vision encoder. These are typically likePhilip [00:18:30]: They're super fineAli [00:18:31]: Less than a billion parameters, right?Philip [00:18:32]: Yeah. It's, - There's a little bit less standardization among vision encodersSwyx [00:18:37]: YeahPhilip [00:18:37]: So the support matrix can be a little bit, sparser. But overall, yeah, it's a pretty, it's a pretty minor component of the overall system. And ultimately what you get out of the system is all of a sudden you have Kimi Vision, GLM weights, and DeepSeek attention all in one model.Open Source Model Grafting and Franken-MergesPhilip [00:18:56]: And that's, I think, a lot of the power and beauty of open source, is that you can take all of these different components and combine them together into a system that's better than anyoneSwyx [00:19:05]: YeahPhilip [00:19:05]: Can be individually.Swyx [00:19:06]: People used to say that you would also do Franken-merges where you would take likePhilip [00:19:10]: YeahSwyx [00:19:10]: Layers from each model.Swyx [00:19:11]: Does anyone do that anymore?Ali [00:19:13]: Well, to your point previously when you were mentioning like, the work that goes into supporting a model when it first comes out, like GLM-5.2 or MiniMax M3 or whatever the case is. Sometimes you do have to like, you do have to switch out some things. Like, for instance, the MiniMax M3 head uses full attention, and with full attention you end up with this like insane bottleneck in spec dec ‘cause you're doing auto-regressive token generation for three tokens, and you're doing this like N squared over all of the tokens that are in your sequence. Your KV cache is like very large because it's not sparse, it's not top K. So we find it better to like, okay, we're gonna replace this, we're gonna replace this layer with a layer from another model that's using like GQA, for instance. And then just with the right training, you can get it to have the same acceptance rate. So it is very possible to retrofit layers from other models and very much needed. If a layer is like inefficient, the training just becomes the challenge, like how do you ensure that you train it properly? Which again to your earlier point is like the mesh between training and inference. As in like you need very good training in order to do fast inference. That's like, I feel like more and more becoming true.Swyx [00:20:21]: Yeah. Anything else on the support side when you say like get it to fully production ready?Loop Detection, Race Conditions, and Non-DeterminismPhilip [00:20:26]: Yeah. I think that there's also a question of just, we can test a model to a pretty extensive degree, but we're trying to get it out quickly and then you see a bunch of other people test it and you get interesting results. There was an issue with, GLM briefly where we had some like mode collapses where it would just output the same token over and over again for certain prompts on certain temperatures. Like once you expose an endpoint to the real world, there's going to be, so many more varieties of things given to it that you're able to, discover and patch things. So it's not just a, day zero process, it's then like for the first week, for the first month, if a model remains popular, like how do you both fix bugs and then continue to push the envelope on performance?Ali [00:21:21]: What do you mean you don't want your model outputting S?Swyx [00:21:24]: Is there loop detection on that stuff, by the way? It still happens like quite a lot, which is surprising.Ali [00:21:30]: We have like we, in our endpoint, like if a model was to output the same token like four plus times, we just cut the generation. We say like, “Oh, sorry, this-- Like try again,” or like we will reprocess the request. ‘Cause we know then, like if it, like if, yeah, it's four times the same token, it's probably collapsed.Swyx [00:21:45]: Yeah. Is there a way to opt out in case I really want that?Ali [00:21:48]: You want that?Ali [00:21:50]: I think there's a way that we have to handle it. I'm not exactly certain, but I feel like in certain models, like when they output something like you can imagine, like a table for instance, and so they want, they wanna draw like 12 dashes and 12 dashes. Yeah, I think there's a way for that to happen. I think we only do it on certain tokens. Like we exclude certain special characters.Swyx [00:22:07]: Yeah.Ali [00:22:07]: So we only do it on like certain like S is the most common almost. GLM-5.2Swyx [00:22:11]: OhAli [00:22:11]: And I think it was DSV 4 as well. Like you'd just have like looping issues where like you literallySwyx [00:22:17]: ItAli [00:22:17]: Just have like S.Swyx [00:22:18]: Yeah. Is there a special, something special about S? No, just randomlyAli [00:22:21]: It just seems to be the one token involved.Swyx [00:22:23]: Yeah. And it'Philip [00:22:24]: Is thereSwyx [00:22:24]: And it's only temperature 0Ali [00:22:27]: NoSwyx [00:22:27]: Even at other temperaturesAli [00:22:27]: Even at like 0.9 or whatever, it will still, it will still collapse.Swyx [00:22:30]: That's weird, right?Ali [00:22:30]: It's, it is an inference problem to be honest, like a software problem. Like oftentimes, the image you run will-- like NVIDIA will release an image for instance, and if we will upstream the changes from their latest TensorRT-LLM image into our stack, we'll find that it fixes it. Or oftentimes this will only happen in an inference engine that you're using like SGLang. But if you were to switch to vLLM, that isn't the case. So it seems to be like an extremely like deterministic software issue and not really a model issue. It's not like a weights problem. Like I'- we'll say like, “Oh, it's a problem with the quant. We did PTQ wrong,” right? But that isn't, that doesn't make sense because the same weights used with a different inference engine does not repeat the problem. And sometimes it's, the kernels that are being used in the backend have like these very subtle sometimes race conditions, where if you were to use this model hosted on one cluster, you will never get this problem.Swyx [00:23:19]: Oh my God.Ali [00:23:19]: But if you host it on a different cluster, you will. And the reason is the KV cache transfer from a node to node in that one cluster is using a slower interconnect than the node to node in another cluster. So that exposes the race, whereas in another cluster it doesn't. So then you end up just like, okay, this model is not gonna be hosted on this cluster. We're gonna host it on, another cluster because that cluster exposed that problem. But then it ends up with like, okay, is it the software? Is it the model weights or is it the hardware?Swyx [00:23:42]: There is a thing about this with temperature 0 still not being deterministic, right?Ali [00:23:46]: Right.Swyx [00:23:46]: Mostly because of hardware. Even at temperature 0 same model, you won't always get the same output.Swyx [00:23:52]: Even-- But I'm surprised by the race condition one because, I thought PyTorch was a graph that like guarantees that you at least, execute things in the right order.Ali [00:24:02]: Well, yeah, true. Like I'm not, I'm not saying that there is. Like well, you have things like PTL optimizations where like you can start a kernel before the end of the previous kernel, and that's like ‘cause you want to do that because there'sSwyx [00:24:12]: It's like pipeliningAli [00:24:12]: Expense. Exactly.Swyx [00:24:13]: Yeah.Ali [00:24:13]: But it'- But you don't do it cleanly. Like you overlap a little bit of the execution. No, it is very possible that the kernel itself, like that one block that is supposed to be running in this instance of time, that kernel itself has a race condition. For instance, like a missing barrier. Like often if you're designing a kernel and you want it to make it to be very fast, if you don't test it extensively, you'll, you'll have certain threads access data points from registers before they've been written to by other threadsSwyx [00:24:36]: YeahAli [00:24:36]: For example, because like your barrier is wrong or your synchronization was wrong. But yeah, like the testing itself is very difficult in those like, andSwyx [00:24:42]: And there's no like borrow checkerAli [00:24:45]: What does that mean?Swyx [00:24:46]: Like Rust. Like the. If you're trying to have like memory safety It sounds like a comparable problem.Ali [00:24:52]: Well, yes, but you're working in CUDA, right, NVIDIA GPUs. Like- You just need a higher level language like modular Maybe that's what modular is supposed to do. I don't know.Quantization Quality and Vendor FidelityVibhu [00:25:00]: How do you see keeping quality of the model? So you talked about all these steps of, okay, you gotta do quantization, train your own speculative decoderAli [00:25:07]: RightVibhu [00:25:07]: Run on different hardware. Looking at other model providers, okay, you kicked off a inference speed race on the consumer end. What goes into keeping quality the same across them, right? Sure, you can run benchmarksAli [00:25:22]: YeahVibhu [00:25:22]: But, like, how do you determine how much quantization are there standards? What goes intoPhilip [00:25:27]: There's a few things on quality. Most inference optimizations are lossless. KV caching, for example. You are just recomputing or preventing recomputing the same values. Speculation, of course, if a draft token is wrong, it gets rejected. The main lossy optimization is quantization. And that really comes down to, number one, data format, number two, which parts of the model you choose to quantize, which layers, and number three, like doing a lot of calibration on the quantized weights, to ensure that you're preserving all the outliers. There's other tricks that you can do, though. A big one is long context, ‘cause one thing you asked at, right at the beginning is, “Oh, what's gonna happen if I send a 200,000 token request in?” So with a long input sequence, you need to, store a lot more information. You need to process a lot more tokens. And so even if a model has a context of a certain length, you might, as an inference provider, choose to build an API with a shorter context length, and of course a full length one as well. Because if someone doesn't need the full million token context, for example, you can get them better performance. I don't know if that's exactly like quality of the model. The way that I think about quality is to what degree are we faithfully serving the original model? If you think of a golden implementation of a model that performs exactly the way the model is designed to perform, I think of quality as how close are we getting to that, 100% fidelity of the model.Philip [00:27:13]: You can also, of course, think about quality from the training side and how do you push yourself past 100%. But when I think about purely inference optimizations, it's getting faster while staying as close to that 100% fidelity mark as possible. And certainly our standard internally is that, like you should not be able to tell the difference between our API and a, official API. I think Kimi in particular does a good job of vendor benchmarking hereAli [00:27:41]: YesPhilip [00:27:41]: Where they haveAli [00:27:42]: They released an actual vendor benchmark.Philip [00:27:43]: Exactly, yeah.Ali [00:27:44]: ‘Cause they accused, some people, Amazon? There was some provider that was not doing very well on Kimi's benchmark.Philip [00:27:50]: Yeah.Philip [00:27:51]: So, with Reflect we probablyVibhu [00:27:52]: This was a long time ago, right?Philip [00:27:54]: No.Ali [00:27:54]: Yeah, like threeVibhu [00:27:55]: They alsoAli [00:27:55]: Four, five months agoVibhu [00:27:57]: This also happened with, I don't remember which model, but they pulled out quite a few, and then they started a whole chart about this. It might have beenPhilip [00:28:03]: Kimi Vendor Verifier.Ali [00:28:04]: Yeah.Philip [00:28:05]: Yeah.Ali [00:28:05]: Yeah, ‘cause you, ‘cause you'd be pissed, right? Like if you'Philip [00:28:07]: Yeah.Ali [00:28:07]: If like if I'm a consumer and I'm using like Amazon's endpoint for instance, and I've used Kimi and I'm like, “Oh my God, like this is bad,” I'm not gonna say, “Oh, Amazon quantized the model in a bad way.” I'm gonna say, “Oh, Kimi sucks.” Right?Philip [00:28:17]: Yeah.Ali [00:28:17]: So it seems like that makes sense.Philip [00:28:19]: Yeah, they care. They care.Vibhu [00:28:21]: Justifiably.Ali [00:28:21]: Yeah, justifiably.Vibhu [00:28:22]: This is probably a stupid question, but just checking, has anything improved from main quantization?Philip [00:28:28]: Yeah.Vibhu [00:28:28]: Like, is quantization always strictly worse?Ali [00:28:30]: Well technicallyVibhu [00:28:32]: NoAli [00:28:32]: It's a lossy. QuantizationPhilip [00:28:33]: YeahAli [00:28:33]: Is a lossy, it's a lossy implementation.Philip [00:28:36]: Speed improvesVibhu [00:28:36]: Speed improves.Ali [00:28:37]: It the number, likeVibhu [00:28:38]: No, I' always look for inverse scaling laws.Philip [00:28:40]: Yeah.Ali [00:28:40]: Yeah.Vibhu [00:28:40]: This is something I learned from Noam Brown, where like things that normally act in one direction sometimes do.Philip [00:28:45]: Well, technically when you run a benchmark, because these models are deterministic, sometimes your,Ali [00:28:52]: YeahPhilip [00:28:52]: NVFP4 quant is like, two basis points higher than yourAli [00:28:56]: No, it's noise. It's noise.Philip [00:28:57]: Yeah, exactly. I'm like, yeah, it's, it's within. That's why I always say within margin of error.Philip [00:29:01]: And I stopped saying that because everyone assumes that what is, well, within some margin of error, we're barely inside of that to the worst, so we're saying. But yeah, sometimes it's just like, gives you a higher output score. But like Ali said, that's noise. To my knowledge, you're not necessarily making the results better. You're just trying to, again, like keep your fidelity as close to 100% to the original model.Layer Selection, KL Divergence, and Better QuantizationAli [00:29:27]: There is, to your point, research that we did on MP. I don't know if you are able to pullPhilip [00:29:31]: YeahAli [00:29:32]: A tweet we did. One of our research interns, Joshua, I think it's a tweet on how we have 20% better quantized GLM-5.2 than NVIDIA. Essentially what we found throughout like this month research is, okay, quantization is a lossy. It's. You're compressing the data from, occupying 16 bits to occupying, four bits, for instance. And so you're losing some information, and you're trying to minimize that. And so when I say that I'm gonna quantize the model, my job becomes how do I find the layers that I can quantize, and how to find the layers to not. For instance, with image models, I don't quantize modulation layers, and I don't quantize out projections because those two are. Like out projection is what you see as the user. Modulation is what the model sees or understands. Right, exactly. And so to his paper, do you have the. It doesn't have the. Yeah. It's a long paper. I don't know if I can findVibhu [00:30:25]: If there's a part to search or it's probably in the thread.Ali [00:30:28]: It's probably in the thread.Vibhu [00:30:29]: Yeah.Ali [00:30:29]: But the long and the short is it is very possible that quantizing more of the model makes the results. Like if I have a model that I quantize layers one, five, and 10, and another model where I only quantize layers one and It is possible that the model in which I quantized more information is going to perform better because the quantization errors have canceled out. And so what Joshua showed in his mathematical proof where he had like a verifier in, is that you can predict which layers are going to have quantization errors that will cancel out with each other, and you choose to quantize those layers. And so the result of doing this mathematical quantization is you end up with a model that's 20% more quantized than another provider, so you get 20% more throughput of it because there's more layers than running an NVFP4, and your quality is better than that other quant because the layers that you chose to quantize have their errors cancel out, like one layer skewed to the right one layer skewed to the left, one layer skewed to the right. Your final logits distribution is more similar to the original distribution of the model, so you have better fidelity. And so the way we proved this was with KL divergence. So instead of just scoring on the benchmarks, we scored the KL divergence between the logit distribution of the quantized model and the logit distribution of the original full precision model, and we showed that with this technique we get. If your probability distribution on the logits which token it wants to select is more of the same as the original model, you're probably gonna end up staying true to the original model. So yeah, so it seems like previously before this, it seemed like the industry was, well, the more you quantize, the worse it's gonna be, ‘cause the more loss you introduce. That's not exactly, not necessarily true. So yeah, doesn't improve it, but can cancel out.Philip [00:31:57]: I think it might be this, but reminds me a good bit about pruning where you can prune off certain layers.Philip [00:32:03]: But very interesting. Didn't know this was a whole paper you guys put out.Ali [00:32:06]: It's. Fun fact, it was originally 72 pages, this paper, and then we decidedPhilip [00:32:11]: WowAli [00:32:11]: We can't tell. We couldn't release it. So it's now 45.Swyx [00:32:15]: Still 39 pages, so very substantive. We talked about evals and all these things and, like what's possible in terms of speedup? Like it's like probably like the numberInference Speedups and BenchmarkingSwyx [00:32:25]: Thing that people do wanna care about, and it's something that you wrote about in your post. Like official API is 70 tokens per second, and you push it up to 90. Is that like a normal thing?Philip [00:32:36]: So what's cool about working in inference, the reason that I think inference is going to be a useful place to do engineering for a long time, is that if you look at highly optimized domains like, say, finance, if you're in finance, you measure how much better you got in basis points. It's like, “Oh, I got five basis points better, like twentieth of 1% better,” that's huge news because everything is so optimized. When we publish optimizations, it's 20%, it's 100% it's 200%. So there's still probably