POPULARITY
Categories
60 episodes with Spectre
34 episodes with Spectre
28 episodes with Spectre
26 episodes with Spectre
16 episodes with Spectre
42 episodes with Spectre
42 episodes with Spectre
12 episodes with Spectre
34 episodes with Spectre
15 episodes with Spectre
19 episodes with Spectre
17 episodes with Spectre
19 episodes with Spectre
48 episodes with Spectre
17 episodes with Spectre
15 episodes with Spectre
30 episodes with Spectre
12 episodes with Spectre
10 episodes with Spectre
13 episodes with Spectre
17 episodes with Spectre
7 episodes with Spectre
13 episodes with Spectre
7 episodes with Spectre
12 episodes with Spectre
12 episodes with Spectre
10 episodes with Spectre
9 episodes with Spectre
7 episodes with Spectre
28 episodes with Spectre
4 episodes with Spectre
15 episodes with Spectre
18 episodes with Spectre
18 episodes with Spectre
7 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
5 episodes with Spectre
10 episodes with Spectre
8 episodes with Spectre
8 episodes with Spectre
8 episodes with Spectre
7 episodes with Spectre
7 episodes with Spectre
25 episodes with Spectre
4 episodes with Spectre
9 episodes with Spectre
8 episodes with Spectre
4 episodes with Spectre
27 episodes with Spectre
7 episodes with Spectre
13 episodes with Spectre
9 episodes with Spectre
7 episodes with Spectre
12 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
9 episodes with Spectre
5 episodes with Spectre
11 episodes with Spectre
6 episodes with Spectre
11 episodes with Spectre
14 episodes with Spectre
6 episodes with Spectre
5 episodes with Spectre
5 episodes with Spectre
6 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
5 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
13 episodes with Spectre
6 episodes with Spectre
3 episodes with Spectre
9 episodes with Spectre
3 episodes with Spectre
7 episodes with Spectre
3 episodes with Spectre
9 episodes with Spectre
7 episodes with Spectre
6 episodes with Spectre
7 episodes with Spectre
4 episodes with Spectre
5 episodes with Spectre
12 episodes with Spectre
18 episodes with Spectre
7 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
10 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
5 episodes with Spectre
5 episodes with Spectre
7 episodes with Spectre
5 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
5 episodes with Spectre
4 episodes with Spectre
6 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
5 episodes with Spectre
8 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
9 episodes with Spectre
3 episodes with Spectre
5 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
3 episodes with Spectre
3 episodes with Spectre
6 episodes with Spectre
5 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
2 episodes with Spectre
3 episodes with Spectre
2 episodes with Spectre
6 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
4 episodes with Spectre
3 episodes with Spectre
3 episodes with Spectre
2 episodes with Spectre
3 episodes with Spectre
5 episodes with Spectre
3 episodes with Spectre
4 episodes with Spectre
9 episodes with Spectre
6 episodes with Spectre
6 episodes with Spectre
9 episodes with Spectre
6 episodes with Spectre
9 episodes with Spectre
4 episodes with Spectre
6 episodes with Spectre
6 episodes with Spectre
6 episodes with Spectre
He's a figure of vengeance. He wears a long cloak. He strikes fear into the hearts of criminals. He's... The Spectre. God's wrath made manifest, The Spectre is the angel of vengeance bonded to the ghost of a dead New York cop, Jim Corrigan. Unsurprisingly, a spirit of vengeance is a character who has teamed up with Batman on more than one occasion. First, we get some of that joyous Bob Haney madness, followed up by Doug Moench and Kelly Jones and finally we go into the New 52 and we see Corrigan as a member of the GCPD's answer to the X-Files. In the Grasp of Shahn-Zi (The Brave and the Bold V.1 # 75) Spirit of Vengeance (Batman V.2 # 540-541) We Do Not Sleep (Gotham By Midnight # 1-5) Check out our current ranking list at www.comicsxf.com/batchat-rankings/ Thanks to Geri Nonnewitz for our podcast logo Support the show on Patreon at www.patreon.com/batchatwithmattandwill
Key Segments & Discussion Breakdown1. Dead End Tours: When the Guide Isn't the Only One Talking[3]Paul presents four firsthand stories from people who experienced real paranormal encounters while taking organized ghost tours[6][14]:The Yellow Fever Cemetery (Savannah, Georgia):The Story: Sarah, a history teacher, attended a night ghost tour with a colleague in Savannah[6][15]. Leaning against a wrought-iron gate near an old cemetery, she felt a heavy, cold hand rest on her shoulder, accompanied by a cold body leaning full weight against her[15]. A voice whispered in her ear: "Why are you standing on my porch, honey?" before physically bumping her[18].The Twist: Five minutes later, the tour guide noted that the exact spot where Sarah stood was the former entrance of a residential house that burned down in 1820[19].The Blair Street Vaults (Edinburgh, Scotland):The Story: Mark, an IT consultant from London, was touring the claustrophobic subterranean vaults of Edinburgh[20]. After stopping to tie his shoe, he looked up to find his tour group gone[20][21]. At a T-junction in the hall, he saw a man in a long wool coat and flat cap who motioned him into a side doorway[22][23].The Twist: The room was completely empty, smelling of dust and decay, and the man vanished[23]. In the corridor stood an old portrait matching the man's appearance[23]. Seconds later, the tour group arrived from behind him, baffled as to how Mark got ahead of them past a locked red velvet rope[24][25].The Haunted Mansion Museum (French Quarter, New Orleans):The Story: Francine visited a converted 1890s mansion museum on a four-person tour[26][27]. Looking into a large vanity mirror, she saw the reflection of a woman in period dress standing in the room behind her[27][28]. When she turned around, no one was there, but looking back into the mirror, the woman was face-to-face with her inside the glass[28].The Manifestation: Retreating to her car, Francine's phone signal died, and a haunting music box melody filled the cabin[29][30]. A white mist formed in the passenger seat into the same ghostly woman, who touched Francine's wrist and wept: *"You have a wonderful mother... She's not cross to you and she's not cruel like my mother... She's the one who put me in my grave. You must pity me."*[30]2. The Chronos Effect & Time Leaks[4][5]The hosts pivot to temporal anomalies where individuals temporarily slip into the past or future[4]:The Moberly–Jourdain Incident (Palace of Versailles, 1901):Oxford professors Charlotte Anne Moberly and Eleanor Jourdain got lost searching for the Petit Trianon[7]. The atmosphere suddenly grew cold and heavy, colors muted like a painting, and they encountered 18th-century figures—including Marie Antoinette—alongside the stagnant stench of historical 1789 decay[35].The Bold Street Slip (Liverpool, 1996):A man named David was searching for a shop when modern pavement turned to cobblestone and modern cars disappeared[8]. He encountered Victorian shops and pedestrians, one of whom recoiled in pure terror at the sight of David and his modern Saab, perceiving him as a ghost[8].Sir Victor Goddard's Flight (Drem Airfield, Scotland, 1935):RAF pilot Sir Victor Goddard flew over an abandoned, dilapidated airfield[9][34]. Suddenly, he saw the airfield fully restored, complete with mechanics in yellow suits and unfamiliar monoplanes[9][34]. Four years later in 1939, the military rebuilt the airfield exactly as he had seen it—indicating a slip into the future[34][40].Theoretical Discussion:Paul and Michelle discuss time paradoxes, parallel timelines, the butterfly effect, and the concept that all existence occurs within one continuous moment of eternal consciousness[10].3. Personal Story: Confusing a Ghost[12]Michelle shares a story from junior high where she and her friends held a home-made séance[12][13].When a spirit began knocking on walls in response to questions, Michelle overwhelmed the entity with contradictory and overly complex conditional commands ("If you're a spirit but not really here, knock three times, then two...")[13].The frustrated ghost banged violently on the wall and vanished forever[13].Notable Quotes"You pay 20 bucks, you follow a person in time-period appropriate clothing through a dark alley, and you hope to see something that makes your skin crawl." — Paul[3]"You aren't editing the book. You're just reading a different page." — Michelle[10]"There is really truly only one moment of consciousness that has ever existed, and it's this moment right now." — Michelle[42] Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
Listen Ad Free https://www.solgoodmedia.com - Listen to hundreds of audiobooks, thousands of short stories, and ambient sounds all ad free!
O que um exilado brasileiro em Paris nos anos 1970 e um apoiador de Bolsonaro em Orlando têm em comum? Para Teresa Cristina Schneider Marques (PUCRS), pós-doutoranda na UERJ com apoio da Faperj, ambos ajudam a entender por que o exílio deixou de significar isolamento e passou a ser estratégia política. Neste episódio, produzido em parceria com o Observatório da Extrema-Direita, David Magalhães (UFU / OED) e Guilherme Casarões (FIU / OED) recebem Teresa para discutir a transnacionalização do bolsonarismo, tema de sua pesquisa de campo recente na Flórida e de trabalhos como “Migrantes internacionais na extrema direita e as frentes de transnacionalização do bolsonarismo nos EUA”. A conversa percorre o voto da diáspora brasileira nos Estados Unidos, o papel das igrejas e das organizações de migrantes, a ofensiva de Eduardo Bolsonaro e Paulo Figueiredo junto ao governo Trump e os limites de um ativismo que pode sair pela culatra. No boletim de notícias, David Magalhães analisa as vitórias da AfD na Saxônia-Anhalt e em Mecklemburgo-Pomerânia Ocidental, a largada da corrida presidencial francesa com Marine Le Pen e Éric Zemmour, o acordo migratório entre Jordan Bardella e Nigel Farage e o improvável encontro entre Steve Bannon e Bernie Sanders, sintoma das fissuras dentro do trumpismo. Na abertura, David apresenta o NETEX, novo núcleo da UFU dedicado ao estudo da transnacionalização e do extremismo. E na dica cultural, o documentário Anatomia do Caos, de Dandara Ferreira, sobre os bastidores da CPI da Covid. Aperte o play! Quer apoiar o Chutando a Escada? Acesse chutandoaescada.com.br/apoio Mande um café usando nossa chave PIX: perguntas@chutandoaescada.com.br Comentários, críticas, sugestões? Escreva pra gente em perguntas@chutandoaescada.com.br Participaram deste episódio: David Magalhães (UFU / OED), Guilherme Casarões (FIU / OED) e Teresa Cristina Schneider Marques (PUCRS). Inserção musical no final: Interpretação de Sarah Hester Ross de “The Day the Nazi Died” (Chumbawamba, 1993). Escute também no Spotify, no YouTube ou Apple Podcasts. Citados no episódio ANDERSON, Benedict. The Spectre of Comparisons: Nationalism, Southeast Asia and the World. London: Verso, 1998. FERREIRA, Dandara (dir.). Anatomia do Caos. Brasil, 2026. Documentário. MARQUES, Teresa Cristina Schneider. Migrantes internacionais na extrema direita e as frentes de transnacionalização do bolsonarismo nos EUA. Washington Brazil Office, 8 out. 2024. https://www.braziloffice.org/pt/artigos/migrantes-internacionais-na-extrema-direita-e-as-frentes-de-transnacionalizao-do-bolsonarismo-nos-eua MUDDE, Cas. The Far Right Today. Cambridge: Polity, 2019. Mencionado em entrevista. RAMOS, Paola. Defectors: The Rise of the Latino Far Right and What It Means for America. New York: Pantheon, 2024. ROLLEMBERG, Denise. Exílio: entre raízes e radares. Rio de Janeiro: Record, 1999. Mencionado em entrevista. SZNAJDER, Mario; RONIGER, Luis. The Politics of Exile in Latin America. Cambridge: Cambridge University Press, 2009. TARROW, Sidney. The New Transnational Activism. Cambridge: Cambridge University Press, 2005. TILLY, Charles; TARROW, Sidney. Contentious Politics. Boulder: Paradigm, 2007. Mencionado em entrevista. Capítulos: 00:00 — Introdução e apresentação do NETEX 03:40 — Do exílio na ditadura ao internacionalismo reacionário 27:40 — Por que a diáspora brasileira nos EUA vota em Bolsonaro? 36:55 — Flórida, Yes Brasil e as redes do bolsonarismo no exterior 46:50 — Eduardo Bolsonaro, Paulo Figueiredo e os limites do ativismo transnacional 53:50 — Futurologia: as redes transnacionais depois de 2026 01:03:30 — Boletim OED: AfD, Le Pen e Zemmour, Bardella e Farage, Bannon e Sanders 01:16:35 — Dica cultural: Anatomia do Caos The post O exílio como estratégia: as redes transnacionais do bolsonarismo appeared first on Chutando a Escada.