like a lot further to go, honestly. Like you'll, you'll know that inference is pretty much solved when researchers start publishing about how they got 1% faster at something.Swyx [00:33:19]: Which by the way, because I am from the finance background, in the ‘70s, that was the margin at the time. When you did quantitative finance research, you would findAli [00:33:27]: And like 20%, tens of percent.Swyx [00:33:29]: That's. Yes.Philip [00:33:29]: Yeah.Swyx [00:33:30]: And now it'Philip [00:33:31]: Tiny fractionsSwyx [00:33:32]: For those people interested, look up Andrew Lo's paper. He had a really interesting illustration of quant, stat arb, distribution, narrowing down from like those kinds of 20% differences in the ‘70s, down to nothing today, which is very cool.Philip [00:33:48]: Exactly, and we're at the beginning of the same type of thing. Now benchmarking is hard. I think anyone will tell you that, and benchmarking provider speeds is hard because there's so many variables that go into it. What hardware are you using? How much load do you have on the system? What's the exact nature of the prompts and input and output sequence lengths? All that stuff. But overall, when you start stacking these improvements, you're looking at multiples. You can look at it. The most common form, of course, is TPS, tokens per second, which is bad naming by us in the industry, ‘cause there's two tokens per second. There's tokens per second, the throughput number, and the latency number.Ali [00:34:31]: TTMT, yeah.Philip [00:34:32]: Like total tokens per second out of the, out of the GPU as a throughput number. Most people only care about tokens per second as the latency number, which we should call ITL, intertoken latency, but we don't.Philip [00:34:44]: Anyway, so you can imagine a standard API without many optimizations for a 1 trillion parameter model operating somewhere in the 30 to 50 tokens per second range for reasonable traffic profile. And we generally see the goal of, pushing to 10X that. But, not necessarily day zero, but by stacking enough optimizations, if you have, say like four optimizations, each of which doubles performance. Or sorry, three optimizations, each of which doubles performance, then you stack that up, that's an 8X gain. That's the order of magnitude that we're working with in this space. We're trying to make things substantially faster, not just go from like 70 to 90.Swyx [00:35:38]: Are you saying you've. You have done that?Philip [00:35:40]: So let's say you have as a reasonable baseline, 30 or 40 tokens per second. You can achieve 10X that. So like on GLM-5.2, if you run it unquantized, perhaps on H100s even, and you're just using an off-the-shelf inference engine with no particular optimizations, no speculator, nothing extra around like KV routing, no disaggregation, you're, you're probably, yeah, looking at that like 30 to 40. You think that's like a reasonable baseline?Swyx [00:36:12]: Right. Right.Philip [00:36:12]: To get to something like 10X, there's a lot of trade-offs that you're making. If we're running at more like a 300, 400 tokens per second range, you are using the best hardware possible. You have a optimized speculator. You have done all of your quantization work. You are Seeing a pretty high cache hit rate. You are running with a reasonably small batch size and a parallelism configuration that is tuned for latency versus throughput, but it is possible. So the spreads that you see if you, like, go on artificial analysis or you go on OpenRouter and you look at, the worst provider to the best provider, oftentimes can hit that range. 10X is of course very aggressive. It's oftentimes maybe more of a four to six times improvement. But that's the performance that makes us really excited, is when we can get these huge gains, not just go from 70 to 90 tokens.Stacking Optimizations: NVFP4, Speculation, and DisaggregationAli [00:37:19]: It's also, like, hardware dependent. Like, ifPhilip [00:37:20]: YeahAli [00:37:20]: If you have a thing where you're serving it on just, like, a node of H100s and then you throw, like, you shard the model across, like, four nodes of B200s. Like, you can definitely increase the speed with just throwing more hardware at it. Like, normalizing for the same exact hardware and the same number of GPUs.Philip [00:37:35]: Yeah. Then you're looking at, like, a two to 4X improvementAli [00:37:38]: Right. RightPhilip [00:37:38]: Depending on the inference optimizations. So yeah, it's. Some of it's, what's the call, and some of it's who's the driver.Vibhu [00:37:46]: If you break down the two to 4X, say the example is run GLM-5.2Ali [00:37:51]: YeahVibhu [00:37:51]: On B200sAli [00:37:53]: YeahVibhu [00:37:53]: Single node, right? What's, like, the cost trade-off for effort to get, like, the last bit of juice out versus what should people just think of, right?Ali [00:38:01]: Spectre quantization. Yeah.Vibhu [00:38:03]: Spectre quantization.Ali [00:38:04]: That's, that's, that's like 95%. LikeVibhu [00:38:06]: And how far does that get you? And how easy is that for the average person to do? So say right I wanna throw the weights of GLM-5.2 on a node of B200s, how easy is it to find speculative decoder- decoder model or already quantized model? How much work goes into it?Philip [00:38:23]: If you're doing it up front, it's quite a lot of work. If you're doing it today, there's going to be people who have published things that you can just, you can just grab some NVFP4 weights. You can grab a speculator. Yeah, if we're thinking about, like, what are the 2Xs we're stacking, going from, BF16 to NVFP4 is, it's not quite a 2X, right? It's like. I think it's about, like, 30 to 40%, from 16 to 8, and then another 30 to 40% multiplied from, 8 to 4. So that doesn't quite get you a 2X, but, like, roughly a 2X. Speculator, roughly a 2X. Disagg on top of that if you're able to get enough hardware and put enough traffic through it, another roughly a 2X. And then you add in some, double-digit percent increase from having just a better runtime with, the latest kernels and stuff behind it. And that's how it stacks up.Ali [00:39:21]: YeahPhilip [00:39:21]: So building each of those, like, building the, quantized weights is, for someone who really knows what they're doing, hours to days of work. Building the speculator, again, like, hours to days of work. And the, disagg setup, hours to days. Well okay, but like once you haveAli [00:39:39]: Once set up. Once set up. YeahPhilip [00:39:40]: Yeah, getting disagg working for the first time, I'm saying, of course, is very difficult.Philip [00:39:44]: The marginal implementationAli [00:39:48]: Like, if you're just grabbing, like if you are a person, like just a normal consumer who has access to, like, a node of B200s and you're wondering, “How can I just host it myself?” You don't need to quantize the model yourself. There's always gonna be, like, an open source quantized checkpoint. NVIDIA's gonna push one out if no one else does. You. Usually, the providers will have their own spec dec that they've trained as well. You don't need to train your own spec dec. You can just use that as well.Philip [00:40:09]: Yeah. Like, GLM-5.2 has its own MTP.Ali [00:40:13]: Right. Right.Vibhu [00:40:14]: What's multi token prediction?Philip [00:40:15]: Yes.Ali [00:40:16]: I'm justVibhu [00:40:16]: Can you explain that?Ali [00:40:16]: I'm just an expert.Ali [00:40:18]: I can do it for you in case I get it wrong?Vibhu [00:40:20]: No.Vibhu [00:40:21]: Yeah, you should correct if we're wrong, but their multi-token prediction can be used for self-speculative decoding.Ali [00:40:27]: I'm not sure. I'm not gonna correct that.Vibhu [00:40:28]: Okay. I'm semi-confident in thatAli [00:40:30]: Okay. YeahVibhu [00:40:30]: But someone can check. But it's useful to paint the story of, okay, not just the average person, but say a company wants to switch from serverless inference I wanna throw this up on. I wanna rent some GPUs, throw it up. These are the steps you take to do significantly faster than just put it behind vLLM.Ali [00:40:48]: Right.Vibhu [00:40:49]: I was waiting for a mention of Dynamo.Vibhu [00:40:51]: I feel like, that's supposed to be the baseline that you measure against.Dynamo, KV Routing, and Disaggregation ToolkitsPhilip [00:40:55]: I would think of Dynamo as less of a box system and more of a toolkit for building with. So when we talk about doing aware routing, when we talk about doing KV offloading, when we talk about doing, PD disaggregation, Dynamo fundamentally is. By the way, Dynamo is an open source library from NVIDIA.Ali [00:41:17]: We've done a pod with KylePhilip [00:41:18]: OkayAli [00:41:19]: Kyle Cranin.Philip [00:41:19]: Cool. So then your listeners know then that it supports all the different inference frameworks. And it is multi hardware, which is interesting.Ali [00:41:28]: But it's just a router, it's not like an optimizer layer.Philip [00:41:30]: Yeah. All it does, like, what Dynamo is good at, it is a library for moving information around your cluster, around your hardware. So if you have, KV cache on one place and you need it to be somewhere else, Dynamo coordinates NIXL for you to move that around.Philip [00:41:49]: That doesn't mean that, like, out of the box, you just say, “Pip install Dynamo,” and then you get, like, a massive performance speed up. It's more of a developer toolkit.Ali [00:42:01]: Yeah. I would have said it would. It comes with a set of defaults that you can then swap out.Philip [00:42:06]: It does. If the industry at large, I think, was, like, rolling out all of these deployments, standard, then I think it would be, like, a credible baseline. But, we've got to, we've got to benchmark against, like, what we're seeing in the wild.Speculative Decoding Methods: Medusa, EAGLE, n-Gram, and Spec-SpecVibhu [00:42:23]: I did wanna talk a little bit more about PD disagg, because that is probably, like, number three after quantized and speculative decoding. In your book though, I was just gonna pull out the book.Philip [00:42:31]: Yeah.Vibhu [00:42:32]: Like section 522 on Medusa, 523 on EAGLEPhilip [00:42:35]: YeahVibhu [00:42:36]: 524 on gram.Philip [00:42:37]: It's 55, would be disaggregationAli [00:42:42]: Yeah. Well, no, I just wanted to dwell a little bitPhilip [00:42:44]: YeahAli [00:42:44]: The other. Like, so what do you choose to include? What do you choose to not to include? Because there was all these other techniques.Philip [00:42:51]: Yeah.Ali [00:42:51]: Are these still relevant? Because I think they came out, like, a year and a half ago maybe.Vibhu [00:42:55]: Medusa is quite old.Philip [00:42:56]: Yeah, Medusa's old.Ali [00:42:58]: It was old.Vibhu [00:42:58]: But is it in the book as a good, here'sPhilip [00:43:01]: BaselineVibhu [00:43:01]: Baseline vanilla understand it?Philip [00:43:02]: Like you should know this.Vibhu [00:43:03]: Like I read the paper, I'm like, “ it makes so much sense.”Philip [00:43:05]: Yeah.Philip [00:43:05]: So with the book, I had a couple goals. One was to give people just a working vocabulary for the space as a whole, and the other was to give them some intuition about how each of these techniques works. As I mentioned in my AI Engineer talk, which is the first public addendum to this, the speculation space has moved much faster than everything else. So yeah, even at the time that I wrote the book Medusa, I very much included as a way for people to understand how the space evolved rather than what the most modern technique is. And now of course, there's DFlash, dSpark. There's, there's newer techniques even than EAGLE, although EAGLE is still very commonly used.Ali [00:43:51]: SpecSpecta.Philip [00:43:52]: Yes. Speculative decoding.Vibhu [00:43:54]: What canAli [00:43:56]: Oh, it's a paper by Tri Dao and it's like, it's doing speculative decodingVibhu [00:44:00]: HuhAli [00:44:01]: For the speculative decoder.Philip [00:44:02]: Oh, in spec- oh my God.Ali [00:44:02]: It's literally just an another. It's like, yeah, that's the most simple way to explain it, and it seems like he got trivial speed ups there. But it seems that the complexity with training, it's almost like in our mind at least, it's almost as complex as training GANs. Like it's like a very delicate balance and oftentimes you, it's just but yeah, it's literally speculative decoding on speculative decoding.Vibhu [00:44:21]: Speculative.Ali [00:44:22]: Yeah. We saw this paper.Vibhu [00:44:24]: It's interesting, right?Ali [00:44:24]: Yeah.Vibhu [00:44:24]: I wouldn't even expect it to be very particular to train, I wouldAli [00:44:29]: Right.Vibhu [00:44:29]: The naive part of me is like, okay, train speculative decoder.Ali [00:44:32]: But like, and it makes sense, like the whole idea of speculative decoding is you. It's like, it's like almost like the iPhone auto predict version but for a normal model, right? Like you're just, you're just, generating three tokens and you're like, okay, I'll do prefill on them. And so you save those three turns for your original model. Now your speculative decoder is doing three turns of auto regression, so why not just have an even smaller model?Ali [00:44:53]: The other question there is what are the size of speculators? So say forPhilip [00:44:58]: Right. It's like a billion parameters.Ali [00:45:01]: Like for MiniMax, it's. Yeah. It's like one layer. It's like one 60th of the original model usually.Philip [00:45:06]: Yeah. I think we should do a paper when we get back to the office.Philip [00:45:10]: SpeculativeAli [00:45:11]: SpeculativePhilip [00:45:11]: Decoding.Ali [00:45:13]: No, it's, it does seem like how, when do you stop? But then it also seems like if you're able to train spec-spec decode for instance, right? Like if you're able to have a small model that is accurately predicts what the intermediate speculator is gonna predict, that is able to predict what the original target model's gonna predict, then why not just use that smallest model directly, right?Vibhu [00:45:34]: Yeah. This isAli [00:45:35]: Like it seems likeVibhu [00:45:35]: Adjacent to the routing problem.Ali [00:45:36]: Right.Vibhu [00:45:36]: Yeah.Ali [00:45:36]: Right.Philip [00:45:37]: The thing with speculators is one of the practical constraints on using them is that you do have to run a small model on the same hardware that you're running the big model on. There is a orchestration and resource competition problem inherent in that, and that is one of the constraints on speculation in general, is that draft tokens cost resources to create and cost software complexity to manage. And so if you have like infinitely recursive speculators, you add in quite a bit of that complexity on the actual implementation within the inference engine as well, not just in the training process.Vibhu [00:46:17]: I was gonna say, I would wonder if you could do similar, like distillation and pruning of, it's the same thing, it's just a model. Can we not just distill a lot of the weights, quantize the speculator, out of my domain? The question that also comes up is, this is all for big server workloads, right? How much of this applies to, say I have this MacBook, I wanna run Gemma really efficiently. Similar problems, not the same?Local AI vs. Data Center InferencePhilip [00:46:45]: Pretty different. I talked to Selo, about this on his podcast a couple weeks ago. The difference between inference engineering for the data center and for production workloads versus inference engineering for local AI, is that we start with fundamentally like different constraints and different goals. With local AI, it's how do I fit this model onto my hardware and then make it less dumb? And with data center influence, it's how do I load this model and then make it less slow? And we care about less dumb, and they care about less slow. But the local AI inference engineering ecosystem, I think has a lot for us to learn from in the data center space. They are experts in various forms of quantization, including dynamic quantization that we just don't touch, in the pruning, in the distillation, in the, layer removal. There'Ali [00:47:42]: Layer removal matters less.Philip [00:47:43]: Yeah. There'Ali [00:47:44]: No one loves pruning really.Philip [00:47:45]: Yeah. Well, but the, but they doVibhu [00:47:46]: Which is surprising, right? But that's, that's a whole different thingPhilip [00:47:48]: Just to fit something on the laptop.Ali [00:47:50]: Right.Philip [00:47:50]: So yeah, it's a, it's an interesting, it's an interesting space. Not necessarily that like their techniques make sense for us to do in the data center, because we have different resources and different goals, but more that the process as well as the openness of that field is something to, admire.Ali [00:48:12]: Yeah. Like to your point, like, certain optimizations that would. Like for instance, Turbo Quantum Sharper, like it made such huge hype on that and we did like a whole deep dive on Twitter and like said, what is it? How does it work? Why is it good or not? And it took off and it was implemented on local devices because your memory bandwidth is so slow on like a MacBook, for instance. But try putting the same thing on like an NVIDIA GPU on a B200 Turbo quant would not be. Like, it would not be used. Like, NVIDIA - Like, NVIDIA made it clear that this is not a good optimization, and we've seen it firsthand where the overhead of doing dequantization, quantization of, in the kernel itself with turbo quant kernel, each end is much slower than the time that you save from doing the bandwidth. ‘Cause on the B200s, you have like 3.5 terabytes per second. You don't need decrease the storage that much. You don't need to do, FP4 KV cache. You don't need to use a requant. There's, there's, there's better optimizations to be made. But on Edge devices, it's extremely important, it's extremely useful. So, seems to be, like, different optimizations there, but then they're all uniquely combined with like all you wanna quantize the model, you wanna do speculative decoding, like certain common prefixes with bothPhilip [00:49:18]: Principles.Ali [00:49:19]: Yeah, exactly. Exactly. Exactly.Philip [00:49:20]: They also do a lot of work on, model parallelism, especially over, heterogeneous topology, where you have, some sparks and they are wired together with, Ethernet, DGX sparks.Ali [00:49:35]: Yeah, this is the Exo Labs guys.Philip [00:49:36]: Yeah. You have, a nu