THE TONDU SPECTRESomething once walked the dark roads around Tondu…In this chilling episode of Time between Times, I tell the extraordinary story of the Tondu Spectre — a terrifying, death-like figure said to have appeared in 1904 and physically attacked a man on a lonely road.Was it a ghost, a hoax, or something far stranger?I'm also joined by writer and researcher Rachel Taylor, who has been researching the ghosts and haunted history of Bridgend. Together we delve into the stories, places and folklore that have given this corner of South Wales such a rich supernatural history.Turn down the lights…Come closer to the fire…And listen carefully.Because in the darkness around Tondu…something was waiting.Welcome to the Time between Times, my friends.www.welshstoryteller.comwww.ko-fi.com/owenstatonwww.patreon.com/owenstaton7Rachel Taylors Substack " Hauntings and History"https://thehauntedhistorian.substack.com/p/have-you-seen-the-ghosts-yet?r=56bicy&utm_campaign=post&utm_medium=webThanks to Rachel Taylor for the conversationTake care friendsNos daOwen x
James Bond is back, and this time he's facing his most personal enemy yet. In our review of SPECTRE (2015), we break down Daniel Craig's fourth outing as 007 as Bond goes head-to-head with the mysterious criminal organization SPECTRE and its leader, Ernst Stavro Blofeld.From the massive action sequences and iconic Bond moments to the performances from Daniel Craig, Christoph Waltz, Léa Seydoux, Dave Bautista, and Monica Bellucci, we're breaking down what worked, what didn't, and whether SPECTRE lives up to the high bar set by SKYFALL.Movie: SPECTRE (2015) Starring: Daniel Craig, Christoph Waltz, Léa Seydoux, Ben Whishaw, Naomie Harris, Dave Bautista & Ralph Fiennes Director: Sam MendesJoin the conversation: Was SPECTRE a worthy follow-up to SKYFALL?Spectre #JamesBond #DanielCraig #007 #MovieReview #SayWhatsReel #Skyfall #ChristophWaltz #bond00:00:00 Non 007 spectre related00:25:50 007 review00:56:10 wrap up notes00:59:40 facts about 007 spectre
Episode Highlights & Topics CoveredPart 1: True Crime — The Belanglo State Forest MurdersThe Victims & Location: Between 1989 and 1992, serial killer Ivan Milat targeted young international backpackers and Australian hikers along New South Wales' Hume Highway, luring them into the remote Belanglo State Forest.Forensic Signature: Orienteers and locals discovered buried remains of seven victims between 1992 and 1993. Autopsies revealed spinal knife wounds designed to paralyze victims, alongside gunshots from a .22 caliber Ruger rifle.The Survival Story: On January 25, 1990, British hitchhiker Paul Onions accepted a ride from Milat. When Milat pulled a gun and rope, Onions bolted into oncoming traffic. Driver Joanne Barry courageously stopped her car with her kids inside to let him dive in and escape.Justice Served: Years later, after seeing news of the forest discoveries, Onions contacted Australian authorities. His positive identification allowed police to execute search warrants on Milat's properties, where they recovered stolen camping gear and matching ballistics. Milat was convicted, inspired the horror film Wolf Creek, and died in prison in 2019.Part 2: Paranormal Ranches & High StrangenessBlind Frog Ranch (Utah): Known for massive glowing orbs that hover over canyons and plunge directly into and out of the ground[11].Stardust Ranch (Arizona): Site of intense cattle mutilations, liquid-metal UFO sightings, and owner John Edmonds' notorious claim of fending off Greys with a samurai sword.San Luis Valley (Colorado): A high-desert basin famous for silent helicopters, glowing orbs, shadow figures, and classic saucer/cigar-shaped craft.Dulce Base (New Mexico): The legendary multi-level subterranean facility beneath the mesa, associated with strange ground hums and multi-species military-alien encounters. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
We make another visit to that cave by the restless sea for this week's episode of The Horror. We'll hear The Spectre Of Tappington, the March 24, 1944, episode of The Weird Circle. Listen to more from The Weird Circle https://traffic.libsyn.com/forcedn/e55e1c7a-e213-4a20-8701-21862bdf1f8a/TheHorror1299.mp3 Download TheHorror1299 | Subscribe | Spotify | Support The Horror The Horror is made possible through the support of [...]
durée : 00:15:44 - Le 18/20 : un jour dans le monde - par : Fabienne Sintes - Alors que le monde subit l'envolée des prix des carburants, on fait le point sur les approvisionnements de la France. Peut-on réduire notre dépendance ? Avec Francis Perrin, directeur de recherche à l'Iris et spécialiste des problématiques énergétiques. - invités : Francis Perrin Chercheur, spécialiste des problématiques énergétiques Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France
I'm joined by one of the sharpest tactical voices in football, Jon Mackenzie, writer and presenter for The Athletic, and author of "The Spectre of Pep." I found this absolutely fascinating and I hope you do too.Support Latte Firm for the price of a coffee a month and enjoy ad-free shows, bonus content and access to giveaways and match tickets - patreon.com/lattefirm.
durée : 00:52:13 - Grand bien vous fasse ! - par : Ali Rebeihi - Entre 700 000 et 1 million de personnes sont atteintes du trouble du spectre autistique en France, dont au moins 100 000 jeunes de moins de 20 ans. Comment diagnostiquer des conditions autistiques ? Comment vivre avec ? Retrouvez témoignages et conseils à l'écoute de cette émission. - équipe : Joseph Hascal, Anna Massardier, Nathalie Poitevin, Noé Brabant, Jérôme Boulet, Ewen Dubée - invités : Maïtena Biraben Animatrice et productrice de télévision franco-suisse, Boris Chaumette Docteur en psychiatre et neurosciences, Adrien Primerano Sociologue Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France
. The Jagged Edge of History: Sergey Ryakovski ("The Hyena")A Nation Falling Apart: During the late 1980s, as the Soviet Union began to collapse, Moscow faced extreme instability, underfunded law enforcement, and widespread public panic[2][6].The Apex Predator Next Door: To his neighbors, Sergey Ryakovski was an ordinary construction worker, husband, and helpful neighbor—but beneath the surface lurked an indiscriminate predator[6][7].No Profile, No Pattern: Operating in total secrecy, he targeted men, women, and the elderly alike with no specific victim type[7].The Giggling Monster: Press dubbed him "The Hyena" due to his uncontrollable, high-pitched wheezing laughter during police questioning, viewing his horrific actions as a mission to "cleanse" and "audit" society[8][9].Can a serial killer operating in the chaos of a collapsing empire ever be brought to justice, or will the darkness swallow his secrets[2]*?*2. This Week in Spooky Coop NewsBigfoot Bounty: Hard Mountain Dew drops a $1 million bounty for verified proof of Bigfoot[11].Psychic Lawsuit: A widow sues a psychic over a $78,500 fee charged to lift an alleged inheritance curse[12].Vulcan World Record: Star Trek fans gather in London to set a Guinness World Record for simultaneous Vulcan salutes[13].Mass Ocean Scream: Over 100 people meet on a San Francisco beach for a collective guttural scream session[13].Emu Chase: Police and drones pursue an escaped emu through downtown New York[14].3. Glitches in the Matrix & Alternate TimelinesThe Flash of Light: The eerie report of Clara, who experienced a bright blue-green flash while driving—only to arrive where her workplace should be and find a shopping mall, along with a completely altered family history[15][16].The Man Who Fell Through Time: Paul and Michelle share a personal encounter on a mountain trail with Asamu Sodom, a man who claimed he accidentally crossed over into our reality[17].A World Without Atomic Bombs: Asamu described growing up in a timeline where Japan stayed out of World War II, Pearl Harbor never happened, and the atomic devastation of Hiroshima never occurred[4].Physical Proof: He carried worn, dog-eared black-and-white photographs and documents that appeared impossible to forge from his reported era[21][22].The Forest Encounter: What terrifying phenomenon took place when Asamu pursued his grieving friend into Japan's Aokigahara forest, only to be overcome by sudden dizziness[23][24]? Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
Listen Ad Free https://www.solgoodmedia.com - Listen to hundreds of audiobooks, thousands of short stories, and ambient sounds all ad free!
Spooky Coupe Episode 58: Samuel Little – The Most Prolific Serial Killer & Jodie's Ghost ClubEpisode SummaryIn Episode 58 of Spooky Coupe, hosts Paul and Michelle dive back into true crime to examine the chilling case of Samuel Little, officially recognized by the FBI as the most prolific serial killer in United States history. From his decades of eluding authorities across state lines to his detailed prison confessions and victim portraits, we unpack how he operated unnoticed for so long. Later in the episode, we share an unsettling listener story submitted by Jodie, chronicling her ghost hunting group's very first investigation of an abandoned house and the bizarre phenomena they encountered inside [1, 35–37].Show Notes & Highlights1. Podcast Catch-Up & The "Spooky Coupe" NameMichelle's Health Update: Michelle shares her relief after starting antibiotics for a long-standing sinus infection.Why "Spooky Coupe"? Paul and Michelle recap how their podcast got its name—inspired by their chickens making frantic noises around midnight, which turned out to be caused by a bear trying to break in.True Crime Return: The hosts discuss their love for the true crime genre and their decision to bring more crime cases to the coop.2. True Crime Case: Samuel LittleThe Body Count: Samuel Little claimed responsibility for 93 murders over four decades, with the FBI's Violent Criminal Apprehension Program (ViCAP) verifying at least 60.Early Life: Born June 7, 1940, in Reynolds, Georgia, and raised by his grandmother in Lorain, Ohio. Little showed aggressive behavior early on, dropped out of high school, and entered a juvenile institution in 1956 at age 16.How He Eluded Capture:Mobility & Indifference: As a transient traveling through Georgia, Florida, Mississippi, Ohio, and California, police departments lacked unified databases to link his crimes across state lines [10–12].Targeting Marginalized Victims: He deliberately targeted vulnerable women in deep poverty, sex work, or suffering from addiction.Modus Operandi: A former professional boxer who fought under the name "The Bosco," Little used his physical strength to knock victims unconscious and strangle them with his hands, leaving minimal forensic evidence [13–15].Legal Close Calls: Little previously beat murder charges in Mississippi (1982, Melinda LaRee) and Florida (Patricia Mount) due to lack of indictment or physical evidence.Capture & Confessions:Arrested in 2012 at a homeless shelter in Louisville, Kentucky; DNA linked him to three cold-case murders in Los Angeles (Carol Alford, Guadalupe Apodaca, and Ernestine Jones).Convicted in 2014 and sentenced to three consecutive life terms.In 2018, Texas Ranger James Holland conducted over 100 hours of interviews with Little, during which he confessed to dozens of additional killings and drew detailed memory portraits of victims to help identify them.Final Years: Little expressed zero remorse, claiming "God put me here to do this" and refusing last rites before dying in prison on December 30, 2020, at age 80 [25–27].3. Listener Paranormal Story: "Jodie's Ghost Club"Forming the Club: Jodie and her friends (Alexa, Louisa, and Maryanne) transitioned from watching Friday night TV ghost shows to conducting real-world investigations.The First Investigation: The group scoped out an abandoned house near town during the day to check for floor safety.Nighttime Encounters:Arrived to a heavy smell of wet rot and noticed broken kitchen cabinet doors were inexplicably back on their hinges.Experienced heavy door slams and creaking doors upstairs.Found two illuminated cat-toy balls—originally left on the floor—mysteriously hanging and flashing from a curtain rod high off the ground.The Aftermath: After capturing one distinct electronic voice phenomenon (EVP) recording, the group packed up and vowed never to return.Memorable Quote"Be good to each other, love one another, take care of each other, be safe." — Paul & Michelle Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
In or feature story a young girl falls in love with a demon and things get ugly. This is a very disturbing story. Read by Robert Crandall. Any reproduction of my voice for any purpose including AI is prohibited. All Rights Reserved. Order bonus episodes today. 8 episodes for only 9 dollars USD Go to http://www.adventuresinaudio.net click on donate and select your payment option. Listen for more details. Thank you for listening.
Episode #: 57Hosts: Paul, MichelleTheme: "The Devil Made Me Do It" — real cases where perpetrators claimed demonic possession as a defense or explanationSegment 1 — Arne Cheyenne Johnson (Connecticut, 1981) [01:30–09:00]Central figure: David Glatzel (11-year-old, allegedly possessed), Arne Johnson (David's sister's boyfriend), victim Alan Bono (Johnson's landlord)Ed and Lorraine Warren involved in the exorcism/caseJohnson reportedly said "come into me" during the exorcism, addressing the demonKilled Bono in a bar fight; claimed he was compelled and couldn't stop his handsJudge Robert Callahan threw out the possession defense as speculative/inadmissible, said CT law "was not designed for metaphysics"Defense fell back on insanity plea — failedConvicted; served reduced sentence, released earlySegment 2 — Jason Dalton (Kalamazoo, MI, Feb. 20, 2016) [09:00–15:00]Full-time insurance adjuster, driving for Uber to save for a Disney family vacation~2 weeks before the shootings, family/acquaintances noticed him talking/arguing with himself, sleep deprivationKilled 6, injured 2, during a spree interspersed with normal Uber ridesClaimed a "devil head" (horned, cow-like) appeared on the Uber app and took over his mind/bodyStuck to the possession story ~3 years until family funds for legal defense ran outPleaded guilty; sentenced to life without paroleSegment 3 — Nikolai Dzhumagaliev, a.k.a. "Metal Fang" (Soviet Union, starting 1979) [15:00–25:30]Nickname from metal dental replacements after losing teeth in a fightDiagnosed with STDs from military service; blamed/hated women, believed Satan intensified this hatredFirst killing in 1979; killed 4 more within 6 months (5 in 6 months total)Assaulted victims, dismembered and cannibalized themCaught for an unrelated bar murder; found insane, hospitalized, released after ~1 yearServed human meat to unaware friends at a cookout; killed another victim at that same partyEscaped custody by jumping from a moving transport vehicle; killed for ~2 years while at largeTurned himself in when he ran out of foodTotal confirmed victims: ~10; admitted to more without providing namesListener Story — "Elena" and "Julian" [26:00–37:30]Submitted after a previous episode about a fly infestation storyElena's 15-year-old son became friends with a new boy, "Julian" (names changed)Every visit from Julian triggered swarms of large black flies; flies vanished shortly after he leftElena's son became moody/withdrawn during the friendshipBreaking point: flies swarming the dinner table during a family dinner with Julian presentElena had her son transferred to a different school; friendship faded, flies stopped completelyHosts' takeaway: pay attention to gut reactions to people, not just physical/security concernsRecurring motif: Each case features a real, documented crime paired with a supernatural "possession" narrative offered by the perpetrator (or, in Elena's story, an ambient/parapsychological phenomenon tied to a specific person). Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
Become a supporter of this podcast: https://www.spreaker.com/podcast/harold-s-old-time-radio-no-1--4206392/support.
Become a supporter of this podcast: https://www.spreaker.com/podcast/harold-s-old-time-radio-no-1--4206392/support.
In this bonus episode for paid subscribers, I spoke with Nina Power about the power of communism as an ideology. Hosted on Acast. See acast.com/privacy for more information.