Zu viel versichert? Zu wenig? Versicherungsexperte Harry Büsser von Comparis erklärt die wichtigste Faustregel: Bevor man eine Police abschliesst, sollte man vier Fragen stellen – und zwar in dieser Reihenfolge. Die 4-Schritte-Faustregel – so geht man vor: Risiko vermeiden: Wer das Risiko gar nicht eingeht, muss es nicht versichern. Z.B. Teure Uhr zuhause lassen statt an den Strand mitnehmen. Im Ausland Bahn statt Mietwagen. Risiko vermindern: Einfache Massnahmen, die Schäden verhindern oder begrenzen. Z.B. Rauchmelder installieren, Helm tragen, Velo abschliessen, Laptop im Auto verdecken. Selbst tragen: Kleine Schäden, die einen nicht in finanzielle Not bringen, einfach aus der eigenen Tasche zahlen. Wer über Jahre Prämien zahlt, gibt oft mehr aus als er je an Schäden hätte. Versichern – aber nur bei echten Grossschäden: Die Schlüsselfrage: Würde mich dieser Schaden finanziell aus der Bahn werfen? Solche Risiken gehören versichert. Z.B. Kaputtes Handy oder zerbrochene Sonnenbrille – ärgerlich, aber kein Notfall. Ein Ambulanzjet aus Südostasien hingegen kostet schnell mehrere zehntausend Franken. Merksatz: «Versichern sollte man nicht den verlorenen Fünfliber, sondern den finanziellen Grossbrand.»

Vi håber, I har haft en skøn sommer og vi glæder os til at sende et helt normalt afsnit af Vin for begyndere for jer i næste uge, hvor det skal handle om Cigales i Castilla y Leon.   Her får I dog en lille hilsen fra os begge og vores sommerferie med vin.   -Jonas og René ..................... Køb vores nyeste bog "Bobler for begyndere og øvede" her: https://www.saxo.com/dk/bobler-for-begyndere_bog_9788773396568 Eller vores bog om vin her: https://www.saxo.com/dk/vin-for-begyndere_bog_9788773391303 Støt Vin for begyndere podcast her https://vinforbegyndere.10er.app/ Besøg os på Facebook og Instagram, hvor man kan se billeder af vinene og få tips til vin og mad sammensætning. https://www.facebook.com/vinforbegyndere https://www.instagram.com/vinforbegyndere Web: https://www.radioteket.dk/ Kontakt: radioteket@radioteket.dk Musik: Jonas Landin Lyt vores bog som lydbog her: Køb den her https://www.saxo.com/dk/vin-for-begyndere-og-oevede_lydbog_9788773397374

Kein Grillfeuer und auch kein Feuerwerk zum 1. August - wegen der Waldbrandgefahr gelten strenge Regeln. Und es drohen Bussen bis zu 20'000 Franken. +++ Weitere Themen: Wandern aus der Perspektive von Menschen mit Handicap - dank einem Audioguide. Und: Noroviren - Lidl ruft Beerenmischung zurück.

Männerabend #303 – Biere aus Franken! Dennis und Reinhold haben Besuch: Neue Woche – neues Glück: Der HopfenDino ist zurück! Nach „Biere aus dem Ruhrpott“ und „Biere aus Bayern“ folgt heute Franken und demnächst Thüringen (mit & bei BlockBräu-Braurein Anna) auf unserer flüssigen Reiseroute! Gemeinsam trinken wir uns heute also durch 5 von Dino handverlesene Biere aus Franken. Heute mit dabei: – Rittmayer Sicherheitshalbe – Brauerei Meister Vollbier – Fässla Zwergla – Special Rauchbier – Meinel Bräu Doppelbock Viel Spaß beim Hören! [Stream] Männerabend – Der Craft Beer Podcast: Spotify – iTunes – Deezer [Social Media] Männerabend – Der Craft Beer Podcast: Instagram – Facebook – Twitter Download: (Rechtsklick -> „Ziel speichern unter“) Männerabend #303 – Biere aus Franken // (adsbygoogle = window.adsbygoogle || []).push({}); // ]] // ]]>

Viral Load, two scheming viziers, and the French fell off.

Das ehemalige EWZ-Unterwerk Selnau, wo das Haus Konstruktiv war, steht jungen Zürcher Kunstschaffenden ab September ein Jahr lang fast kostenlos zur Verfügung. Nach Kritik aus der Kunstszene verzichtet das EWZ auf die bisher verlangten Betriebskosten – von zuerst 100'000 und dann 40'000 Franken. Weitere Themen: · Der FC Zürich startet mit einer 2:1-Niederlage beim FC St. Gallen in die neue Super-League-Saison: Umeh Emmanuel bringt den FCZ kurz nach der Pause in Führung, doch Lukas Görtler und Andrin Hunziker drehen die Partie innerhalb von vier Minuten. Für FCZ-Trainer Marcel Koller endet das erste Meisterschaftsspiel mit Zürich ohne Punkte. GC spielt gegen Lausanne ein 1:1-Unentschieden. · In der Schweiz lebt bereits jede 55. Person mit Demenz: Rund 160'000 Menschen sind betroffen, bis 2050 dürfte sich diese Zahl verdoppeln. Die Krankheit belastet auch Angehörige, die unter anderem in speziellen Ferienwochen Entlastung finden. Darum geht es in der Regionaljournal-Sommerserie zum Thema «Time out». · Wetter in Schaffhausen und Zürich: Der Montag beginnt mit vielen Wolken, vereinzelt ist etwas Regen möglich. Am Nachmittag setzt sich zunehmend die Sonne durch. Die Höchsttemperaturen liegen bei bis zu 27 Grad in Zürich und 28 Grad in Schaffhausen.