Spooky Coop Podcast — Episode 56: Where Do UFOs Come From?Are UFOs and UAPs visitors from another world—or could they be closer than we think?This week, Paul and Michelle explore the stranger theories behind the UFO phenomenon: hidden civilizations beneath the Earth, unidentified objects moving through the ocean, interdimensional beings, lunar and Martian bases, demons, angels, and the possibility that UFOs may reflect our own collective fears and beliefs.They also debut “This Week in Spooky Coop News,” with odd headlines from Hungary's gravedigging championship to a possible cryptid photo in a New York marsh, a shoe-tying stunt that caused public confusion, and more.Finally, they share Mandy's eerie account of a home overwhelmed by enormous, biting flies. After exterminators could find no source and conventional solutions failed, one house blessing allegedly brought the infestation to an immediate and unexplained end.Have you seen a UFO, witnessed something strange in the sky or water, or experienced a haunting of your own? Send your story to Spooky Coop. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
***Producer's note: TRN Podcast host Nick Estes joined Red Media-produced TYMSYY podcast, hosted by Sardana Nikolaeva and Masha Kardashevskaya, to talk about anti-communism, pretendianism, and the outsized role of Ward Churchill. Subscribe to TYMSYY if you don't already and help us grow the show*** This is the first installment of our collaborative series on Indigenous Anti-Communism with Nick Estes of The Red Nation Podcast. In this episode, we first introduce our intentions for the series and discuss why it matters to complicate and systematize the various discourses on anti-communism. We then turn to Ward Churchill (1947-2026), the American author and academic, and his edited volume "Marxism and Native Americans" (1983). Featuring Winona LaDuke, Vine Deloria Jr., Russell Means, and various Marxist writers, the volume became a foundational text for anyone interested in debates over Marxism and its relevance to Native American politics. However, as we learn in this episode, the volume's history and Ward Churchill's own legacy may not be as theoretically clear-cut as they seem. Nick Estes illuminates how the volume's anti-Marxist arguments distorted Indigenous understanding of Marxism and communism, which is reflected in the continued misunderstanding of the histories of anti-imperialist Indigenous solidarities. If you enjoyed the first episode of our series "Indigenous Anti-Communism" and don't want to miss upcoming discussions, please subscribe or follow us wherever you get your podcast. Watch the video edition on The Red Nation Podcast YouTube channel https://youtu.be/7qoDHlp5Mm0 Music: Anastasia Alexxeeva/Анастасия Алексеева - Djohogoj/Дьөһөгөй Produced by Red Media www.patreon.com/redmediapr
Spooky Coop Podcast — Episode 55: Unleashing the Hellhounds This week, Paul and Michelle venture into the eerie folklore of hellhounds: phantom black dogs said to appear with glowing red eyes, silent footsteps, and a warning of death. From Cerberus and Anubis to Britain's Black Shuck, the Cadejo of Central America, and the terrifying 1577 church encounters at Bungay and Blythburgh, they explore why stories of supernatural dogs have endured across cultures and centuries. Could these creatures be omens, guardians of the afterlife—or something far more real? Then, stay tuned for a listener-submitted encounter from Oregon's Crater Lake National Park. A couple on a remote trail spot what appears to be a lost child at the edge of the woods—until they realize the child may be something else entirely. Have you ever encountered a hellhound, black dog, Bigfoot, or another unexplained creature? Send us your story at sendittospookycoop@gmail.com. Find every episode and more spooky extras at spookycoop.com. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
This week on The Horror, we return to The Hall Of Fantasy for their tale from May 11, 1953, titled, The Spectre Of Denston Castle. Listen to more from The Hall Of Fantasy https://traffic.libsyn.com/forcedn/e55e1c7a-e213-4a20-8701-21862bdf1f8a/TheHorror1296.mp3 Download TheHorror1296 | Subscribe | Spotify | Support The Horror The Horror is made possible through the support of its listeners. If you'd like to help out, [...]
This episode we cover three different stories featuring ventriloquist dummies. First, the Spectre and Percival Popp use a dummy to capture "The Mole" from More Fun Comics 92. Then Roy Raymond tries to solve the mystery of a sentient dummy in Detective Comics 211. Finally, Abel tells the tale of a ventriloquist dummy out for revenge in House of Secrets 117. Don't be a dummy, listen to it now! Email us at theearth2podcast@gmail.com Facebook www.facebook.com/theearth2podcast Instagram www.instagram.com/theearth2podcast Twitter www.twitter.com/podcast_earth2 Leave us a Voicemail at www.speakpipe.com/theearth2podcast Find our Linktree at https://linktr.ee/theearth2podcast #dccomics #dcmultiverse #TheSpectre #PercivalPopp #Spectre #RoyRaymond #HouseofSecrets
Episode 54: Phantom Radio & Ghostly Phone Calls What happens when the airwaves get strange—and the caller may no longer be among the living? In this episode of Spooky Coop, Michelle and Paul investigate mysterious broadcasts that have unsettled listeners for decades. From the eerie 1977 Southern Television signal interruption and Chicago's legendary Max Headroom hijacking to Cold War numbers stations and Russia's ever-buzzing shortwave mystery, “The Buzzer,” they explore the real history behind broadcasts that sound like they came from somewhere—or someone—unexpected. Then, they turn to one of the most personal paranormal experiences people report: phone calls from the dead. Paul and Michelle share stories of static-filled calls, familiar voices, unexplained voicemails, Morse-code messages, and a grandmother's warning that may have prevented a tragedy. Are these technical glitches, grief reaching for meaning, or signs that our loved ones can still find a way to say, “I'm okay”? Expect spooky signals, strange codes, bad impressions, tender stories, and at least one joke about collect calls from the afterlife. In This Episode The “Vrillon” television interruption in Southern England The Max Headroom broadcast hijacking in Chicago Numbers stations and their Cold War mystery Russia's shortwave station UVB-76, also known as “The Buzzer” The haunting Lincolnshire Poacher broadcasts Reported calls, voicemails, and messages from departed loved ones Michelle's personal story of a possible warning from beyond A Note from the Coop The stories in this episode are shared as reported experiences, folklore, and unexplained events. We invite you to listen with curiosity, compassion, and a healthy respect for your phone's spam filter. If you enjoy the show, please subscribe, share the episode with a fellow spooky friend, and send us your own strange stories. Until next time: be safe, be kind, and keep one ear on the airwaves. See you in the Spooky Coop. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