Die Privatbank Vontobel verzeichnet mit 216 Mio. Franken einen Rekord-Halbjahresgewinn. Co-Geschäftsleiter Georg Schubiger schreibt dies der hohen Handels- und Beratungsaktivität zu. Die Themen: Edelmetalle, KI-Aktien und Halbleiterindustrie generierten besonders viel Nachfrage und Transaktionen. SMI: +0.8%

In unserer Sommerserie geht es um Netzwerke: Heute mit einem Geflecht aus Stahldrähten, das Leben rettet. In den Bergen schützen Stahlnetze Strassen, Wanderwege und Schienen der RhB. Weiter in der Sendung: · Das Bündner Obergericht hat in einem Verfahren gegen einen Asylbewerber aus Burundi Fehler gemacht. Der Asylbewerber musste deshalb aus der Haft entlassen werden. · Der Gästesektor für Fans des FC St. Gallen für das Auswärtsspiel gegen Benfica Lissabon ist gesperrt. Trotzdem gibt es für St. Galler Fans die Möglichkeit, das Europa-League-Qualifikationsspiel in Portugal im Stadion zu verfolgen. Kostenpunkt für Tickets auf der Haupttribüne:1600 bis 2000 Franken.

Migrolino zieht den Streit um Sonntagsarbeit in Zürich und Winterthur vor das Verwaltungsgericht. Das Unternehmen stuft seine «Fresh»-Filialen als Gastrobetriebe ein. Die Zürcher Volkswirtschaftsdirektion betrachtet sie dagegen primär als Verkaufsläden und verweigert die Ausnahme vom Sonntagsschutz. Weitere Themen: · Wegen der Trockenheit gilt in den Kantonen Zürich und Schaffhausen ein Feuerverbot in Waldnähe: Mehr als 100 Zürcher politische Gemeinden haben zusätzlich strengere Regeln erlassen und teilweise Feuerwerk sowie Holzkohlegrills im ganzen Gemeinde- oder Stadtgebiet verboten. Für den 1. August sind vielerorts nur Gas- und Elektrogrills erlaubt. · Steinschlagschutznetze können Leben retten und sind besonders für die Rhätische Bahn in den Bergen von grosser Bedeutung: Sie investiert jährlich Millionen von Franken in Schutzbauten. Beim Bau neuer Netze stellen über 100 Jahre alte Bahnschienen eine besondere Herausforderung dar. · Wetter von SRF Meteo: Heute ist es meist sonnig, am Morgen ziehen jedoch einige Wolkenfelder von Norden nach Süden. Im Tagesverlauf bilden sich Quellwolken, die Höchstwerte liegen zwischen 20 Grad auf dem Hörnli und 27 Grad in Bülach und Schaffhausen.

Am 1. August feiert die Schweiz ihren Nationalfeiertag. Wegen der grossen Trockenheit gelten in den Kantonen Zürich und Schaffhausen jedoch strenge Regeln für Raketen, Feuerwerk und Höhenfeuer. Da die Vorschriften je nach Gemeinde variieren, lohnt sich ein Blick auf die Website der eigenen Gemeinde. Weitere Themen: · Nach dem «Fall Maisano» am Universitätsspital Zürich fordert die Eidgenössische Qualitätskommission, dass Spitäler die Interessenbindungen ihrer Ärzt:innen regelmässig offenlegen: Damit sollen finanzielle Interessenkonflikte verhindert werden, nachdem ein Bericht den Einsatz eines Medizinprodukts mit bis zu 70 Todesfällen bei Herzpatient:innen in Verbindung gebracht hatte. · Stadt Zürich will die Zahl der E-Trottis und anderer Mietfahrzeuge laut einem Medienbericht von mehr als 7000 auf rund 3000 senken: Zudem sollen künftig nur noch drei statt acht Anbieter:innen zugelassen werden, wobei die Umsetzung in den kommenden Monaten geplant ist. · Die Schaffhauser Polizei hat einen 23-jährigen mutmasslichen Betrüger festgenommen, der sich gegenüber einem 84-jährigen Mann als Polizist ausgegeben und 50'000 Franken erbeutet haben soll: Bei der versuchten Übergabe weiterer 10'000 Franken wurde der Mann verhaftet, das Geld konnte dem Opfer zurückgegeben werden.

Frank Richter ist zurück im Studiocast. Diesmal aber nicht zum Arbeiten, wie er gleich zu Beginn klarstellt, sondern zum Plaudern. Und das lohnt sich: Der Comedian und Christoph Soltmannowski teilen einen Background im People-Journalismus, und wenn zwei Ehemalige aus dieser Branche zusammensitzen, kommen Geschichten auf den Tisch, die man so noch nie gehört hat.Frank erzählt von seiner Zeit bei «Glanz & Gloria»: von der Managerin, die eine Frage nach Beatrice Eglis Kindheitsinstrument strich, weil sie «zu privat» sei. Von der Anwaltskanzlei, die pro «Schweini» 10 000 Euro forderte, weil man Bastian Schweinsteiger gefälligst beim vollen Namen zu nennen habe. Und von der Erkenntnis, dass die grössten Stars meist die unkompliziertesten sind, während das Management drumherum die Probleme macht. Christoph hält dagegen: mit einem Bond-Girl, einer Aufpasserin hinter seinem Rücken und einem Weltstar, dessen Presseassistentin keine Fotos wollte. Wegen der grauen Haare.Dazwischen geht es um das Handwerk: Wie wird aus einem TV-Journalisten ein Comedian? Frank erzählt, wie ihn das Blick Live Quiz vor die Kamera brachte und warum er sich am Ende zwischen SRF und Blick entscheiden musste. Er erklärt, weshalb schwarzer Humor in der Schweiz mainstreamtauglich verpackt sein muss, warum das deutsche Publikum schneller lacht und wieso er auf dem Land manchmal bremsen muss, damit die Pointen überhaupt ankommen. Und er hat eine Beobachtung mitgebracht, die hängen bleibt: Gen-Z-Zuschauerinnen schauen bei derben Witzen erst nach links und rechts, ob die anderen lachen. Die Lizenz zum Lachen holt man sich heute offenbar beim Sitznachbarn ab.Natürlich ist auch sein aktuelles Programm «Love My Job» Thema. Der Titel ist ironisch gemeint, denn Frank kann den LinkedIn-Hashtag nicht ausstehen: Wer vom Betriebsausflug auf den Pilatus ein Gruppenfoto mit #lovemyjob postet, will laut ihm vor allem eines, nämlich dem Chef Honig ums Maul schmieren. Über Social Media hat er trotzdem seine Meinung geändert. Sein Fazit nach dem späten Einstieg: Am Anfang ist es cringe, später ist es Content. Und es füllt die Theater.Ausserdem: Wie Sacha Baron Cohen als Ali G an sein Trump-Interview kam, warum Donald Trump überhaupt in «Home Alone 2» mitspielt, was Macaulay Culkin heute macht, wie Joel von Mutzenbecher als Regisseur mit Frank 32 Figuren samt eigener Stimme und Körperhaltung erarbeitet hat und weshalb Frank als kokainsüchtiger HR-Chef in einem Kurzfilm einen Ritterschlag erhielt.Eine Stunde über Comedy, Promis und die Frage, wie viel Inszenierung in der Spontaneität steckt. Reinhören lohnt sich.Alle Termine zu «Love My Job» gibt es auf Frank Richters Website und seinen Social-Media-Kanälen.Kapitel:(00:00) Intro: Taylor Swift und der starke Franken(01:40) Von der Lokalzeitung zu «Glanz & Gloria»(03:47) Beatrice Egli und die zwei Managerinnen(05:34) Bond-Girl, Prinz Albert und Paul McCartneys Haare(08:46) Haartransplantationen und graue Bartkränze(10:17) Blick Live Quiz und der Abschied vom SRF(14:37) Schwarzer Humor und die Grenzen der Comedy(16:55) Warum die Gen Z erst nach links und rechts schaut(20:20) Ali G, Trump und «Home Alone 2»(22:52) Kinderstars: Culkin, Jodie Foster, Sofia Coppola(26:01) Schlagfertigkeit: alles inszeniert?(28:53) Lady Gaga im Hallenstadion und Conan O'Brien(30:18) Deutsches Publikum vs. Schweizer Publikum(32:35) Stadt, Land und der Kantönligeist(35:07) «Love My Job»: Warum der Hashtag nervt(38:28) Social Media: Am Anfang cringe, später Content(42:16) Das vierte Programm und die Tour(43:21) Schauspiel: der kokainsüchtige HR-Chef(45:17) Joel von Mutzenbecher und 32 Figuren(46:44) Böse Mails: Schweini und andere EmpörteTickets gibt es bei Ticketcorner:https://www.ticketcorner.ch/search/?a...Mehr zu Frank Richter aufhttps://www.frankrichter.ch/

Cage joins the show live from Fanatics Fest to share his biggest takeaways from one of the hobby's premier events before Al Crisafulli of Love of the Game Auctions joins us for an in-depth discussion about the hobby's most talked-about card. We examine the restored "Franken-Wagner" from every angle, discussing restoration versus alteration, transparency, provenance, donor cards, and whether a reconstructed T206 Wagner should still be considered an authentic piece of hobby history. Al also shares fascinating comparisons to the restoration of iconic automobiles and other historic collectibles, providing important context for one of the hobby's most polarizing debates. Whether you love it, hate it, or simply haven't made up your mind, this conversation is one every collector should hear.

Die Hitze hat nicht nur Auswirkungen auf Fische und die Landwirtschaft. Aufgrund der Hitze und des geschmolzenen Schnees ergeben sich auch neue Herausforderungen im Hochgebirge. Einige Touren haben sich so verändert, dass Berggänger vorsichtig sein müssen, warnt der SAC. Weitere Themen: · Cyberkriminelle greifen Thurgauer Stadler Rail an: Vor ein paar Tagen ist Stadler Rail von Cyberkriminellen angegriffen worden. Die Firma hat das geforderte Lösegeld nicht gezahlt und bei der Kantonspolizei Strafanzeige eingereicht. · Bündner Polizei nimmt Betrüger fest: Ein 81-jähriger Mann wurde in Graubünden um mehrere zehntausend Franken gebracht. Die Kantonspolizei konnte zwei Männer ausfindig machen: einen 21-jährigen Kasachen und einen 24-jährigen Ukrainer. · Hin und Her um Bepflanzung in Wil: Die Stadt Wil unternimmt einen neuen Anlauf, um die Ladenstrasse und die obere Bahnhofstrasse mit einer Bepflanzung zu versehen. Im Moment stehen dort 28 Bäume – nicht mehr. · Sommerserie «Netzwerke» der Regionaljournale: Die Zahl der Fachkräfte, die aus dem Ausland kommen, um hier zu arbeiten, steigt. In der Schweiz wirklich anzukommen, ist jedoch deutlich schwieriger. Was hier helfen kann: Ein Netzwerk von Menschen, die diese Erfahrungen bereits gemacht haben.

Mit 16 Jahren verliert Fabian Kappeler auf dem Weg in die Berufsschule die Kontrolle über sein Motorrad. Der Unfall verändert sein Leben von einem Moment auf den anderen. Seither sitzt er im Rollstuhl.Plötzlich ist vieles anders als geplant: Ziele müssen neu definiert, Wege neu gefunden und Zukunftspläne überdacht werden. Doch statt aufzugeben, sucht Fabian pragmatisch nach Lösungen und macht unbeirrt weiter. Getragen von seiner Familie und seinen engsten Freunden.In dieser Folge von Querschnitt sprechen Fabian und seine Mutter Claudia offen über seinen Unfall, die Zeit danach und darüber, was Resilienz für sie bedeutet. Wie findet man neue Perspektiven, wenn das Leben eine unerwartete Wendung nimmt? Was gibt Kraft in schwierigen Momenten? Und wie gelingt es, trotz Rückschlägen nach vorne zu blicken?Eine ehrliche und inspirierende Geschichte über Mut, Anpassungsfähigkeit und die Kraft, weiterzumachen, wenn Pläne sich ändern.Moderation: Elena HirtDie Gönnervereinigung der Schweizer Paraplegiker-Stiftung (SPS) zählt 2 Millionen Mitglieder, die mit ihrem Mitgliederbeitrag das Leistungsnetz der Schweizer Paraplegiker-Gruppe ermöglichen und querschnittgelähmte Menschen unterstützen. Jeden zweiten Tag führt ein Unfall zu einer Querschnittlähmung. Mitglieder, die nach einem Unfall querschnittgelähmt und lebenslang auf den Rollstuhl angewiesen sind, erhalten 250 000 Franken – rasch und unbürokratisch. Weitere Infos: paraplegie.ch

Männerabend #302 – Biere aus Bayern! Dennis und Reinhold haben Besuch von unserem Lieblings-Dino: Dem HopfenDino! Nach dem großen Erfolg von „Männerabend #294 – Biere aus dem Ruhrpott“ wollen wir die regionalen Bier-Folgen fortsetzen. Nach dem Ruhrpott folgen jetzt also in den nächsten Wochen Bayern, Franken und Thüringen auf unserer flüssigen Reiseroute! Gemeinsam trinken wir uns heute also durch 5 von Dino handverlesene Biere aus dem blau-weißen Freistaat und, quasi als Geburtstags-Geschenk, für Reinhold (17.07.) gibt es auch noch den Jacob Weißbier-Kalender. Die Fotos des aktuellen Kalenders könnt ihr auch jeweils auf dem Jacob-Instagram-Kanal anschauen (KLICK). Heute mit dabei: – Giesinger Original Münchner Hell – Dampfbierbrauerei Zwiesel Dampfbier – Jacob Naturtrübes Hefe Weissbier – Zötler Maibock Hopf – Muospacher Bockfotzn Viel Spaß beim Hören! [Stream] Männerabend – Der Craft Beer Podcast: Spotify – iTunes – Deezer [Social Media] Männerabend – Der Craft Beer Podcast: Instagram – Facebook – Twitter Download: (Rechtsklick -> „Ziel speichern unter“) Männerabend #302 – Biere aus Bayern // (adsbygoogle = window.adsbygoogle || []).push({}); // ]] // ]]>

This week on Sports Cards Live, we kick off Hobby Palooza with a look at one of the busiest weekends the hobby has seen in years. We discuss the growing buzz around the restored "Franken-Wagner," why it has become the hobby's biggest conversation, what we're hearing from Fanatics Fest, and how West Coast Card Show is faring while competing with multiple major events. We also dive into the exploding premium being paid for elite eye appeal, highlighted by the record-setting Worldwide Gum Babe Ruth sale, and what it could signal for the vintage market. Plus, rumors from the show floor, thoughts on the National just around the corner, and plenty more hobby conversation.

Hier geht's zum Verhandlungs-Bootcamp: https://verhandlungs-bootcamp.com/AMEX-Karte beantragen: https://www.americanexpress.ch/de/karten/privatkunden-karten/platinum-card/ CODE: Platinum Card®️ beantragen und den Promo-Code: FA5BGQSJ4 im Online-Kartenantrag eingeben.Verhandeln passiert nicht nur im Büro — es passiert überall. Frédéric teilt drei echte Erlebnisse aus dem Urlaub: ein Mietwagen in Mexiko, ein Upgrade auf dem Flug von Toronto und ein Airbnb-Deal direkt beim Gastgeber. Drei Situationen, drei Lektionen — sofort anwendbar.In dieser Folge erfährst du:Wie ein einziger Satz am Mietwagenschalter über 600 Franken spartWarum meine Schwester Business Class flog — und die anderen nichtWie du bei Airbnb direkt beim Gastgeber buchst und >10% sparstWarum Urlaub die beste Verhandlungsschule istDie drei Grundprinzipien, die überall funktionieren — im Urlaub und im BusinessHier geht's zum Verhandlungs-Bootcamp: https://verhandlungs-bootcamp.com/Sichere dir deinen Platz und starte durch!Wenn du mit mir mal persönlich und live sprechen willst, dann buche dir hier einen Termin für deine kostenlose Verhandlungsstrategie:Abonniere diesen PodcastVernetzen wir uns auf Linkedin: https://www.linkedin.com/in/fredericmathier/Termin mit Frédéric buchen: www.fredericmathier.comDanke für deine ***** Bewertung auf iTunes oder SpotifyInstagram: https://www.instagram.com/frederic_mathier/TikTok: https://www.tiktok.com/@frederic_mathier/Wünsche dir erfolgreiche VerhandlungenFrédéric Mathier

Die insolvente Basler Confiserie Schiesser soll weitergeführt werden: Ein Interessent bietet rund 81'000 Franken für die Wort-Bild-Marke und das bewegliche Inventar Ausserdem: · Kundendaten der Industriellen Werke Basel entwendet

Der Spezialchemiekonzern Ems steigert den Gewinn im ersten Halbjahr 2026 – trotz höherer Rohstoffpreise, starkem Franken und anspruchsvollem Auto-Markt. Dafür bringt der KI-Sektor dem Unternehmen immer mehr ein: Jedes dritte Rechenzentrum kühle mit Ems-Komponenten, sagt Chefin Magdalena Martullo. SMI +0.1%

Der 82-Millionen-Kredit ist von der Gemeindeversammlung abgesegnet. Das Ja bedeutet, dass die Umsiedlung von Brienzerinnen und Brienzern vorangetrieben werden kann. Zulasten der Gemeinde Albula/Alvra fallen gerade mal noch maximal 360'000 Franken. Weitere Themen: · Wasserversorger rund um Wil stossen bei der aktuellen Hitzeperiode an ihre Grenzen. · In Buchs wurde der erste Japankäfer im Kanton St. Gallen gesichtet. · Bars, Public Viewings und Behörden sind mit ungewöhnlicher Situation konfrontiert, da das nächste Spiel der Schweizer Fussballnationalmannschaft morgens um 3.00 Uhr losgeht. · Im geplanten neuen Stadtpark von Wil soll es doch keinen Badeweiher geben. · Das Kulturfestival St. Gallen feiert sein 20-Jahr-Jubiläum.

430 Franken für ein Taxi, unnötige Amazon Käufe und die Frage, warum man ausgerechnet danach plötzlich anfängt, überall sparen zu wollen.