The Anfield Wrap talks to tactics expert Jon Mackenzie about his new book, 'The Spectre of Pep'. An analysis into influence of Pep Guardiola on football at all levels. Neil Atkinson hosts... Learn more about your ad choices. Visit podcastchoices.com/adchoices
Shobana Shankar discusses her book, An Uneasy Embrace: Africa, India and the Spectre of Race. Prof. Shankar explains how Africans and Indians make and unmake their differences. While decolonization brought Africans and Indians together to challenge Euro-American style white supremacy, discord over caste, religion, sex and skin color simmered beneath the rhetoric of Afro-Asian solidarity. […]
Spooky Coop Episode 53: Rocks of Thornton Road & Spirits of Hawaii This week, Paul and Michelle investigate two very different kinds of paranormal mystery: one loud, chaotic, and possibly armed with an invisible rock collection; the other quiet, luminous, and full of forest magic. First, they travel back to 1980s Birmingham, England, and the baffling reports from Thornton Road. Residents described windows shattering, roof tiles breaking, and stones appearing to strike homes from impossible angles. Police considered a human explanation involving a catapult, but the alleged activity—and the lack of a culprit—left many neighbors wondering whether something far stranger was behind the flying rocks. Then, the Coop heads to Kauaʻi, Hawaii, for a listener's personal and uplifting story of encounters with the Menehune tradition and small, fairy-like beings in the forest. From mysterious woodland visitors and healing herbs to blessings for children and one unforgettable question about whether winged beings hatch from eggs, this is a tale of wonder, kindness, and the occasional supernatural giggle. In this episode: The Thornton Road “rock-throwing poltergeist” and its years of unexplained activity Why a secret mega-catapult may not be the most satisfying explanation Paul's theory about a ghost named Frederick with excellent aim A listener's personal account of the Menehune and fairy-like beings in Kauaʻi Forest magic, medicinal herbs, mysterious “doorways,” and tiny packets of salt Why it's always wise to be kind, stay curious, and maybe keep your windows reinforced Listener note: These stories are presented as reported experiences, folklore, and paranormal accounts—not as proven facts. We share them in the spirit of curiosity, respect, and a healthy appreciation for the unexplained. Have a strange story of your own? Send it our way—and until next time, be safe, be kind to each other, and we'll see you in the Spooky Coop. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
In the latest edition of Caught Offside, Andrew Gundling and JJ Devaney are joined by legendary Premier League fullback, Ashley Cole, talking Arsenal's dominance, the upcoming Premier League season as whole, his interest in coaching and much much more!Plus, we'll share our thoughts on a truly devastating loss for Celtic against LASK and their cast of American heroes. We'll also take a look at some of the wild financial figures being thrown around in the transfer window and try to make sense of it and we'll close things out with another edition of "This Week in Dutch Bluntness."For even more Caught Offside content, get on over to Caught Offside Plus right now! It's time for a Pep-talk! We've got a special interview for you with John Mackenzie to discuss his soon-to-be-released book about Pep Guardiola, "The Spectre of Pep: How Guardiola Haunts Modern Football Tactics."For all the latest merch, get over to https://caughtoffsidepod.com/ ---Reddit: https://www.reddit.com/r/CaughtOffsidePod/X: https://twitter.com/COsoccerpodInstagram: https://www.instagram.com/caughtoffsidepod/Email: CaughtOffsidePod@gmail.com Hosted on Acast. See acast.com/privacy for more information.
Paul and Michelle investigate two fresh cryptid mysteries making waves in 2026. First, they head to Highway 292 near Spartanburg, South Carolina, where two late-night drivers reported seeing a huge, dark, hair-covered figure move upright into the woods. Was it Bigfoot, a black bear, or someone's especially unusual uncle after a cookout? Then it's off to the misty Scottish Highlands and Loch Ness, where new reports and videos of strange dark shapes, sudden wakes, and unexplained movement have once again stirred up Nessie fever. The hosts explore why the Loch Ness Monster still captures our imagination—and whether these sightings offer more than just a mystery. In this episode: A reported South Carolina Sasquatch encounter near Lake Cooley Why nighttime sightings are so difficult to confirm Fresh Loch Ness reports and online footage from 2026 The enduring appeal of Nessie, Bigfoot, and the unexplained How to safely document a strange sighting Listener question: Are cryptids becoming bolder, or are we simply paying closer attention? What would convince you that what you saw was definitely not a bear? Send your strange stories and theories to: sendittospookycoop@gmail.com More Loch Ness sighting reports: lochnesssightings.com Keep your flashlight charged, your mind open, and one eye on the water. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
In the latest edition of Caught Offside, Andrew Gundling and JJ Devaney are looking back on the good and bad of the English Premier League's opening weekend. We'll discuss Manchester United's "passionless" defeat at Hull City as well as Tottenham's deja vu performance against Brentford. We'll also weigh in on Brighton's dismantling of Aston Villa and Arsenal looking very much like the defending champions that they are.And finally, we'll look at some of the situations that have caused fans to turn against their own players - Enzo Fernandez at Chelsea, Ollie Watkins at Villa and Julian Alvarez at Atletico Madrid.For even more Caught Offside content, get on over to Caught Offside Plus right now! Later this week, it's time for a Pep-talk! We'll have a special interview for you with John Mackenzie to discuss his soon-to-be-released book about Pep Guardiola, "The Spectre of Pep: How Guardiola Haunts Modern Football Tactics."For all the latest merch, get over to https://caughtoffsidepod.com/ ---Reddit: https://www.reddit.com/r/CaughtOffsidePod/X: https://twitter.com/COsoccerpodInstagram: https://www.instagram.com/caughtoffsidepod/Email: CaughtOffsidePod@gmail.com Hosted on Acast. See acast.com/privacy for more information.
Stamford Chidge interviews Jon McKenzie about his new book: 'The Spectre of Pep - How Guardiola Haunts Modern Football Tactics'.A natural successor to the brilliant 'Inverting the Pyramid' by Jonathan Wilson, Jon's new book put Guardiola in to a historical context before exploring his impact on football over the last 20 years and poses the question of what his legacy will be.Jon is a writer and presenter on The Athletic FC and Tifo and is a regular on the FanCast's Opppositon View.The book can be purchased here: https://geni.us/TheSpectreOfPep Hosted on Acast. See acast.com/privacy for more information.