Welcome to another installment of the ChinaTalk radio show! Today, we're diving into Taiwan's war on green energy. Shenanigans abound in this episode, including: The lights-out scenario — Taiwan only holds 11 days of LNG reserves, and 97% of the island's energy is imported, but the ruling party phased out nuclear and botched the renewable rollout anyway. The offshore wind graveyard — how made-in-Taiwan components drove developers to abandon the world's best offshore wind sites, The Taipower unbundling reversal — and the Kafkaesque system that keeps electricity prices dirt cheap despite the Iran war. “Green energy cockroaches” — why corruption is Taiwan's dirtiest secret, and how the Taiwanese public came to associate renewables with scandal, The nuclear U-turn — How President Lai Ching-te walked back forty years of "Non-Nuclear Homeland" orthodoxy to restart Taiwan's nuclear reactors. A transcript of this show with embedded source links is available on the ChinaTalk substack. This episode was produced by Lily Ottinger and Aqib Zakaria. Special thanks to "Jason Feng," Angelica Oung, Ricky Huang, Tsaiying Lu (DSET), and Yu-Hsuan Yeh (formerly of CSIS and DSET) for their time and expertise. Everyone's views are their own and don't represent any organization. If you want to learn more, check out Angelica's ongoing work on her two Substacks, Taipology and Elemental Energy. You can also check out Ricky's two podcasts, where he hosts cross-partisan debates about energy policy and more. "Jason's" voice was anonymized with ElevenLabs' text-to-speech tools. Finally, we know Angelica is a controversial figure, but we decided to interview her because, on energy policy specifically, her views are shared by a not-insubstantial portion of the Taiwanese public. [See: this poll which reported that 59% of the Taiwanese public didn't feel confident that Lai's administration could protect Taiwan from power outages, and this poll from June 2025 that shows a near-even split in public opinion for and against the non-nuclear homeland policy.]  Outro song lyrics: 「燈火 Taiwan」 (Lights of Taiwan) [Verse 1] The AC stopped humming on August day eight Aunties in the market, no fan on their face Eleven days of gas, forty-two of coal Then the island goes dark, and the story gets old O-lóng-mn̂g, o-lóng-mn̂g (黑黑暗暗, pitch black) We knew this would come, but we looked away [Pre-Chorus] Forty years they said hūi-hi̍k (非核, non-nuclear) Forty years of dreaming we could wish it all away But the strait is a wind tunnel, and the sun still shines While we burned the future for cheaper times [Chorus] Góa ê kò͘-hiong, lí kám ū thêng-thāu? (我的故鄉, 你敢有聽著? — My homeland, can you hear?) The Franken-reactor sleeps beneath the hill Crystal Yang drank the water, but the people got ill Góa ê kò͘-hiong, lí ài kiàⁿ-khí-lâi (我的故鄉, 你愛起來 — My homeland, you must rise) Not nuclear OR green — we need both to survive [Verse 2] Round 3.1, Round 3.2, localization chains RWE went home, EnBW felt the pain Yunlin's turbines turning, three times the cost While the lūi-chhù (綠能蟑螂, green cockroaches) ate what we lost Behind the meter, batteries wait Zero price auction — we sealed our own fate [Pre-Chorus] Taipower's black box, CPI's lie TSMC pays more so the auntie don't cry But the data centers can't grow, AI waits at the door While we argue if nuclear is sin or chó͘ (善或惡, good or evil) [Chorus] Góa ê kò͘-hiong, lí kám ū thêng-thāu? The Franken-reactor sleeps beneath the hill Crystal Yang drank the water, but the people got ill Góa ê kò͘-hiong, lí ài kiàⁿ-khí-lâi Not nuclear OR green — we need both to survive [Bridge] (Spoken, over soft piano) March 22nd, 2026 Lai Ching-te said the words nobody wanted to hear Kò͘-hiong needs power Not slogans, not pride, not forty years of fear [Final Chorus] Góa ê kò͘-hiong, lí kám ū thêng-thāu? The blockade is coming, the Hormuz is closed Spot market gas at 140% — who knows? Góa ê kò͘-hiong, lí ài kiàⁿ-khí-lâi Distributed and hardened, let the sun and wind rise With nuclear beside them — open both your eyes [Outro] O-lóng-mn̂g, mài koh o-lóng-mn̂g (黑黑暗暗, 莫閣黑黑暗暗 — Darkness, don't be dark again) Kiàⁿ-khí-lâi, Tâi-oân (起來, 台灣 — Rise up, Taiwan) Kiàⁿ-khí-lâi... ChinaTalk is an audience-supported publication. If you'd like to help us produce more content like this, please consider a paid subscription on Substack. Learn more about your ad choices. Visit megaphone.fm/adchoices

Welcome to another installment of the ChinaTalk radio show! Today, we're diving into Taiwan's war on green energy. Shenanigans abound in this episode, including: The lights-out scenario — Taiwan only holds 11 days of LNG reserves, and 97% of the island's energy is imported, but the ruling party phased out nuclear and botched the renewable rollout anyway. The offshore wind graveyard — how made-in-Taiwan components drove developers to abandon the world's best offshore wind sites, The Taipower unbundling reversal — and the Kafkaesque system that keeps electricity prices dirt cheap despite the Iran war. “Green energy cockroaches” — why corruption is Taiwan's dirtiest secret, and how the Taiwanese public came to associate renewables with scandal, The nuclear U-turn — How President Lai Ching-te walked back forty years of "Non-Nuclear Homeland" orthodoxy to restart Taiwan's nuclear reactors. A transcript of this show with embedded source links is available on the ChinaTalk substack. This episode was produced by Lily Ottinger and Aqib Zakaria. Special thanks to "Jason Feng," Angelica Oung, Ricky Huang, Tsaiying Lu (DSET), and Yu-Hsuan Yeh (formerly of CSIS and DSET) for their time and expertise. Everyone's views are their own and don't represent any organization. If you want to learn more, check out Angelica's ongoing work on her two Substacks, Taipology and Elemental Energy. You can also check out Ricky's two podcasts, where he hosts cross-partisan debates about energy policy and more. "Jason's" voice was anonymized with ElevenLabs' text-to-speech tools. Finally, we know Angelica is a controversial figure, but we decided to interview her because, on energy policy specifically, her views are shared by a not-insubstantial portion of the Taiwanese public. [See: this poll which reported that 59% of the Taiwanese public didn't feel confident that Lai's administration could protect Taiwan from power outages, and this poll from June 2025 that shows a near-even split in public opinion for and against the non-nuclear homeland policy.]  Outro song lyrics: 「燈火 Taiwan」 (Lights of Taiwan) [Verse 1] The AC stopped humming on August day eight Aunties in the market, no fan on their face Eleven days of gas, forty-two of coal Then the island goes dark, and the story gets old O-lóng-mn̂g, o-lóng-mn̂g (黑黑暗暗, pitch black) We knew this would come, but we looked away [Pre-Chorus] Forty years they said hūi-hi̍k (非核, non-nuclear) Forty years of dreaming we could wish it all away But the strait is a wind tunnel, and the sun still shines While we burned the future for cheaper times [Chorus] Góa ê kò͘-hiong, lí kám ū thêng-thāu? (我的故鄉, 你敢有聽著? — My homeland, can you hear?) The Franken-reactor sleeps beneath the hill Crystal Yang drank the water, but the people got ill Góa ê kò͘-hiong, lí ài kiàⁿ-khí-lâi (我的故鄉, 你愛起來 — My homeland, you must rise) Not nuclear OR green — we need both to survive [Verse 2] Round 3.1, Round 3.2, localization chains RWE went home, EnBW felt the pain Yunlin's turbines turning, three times the cost While the lūi-chhù (綠能蟑螂, green cockroaches) ate what we lost Behind the meter, batteries wait Zero price auction — we sealed our own fate [Pre-Chorus] Taipower's black box, CPI's lie TSMC pays more so the auntie don't cry But the data centers can't grow, AI waits at the door While we argue if nuclear is sin or chó͘ (善或惡, good or evil) [Chorus] Góa ê kò͘-hiong, lí kám ū thêng-thāu? The Franken-reactor sleeps beneath the hill Crystal Yang drank the water, but the people got ill Góa ê kò͘-hiong, lí ài kiàⁿ-khí-lâi Not nuclear OR green — we need both to survive [Bridge] (Spoken, over soft piano) March 22nd, 2026 Lai Ching-te said the words nobody wanted to hear Kò͘-hiong needs power Not slogans, not pride, not forty years of fear [Final Chorus] Góa ê kò͘-hiong, lí kám ū thêng-thāu? The blockade is coming, the Hormuz is closed Spot market gas at 140% — who knows? Góa ê kò͘-hiong, lí ài kiàⁿ-khí-lâi Distributed and hardened, let the sun and wind rise With nuclear beside them — open both your eyes [Outro] O-lóng-mn̂g, mài koh o-lóng-mn̂g (黑黑暗暗, 莫閣黑黑暗暗 — Darkness, don't be dark again) Kiàⁿ-khí-lâi, Tâi-oân (起來, 台灣 — Rise up, Taiwan) Kiàⁿ-khí-lâi... ChinaTalk is an audience-supported publication. If you'd like to help us produce more content like this, please consider a paid subscription on Substack. Learn more about your ad choices. Visit megaphone.fm/adchoices

Is your organization running on a Franken-stack.That unmanaged tech stack that is layers of disconnected software tools piled on top of each other. In the race to add more technology as companies scale, managing tech stacks doesn't look like a priority.And yet, vital customer data - that kills net margins - can get dropped between marketing attribution and CRM logs. Today we're joined by Laura Farkas, LMNts Marketing, which provides revenue system optimisation, funnels and automation.She's a specialist in building inbound distribution networks that clients can control.On this episode of A Seat at The Table, we'll be discussing•Why is your connection count on LinkedIn completely irrelevant to your actual inbound revenue?•What is a Franken-stack and how does it drain operating margins from a mid-market company?•Why are traditional marketing retainers and basic lead generation agencies failing?•How does a twenty-million-pound B2B enterprise structure a proper revenue systems audit?•How can a business owner use intentional human imperfection to defeat automated software noise?So let's sit down with Laura and learn how to wrestle that monstrous tech-stack into a lean, sales driving machine.Connect with Laura Farkas:Main book: Paternoster Marketing https://www.amazon.com/Paternoster-Marketing-Never-Leads-Again-ebook/dp/B0FTXKKQB6 Laura's Revenue Systems page, based on the Paternoster Marketing concept:  https://marketingfunnel.website/b2b-revenue-system/ LinkedIn: https://www.linkedin.com/in/itslaurafarkas/YouTube: https://www.youtube.com/@LauraFarkasLMNtsMarketing SEND US A MESSAGEVisit A Seat at The Table's website at https://seat.fm