Paranormal Stories: Mimics, Swamps, and Unready Harvests Welcome back to the Spooky Coop! Grab your chocolate, adjust your microphones, and get ready to climb into the coop with your hosts, Paul and Michelle [1, 8]. This week, we are going "for broke" with a massive bumper episode featuring six terrifying, listener-submitted stories of cryptids, alien abductions, and bone-chilling mirror entities [1, 33, 45]. We've kept it PG (mostly!) and cleaned up some colorful language, but the scares in these stories are completely unedited [6, 7]. ⏱️ Episode Timestamps & Story Segments Story 1: The Appalachian Mimic Story 2: The Honey Island Swamp Monster Story 3: The Skyscraper Crawler Story 4: The Orchard Abduction Story 5: The Rockies Harvest Story 6: The Reflection in the Rain Bonus Segment: The Vanishing Trucker]
Paul and Michelle explore a collection of eerie cases of Spontaneous Human Combustion and listener-submitted encounters of personal hauntings: The mystery of Spontaneous Human Combustion, including the cases of Mary Reeser and Michael Faherty. A middle-school teacher's escalating experiences with moving objects, unexplained messages, and a presence that may be attached to him rather than his home. An architect who discovers a room that should not exist behind an attic wall. Are these paranormal encounters, failures of perception, or something stranger? Join Spooky Coop as Paul and Michelle follow the stories—and the shadows. Episode Credits:Written, edited and produced by Paul & MichelleArtwork created by MIchelle & PaulMusic written and produced by I.C.D.LINKS: Listen to us here: SpookyCoop , find us @ spookycoop.com and follow us on: FaceBook | BlueSky | X (formerly Twitter) | Threads | Instagram | TikTok
What happens to football when the manager who shaped the modern game finally leaves? Jon MacKenzie, author of The Spectre of Pep: How Guardiola Haunts Modern Football, joins the Brazilian Shirt Name Podcast alongside Seb White to examine Pep Guardiola's extraordinary influence on the modern game and what happens after him. From playing out from the back and controlling space to high pressing, positional play and the growing tactical obsession with control, Guardiola's ideas have spread from the very top of the Premier League all the way down the football pyramid. But did that dominance eventually become a weakness? The conversation explores why Guardiola's later Manchester City sides became more risk-averse, whether his system restricted players such as Jack Grealish and Phil Foden, the criticism that Pep was simply a “chequebook manager”, and why football may now be moving away from the tactical consensus he created. Plus, what could Guardiola do next? International football? Another club? Or has one of football's great obsessives finally discovered that there is more to life than the game? And, naturally, there's some Michel Foucault, Martin Heidegger and Jacques Derrida along the way. Our good friends at Stanchion Books are putting on a second event with Jon. The first one sold out, some of us will be there so you should get your tickets before they sell out. You can pre-order the book from them as well.” https://www.eventbrite.co.uk/e/the-spectre-of-pep-guardiola-book-launch-with-jon-mackenzie-tickets-1997680941472? Join the Brazilian Shirt Name Whatsapp Channel: https://whatsapp.com/channel/0029VbBNgO58PgsAgQXRP32T
After the extraordinarily low lows of last week's Star Trek episodes, this week's batch feels like a welcome respite. Up to bat this week is previously-seen guest star Diana Muldaur in the psychedelic ugly alien freakout Is There in Truth No Beauty, followed by the good ol' fashioned wild west romp Spectre of the Gun!CHAPTERS:(00:00:00) - The Nextlander Watchcast Episode 187: Star Trek: Is There in Truth No Beauty and Spectre of the Gun(00:00:39) - Intro.(00:01:58) - Jumping right into Is There in Truth No Beauty, an impossible title to say correctly the first time.(00:15:26) - Don't look at the ugly alien. You'll go crazy.(00:17:57) - Pretty lady onboard, let's get weird about it.(00:29:13) - Goddammit, Larry.(00:37:49) - Show the pretty lady flowers while we step out with her ambassador.(00:44:11) - Piloting Spock around like it's the ambassador's vacation body.(00:47:57) - Spock goes crazy, and it barely matters.(00:51:41) - Spock's good, Dr. Jones got her mindlink, and Larry's dead, so I guess this is a happy ending.(00:56:37) - Break!(00:57:04) - We're back, and it's time to talk about Spectre of the Gun (and some OK Corral history)!(01:14:33) - A neon probe with a big booming snake head says to stay outta this space.(01:18:47) - A half-remembered idea of Tombstone, Arizona, where everybody is pointing you toward doom.(01:25:08) - A force field prevents escape, gas grenades are constructed, and Chekov "dies".(01:32:06) - Spock sees through the Matrix.(01:39:06) - Huh, you defeated our hastily assembled wild west trap. I guess you guys are cool.(01:44:21) - Final thoughts, and housekeeping for next week's episodes.(01:47:03) - Outro.
Four bands under the knife this week
The Spectre of Pep: How Guardiola Haunts Modern Football Tactics (Seven Dials, 2026) by Jon Mackenzie A spectre is haunting European football. Watch almost any game, no matter the level, and you will notice it. From inverted full backs to extreme high lines, the tactical threads of the match will eventually converge onto a singular point: Pep Guardiola. For some though, his near total dominance has come at the cost of the game itself, with modern football's protagonists caught between acknowledging Guardiola's greatness and wanting to evolve the sport beyond him. In The Spectre of Pep, acclaimed tactics writer Jon Mackenzie explores this contradiction. But this is much more than the story of just one man. After tracing Guardiola's ascendancy, displacing giants like Jose Mourinho and Antonio Conte, it examines how other elite coaches have fought to establish their own legacies within this landscape - including the likes of Jurgen Klopp, Roberto De Zerbi, Luis Enrique and Mikel Arteta. And with his time at Manchester City coming to an end - in a world where his influence is arguably diminishing - it also speculates on what the next generation of tactics may look like, and how Guardiola's legacy will be viewed in the future. If you want to understand football today, you have to understand the tactical spectre of Pep Guardiola. To do that, you need this book. Jon Mackenzie is a writer and presenter working for The Athletic. Samee Siddiqui is Assistant Professor of World History at Drury University. Learn more about your ad choices. Visit megaphone.fm/adchoices Support our show by becoming a premium member! https://newbooksnetwork.supportingcast.fm/new-books-network
The Spectre of Pep: How Guardiola Haunts Modern Football Tactics (Seven Dials, 2026) by Jon Mackenzie A spectre is haunting European football. Watch almost any game, no matter the level, and you will notice it. From inverted full backs to extreme high lines, the tactical threads of the match will eventually converge onto a singular point: Pep Guardiola. For some though, his near total dominance has come at the cost of the game itself, with modern football's protagonists caught between acknowledging Guardiola's greatness and wanting to evolve the sport beyond him. In The Spectre of Pep, acclaimed tactics writer Jon Mackenzie explores this contradiction. But this is much more than the story of just one man. After tracing Guardiola's ascendancy, displacing giants like Jose Mourinho and Antonio Conte, it examines how other elite coaches have fought to establish their own legacies within this landscape - including the likes of Jurgen Klopp, Roberto De Zerbi, Luis Enrique and Mikel Arteta. And with his time at Manchester City coming to an end - in a world where his influence is arguably diminishing - it also speculates on what the next generation of tactics may look like, and how Guardiola's legacy will be viewed in the future. If you want to understand football today, you have to understand the tactical spectre of Pep Guardiola. To do that, you need this book. Jon Mackenzie is a writer and presenter working for The Athletic. Samee Siddiqui is Assistant Professor of World History at Drury University. Learn more about your ad choices. Visit megaphone.fm/adchoices Support our show by becoming a premium member! https://newbooksnetwork.supportingcast.fm/sports