Die Sommerferien stehen vor der Tür - und mit ihnen die große Frage: Kann Urlaub mit Kindern überhaupt entspannend sein? Zwischen Organisation, Kosten, Erwartungen und den Bedürfnissen aller Familienmitglieder wird die gemeinsame Auszeit oft zum Balanceakt.Für diese Folge von "Zwischen Hamburg und Haiti" besucht Reporterin Clea Schnitzlein einen Ferienhof in Franken. Dort erleben Kinder Ferien zwischen Ziegenfütterung, Traktorfahrten und Lagerfeuer. Eltern erzählen, warum sie sich bewusst für Urlaub auf dem Bauernhof entscheiden und was für sie eine gelungene Familienreise ausmacht.Außerdem spricht die Tourismussoziologin Kerstin Heuwinkel über aktuelle Trends im Familienurlaub, den wachsenden Anspruch an die perfekte Auszeit und darüber, wie sich veränderte Familienstrukturen auf das Reiseverhalten auswirken.Doch die Folge blickt auch auf die Realität vieler Familien: Nicht alle können sich eine Urlaubsreise leisten. Alleinerziehende berichten von finanziellen Herausforderungen und davon, wie sie ihren Kindern trotzdem unvergessliche Ferien ermöglichen wollen.weitere Informationen:Tourismussoziologin Kerstin Heuwinkelhttps://kerstin-heuwinkel.de/Zahlen des Deutschen Tourismusverbandes 2025https://www.deutschertourismusverband.de/fileadmin/user_upload/Footer/Presse/Zahlen-Daten-Fakten_2025.pdfReportage: Wenn die Ferienzeit zur finanziellen Herausforderung wirdhttps://www.ndr.de/nachrichten/info/jugenderholungswerk-100.htmlPodcast-Tipp:Wenn Ihr wissen wollt, wie andere Eltern ihren Alltag gestalten, hört doch mal rein in "Eltern ohne Filter". Mütter und Väter erzählen ungefiltert von ihrem Leben als Eltern. Vom irrsinnigen Glück. Vom ganz normalen Wahnsinn. Und von ihren dunklen Momenten. Es ist und bleibt ein Abenteuer.Eltern ohne Filter · Neue Folgen - Jetzt Podcast anhören!https://www.ardsounds.de/sendung/urn:ard:show:821f9bc1b027e65d/

Immer wieder werden in Venezuela Nachbeben registriert, nach den zwei schweren Erdbeben, die weite Teile des Landes verwüstet haben. Auch für grössere Nachbeben besteht ein Risiko, erklärt Stefan Wiemer, Leiter des Schweizerischen Erdbebendiensts. Weitere Themen: · Nach Venezuela reist auch ein Rettungsteam aus der Schweiz. Mit dabei sind acht Hunde vom Rettungsdienst Redog, einem Verein, der Such- und Rettungshunde ausbildet. Zentralpräsidentin Linda Hornisberger hat den Einsatz koordiniert. · Wer ein iPad oder ein Macbook kaufen möchte, muss tiefer in die Tasche greifen. Apple ehöht nämlich weltweit seine Preise. Für einen Laptop bedeutet schnell Extrakosten in Höhe von 200 Franken. Jan Baumann erklärt, warum Apple diesen Schritt macht. · Am Montag wird die Google-Tochter Alphabet in den Dow Jones aufgenommen. Für Alphabet ist das nicht nur eine gute Nachricht. Grund ist: Der sogenannte «dow curse», also der «Fluch des Dow Jones». Was es damit auf sich hat, erklärt US-Börsenkorrespondent Jens Korte. · Europa leidet unter der Hitze. In Frankreich fallen AKWs und Züge aus, Schulen bleiben geschlossen, die Krankenhäuser kommen an ihre Grenzen. Auch in Spanien werden Rekordtemperaturen gemessen. Und doch scheinen die Probleme dort weniger gravierend zu sein als etwa in Frankreich. Was macht Spanien anders? Die Einschätzung von Julia Macher, freie Journalistin in Barcelona. · Nicht nur Menschen, sondern auch Tiere passen sich an die Hitze an. Zum Beispiel Vögel. Das funktioniert ähnlich wie bei hechelnden Hunden, erklärt Livio Rey von der Vogelwarte Sempach.

Clive Davis/Diddy rumors, gross Frankenstein rabbits, cute cat sleep habit, Austin Metcalfs dad goes off, Tulsi Gabbard finally does something and more! Check out our amazing sponsors! Use code 'NERD' on binoid.com to get the best thc product on the market and save 25% using the code + get free shipping! Twitter/Mewe/Parler/Gettr/Rumble/tiktok: @voicesofmisery Gmail: voicesofmiserypodcast@gmail.com Discord server: voices of misery podcast https://tinyurl.com/VoMPodcastTees Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Have an idea or tip? Send us a text!You can put a famous face next to a brand and still end up with marketing that goes nowhere. What actually moves people is trust, specificity, and a story that feels human, especially in the photo imaging industry where the product is memory, emotion, and identity.Gary Pageau sits down with Katherine Tuominen from Catalyst Brand Strategy to unpack ethical marketing in a practical way: serving the customer's needs, avoiding misleading edits or “Franken-grabbing,” and building a message you can repeat without sounding robotic. We get concrete about brand storytelling, from finding the turning point that sparked the business to adding the gritty details that make someone say, “I've been there.” Tuominen shares how to keep your core narrative consistent while refreshing it with timely lenses like AI, anniversaries, and local moments.We also talk photography business marketing beyond discounts, including how to sell the value of photo printing, photo books, wall art, and prints by leaning into tactile, analog appeal. If you're thinking about in person events, we break down “event activations,” strategic partnerships, sponsors, and how to track ROI with QR codes, UTMs, email nurture campaigns, and clear objectives. Wrap it all with a hard truth for small business owners: growth often requires saying no and focusing on what actually moves the needle.Energize your sales with Shareme.chat, the proven texting platform. ShareMe.Chat ShareMe.Chat platform uses chat-to-text on your website to keep your customers connected and buying!MediaclipMediaclip strives to continuously enhance the user experience while dramatically increasing revenue.Buzzsprout - Let's get your podcast launched!Start for FREEIndependent Photo ImagersIPI is a member + trade association and a cooperative buying group in the photo + print industry.Photo Imaging CONNECTThe Photo Imaging CONNECT conference, March 2027, at the RIO Hotel and Resort in Las Vegas, NDisclaimer: This post contains affiliate links. If you make a purchase, I may receive a commission at no extra cost to you.Support the showSign up for the Dead Pixels Society newsletter at http://bit.ly/DeadPixelsSignUp.Contact us at gary@thedeadpixelssociety.comVisit our LinkedIn group, Photo/Digital Imaging Network, and Facebook group,  The Dead Pixels Society. Leave a review on Apple and Podchaser. Are you interested in being a guest? Click here for details.Hosted and produced by Gary PageauAnnouncer: Erin Manning

We've seen spoof horror flicks before, but have we seen one that includes a bunch of stoners heading to Romania to party and get attacked by vampires? Unfortunately, we have. Join your bloodsucking hosts, Sam and Liz as they talk about Franken-bodies, wild west train shootouts and the dangers of accidentally shooting yourself with a crossbow in 2009's Transylmania. Want more screams and laughs? Join our Fright Club at http://patreon.com/frightmicpodcast and get access to tons more episodes, discussions, rankings, watch parties and more!Fright Mic is an independent horror podcast. We would love to have you join our Fright Fam by following us on all our socials!PATREONMERCHFacebookFRIGHT CLUBInstagramBlueskyTwitterTiktokDiscordSupport the show

 Guest Host Heather McGhee, author of the groundbreaking book “The Sum of Us” and expert in economic and social policy, is joined by The Atlantic's Adam Serwer! They discuss a number of issues facing the country, including the continued unrest in Minnesota and the No Kings protests happening nationwide. From the steps of the Supreme Court to the streets of Minneapolis, McGhee and Serwer explore how the struggle for Black freedom has historically paved the way for immigrant rights and why that "inheritance" is currently under its most aggressive attack in generations.Heather and Adam also examine the recent SCOTUS arguments on birthright citizenship and how this could create a whole new class of undocumented immigrants in this country, particularly Chinese immigrants. They also examine Adam's recent writing following the death of Rev. Jesse Jackson  about what Rev. Jackson's message meant to the movement. They dissect his 1980's vision of cross-racial, cross-class solidarity and explain why his "Patchwork Quilt" philosophy is more relevant today than ever.LEARN more about Heather McGhee: https://heathermcghee.com/READ Adam Serwer's writing in The Atlantic: https://www.theatlantic.com/author/adam-serwer/SUPPORT THE SHOW BY VISITING OUR SPONSORS:Visit American Giant to get their Classic Full Zip Hoodie and other cold weather staples. Get 20% off your first order by entering the code FRANKEN at checkout! https://www.american-giant.comRefresh your wardrobe with Quince! Get free shipping and 365-Day returns at https://www.quince.com/franken