The Spectre of Pep: How Guardiola Haunts Modern Football Tactics (Seven Dials, 2026) by Jon Mackenzie A spectre is haunting European football. Watch almost any game, no matter the level, and you will notice it. From inverted full backs to extreme high lines, the tactical threads of the match will eventually converge onto a singular point: Pep Guardiola. For some though, his near total dominance has come at the cost of the game itself, with modern football's protagonists caught between acknowledging Guardiola's greatness and wanting to evolve the sport beyond him. In The Spectre of Pep, acclaimed tactics writer Jon Mackenzie explores this contradiction. But this is much more than the story of just one man. After tracing Guardiola's ascendancy, displacing giants like Jose Mourinho and Antonio Conte, it examines how other elite coaches have fought to establish their own legacies within this landscape - including the likes of Jurgen Klopp, Roberto De Zerbi, Luis Enrique and Mikel Arteta. And with his time at Manchester City coming to an end - in a world where his influence is arguably diminishing - it also speculates on what the next generation of tactics may look like, and how Guardiola's legacy will be viewed in the future. If you want to understand football today, you have to understand the tactical spectre of Pep Guardiola. To do that, you need this book. Jon Mackenzie is a writer and presenter working for The Athletic. Samee Siddiqui is Assistant Professor of World History at Drury University. Learn more about your ad choices. Visit megaphone.fm/adchoices Support our show by becoming a premium member! https://newbooksnetwork.supportingcast.fm/european-studies
Night Gallery Season 3 Episode 4 Spectre in Tap-Shoes
The Blue Moon Podcast is back for YET another season... On today's show, David Mooney is joined by City fan Ciaran Murray and by the Daily Mail's Jack Gaughan to look back over the pre-season tour of Asia for Manchester City and to discuss what's new about the Enzo Maresca era at the club. The Athletic's Sam Lee has been speaking to the players on the tour, so he's here to look at what they've said about what's different tactically. Plus, we speak to Tifo's Jon Mackenzie, whose new book "The Spectre of Pep: How Guardiola Haunts Modern Football Tactics", looks at how much City's former manager has influenced how we all see the game. We take a look ahead to the Community Shield tie with Arsenal on Sunday, plus (and we make no apologies for this) we're starting off as weirdly as we mean to go on by taking a look at all the times that City have played on the same day as a partial or total solar eclipse. ========== To get more podcasts or to listen without the ads, join our Patreon. It's just £2 per month for all the extra content and you can get a 7-day free trial first: https://www.patreon.com/BlueMoonPodcast And why not gift a Patreon subscription to a friend or family member? More details: https://www.patreon.com/BlueMoonPodcast/gift Learn more about your ad choices. Visit podcastchoices.com/adchoices
We started this set of broadcasts in trouble, and it's only getting worse! Why is it important to have conversations about subjects that make us uncomfortable? Tune in to The Public Square® today to hear more. Topic: Politics The Public Square® with host Dave Zanotti thepublicsquare.com Air Date: Friday, August 7, 2026
durée : 00:02:30 - Le 6/9 de l'été - L'inquiétude monte en Europe car les stocks de gaz sont à un niveau exceptionnellement faible. Comment l'expliquer ? - équipe : Romain Gueugneau Vous aimez ce podcast ? Pour écouter tous les épisodes sans limite, rendez-vous sur Radio France
Jace and Rocky break down all 10 DC releases for the week of August 5, 2026. In Absolute Superman #22, Superman risks being poisoned by Metallo's kryptonite blood because he knows reaching Christopher Smith emotionally is the only way to end the fight. In Batman #12, Verity is living up to the Pennyworth name as mention of her training and her quick thinking impresses Bruce and Damian while five highly trained killers each demonstrate their methods in a series of murders. Could their missions be tied to Gotham's newest threat? In Absolute Green Lantern #17, Jo confronts Mu as the issue adds more fascinating cosmic concepts, but seventeen issues of undefined terminology leave even the central conflict difficult to understand. But the issue does deliver the origin of Absolute Jessica Cruz and a great scene of Tomar-Re being fired like a bullet. In Adventures of Superman: Book of El #11, even though Superman has been reunited with Osul and Otho, the battle is not over. Kal-El is determined to stop Kryl-Ux even while trying to save him from himself and the grief of losing his family that has twisted his soul. In Deadman #3, Deadman travels through Hell to rescue a soul Rama Kushna says does not belong there, only for him to be shocked when Lucifer shows him proof the soul is where it belongs. The devil's got receipts. In Batgirl #22, Cassandra learns that Victoria was both her childhood friend and rival assassin. Both girls were manipulated by David Cain and Dr. Forget-Me-Not, respectively, in an experiment to build the perfect killer. The resolution for this one feels a little abrupt, but the possibility that Victoria may still be alive hints at more to come for this storyline in the future. In JSA #22, Jim Corrigan refuses to reclaim the Spectre, forcing the Spirit of Vengeance to seek a surprising new host from within the JSA. Kid Eternity fulfills her previously unknown mission to find that host, but turns down the chance to recover her memories and move on to the afterlife so she can remain with the team and continue to make a difference. In Supergirl: Survived #3, Kara's love for Kal drives her to attempt the impossible, using a Blue Lantern ring to recreate Krypton for both Kal-El and Hal Jordan. It's an emotionally powerful moment, but the issue takes an unexpected turn when an Elseworlds version of Mister Mxyzptlk arrives, shifting the story away from the hard sci-fi direction it had been building toward. In Gotham Academy: First Year #6, the final issue reveals more about Olive Silverlock, but the predetermined need to position every character for the beginning of Gotham Academy Vol. 1 leaves the story feeling paint-by-numbers. The rushed maneuvering from the end of issue #4 makes the conclusion feel both overstuffed and choppy, while surprisingly little of consequence actually happens. In 100 Bullets: The U.S. of Anger #2, the issue provides so little usable context that neither a new reader nor someone familiar with the original series can clearly explain who the characters are or what is happening. The series does seem to be appropriately named, though, as the anger the issue is steeped in feels all too real. While the story hasn't come together yet, the simmering anger of the U.S. reflected here feels authentic. They also give a rundown of this week's collected editions, reprints and facsimile releases. As always, all books are ranked from top to bottom and each host gives a Book of the Week pick. 00:00 — DC Week Overview, SDCC Aftermath & Comic Promotion 17:49 — Absolute Superman #22 35:27 — Deadman #3 43:03 — Absolute Green Lantern #17 57:44 — Supergirl: Survived #3 1:04:19 — Batman #12 1:13:51 — Batgirl #22 1:18:18 — Gotham Academy: First Year #6 1:21:19 — Adventures of Superman: Book of El #11 1:29:23 — 100 Bullets: The U.S. of Anger #2 1:37:20 — JSA #22 1:45:42 — Collected Editions, Reprints & Facsimiles 1:53:17 — Rankings & Books of the Week
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
Legendary screenwriter, producer, and director John Logan (MICHAEL, SWEENEY TODD, GLADIATOR, THE AVIATOR, HUGO, ANY GIVEN SUNDAY, SKYFALL, SPECTRE, PENNY DREADFUL, RANGO, ALIEN: COVENANT, RED, THE LAST SAMURAI, STAR TREK: NEMESIS, THEY/THEM) returns to continue discussing his incredible career with Adam and Joe. Tired of commercials? Support THE MOVIE CRYPT for just $1 a month and start getting every episode commercial-free! Visit www.Patreon.com/TheMovieCrypt to sign-up today!
© 2020-2026 Ivy Podcast Discovery LLC
