Podcasts about Speculation

Shannon's 7:30 Dirty 8-25-2026 See omnystudio.com/listener for privacy information.

Closing arguments in the trial for Lindsay Clancy, the Massachusetts mom accused of killing her three children, could begin as soon as Monday as the prosecution continues calling rebuttal witnesses. Caroline Polisi has more.GOP Sen. Ted Cruz joins "CBS Mornings" to discuss his new book, "Going Further: The Incomparable Clarence Thomas," and weigh in on the U.S.-Canada trade feud and the Senate race in Texas between Ken Paxton and James Talarico.Prince Harry and Meghan Markle's surprising announcement has fans speculating if Meghan Markle may be eyeing a return to acting. Roya Nikkhah reports.Dr. Sasha Hamdani, a board-certified psychiatrist and ADHD clinical specialist, joins "CBS Mornings" to discuss the relationship between ADHD and Rejection Sensitive Dysphoria, also known as RSD.Levi Zhang and his parents share some recipes from their new cookbook and talk about how their viral videos started.A former Ohio principal lost his job after allowing a homeless student to remain in school after the district unenrolled him. He and his wife eventually became the student's legal guardians. CBS News contributor David Begnaud shares the story.

Speculation about SkyCity's future has seen the company's share price end the day up 9.9 percent to $0.67. SkyCity management have talked about a strategic review for its Adelaide, South Australia casino business and is progressing the sale of its Grand Hotel asset, which would reduce SkyCity's debt levels. Harbour Asset Management expert Shane Solly explained further. LISTEN ABOVESee omnystudio.com/listener for privacy information.

MAFS 20 Eps 16-18 Recap: A Perfect Match Podcast Watch the podcast on YouTube! A Perfect Match dives into Married at First Sight Season 20, episodes 16-18, as hosts Aysha and Jason Reed guide listeners through high-stakes retreats, shifting alliances, and relationship turbulence on the eve of Decision Day. With just a week left before the final choices, the Seattle couples face truth-telling games, awkward revelations, and conflicts that test every bond, while production changes add even more uncertainty to their journeys. This podcast episode unpacks the fallout from the group retreat, where couples like Tori and Felipe, Mecca and Belle Jolie, and Devin and Kaitlyn encounter major turning points. The hosts break down the impact of messy group games that force participants to confront how others see them, including Felipe's discomfort at winning every “bad husband” award and Mecca's struggle to take responsibility for repeatedly hurting Belle Jolie. The discussion also highlights Kaitlyn and Devin's ongoing communication issues, revealing how missed signals and lack of affection put their relationship in crisis. Meanwhile, the hosts consider the consequences of new format twists: experts no longer hosting the reunion and couples facing each other alone for Decision Day decisions. Key moments and insights include: The retreat's “most likely” group game exposing unresolved issues, especially with Felipe and Tori's partnership Sean's controversial comment to Nikki, rating her a Seattle 10 but a New York 7, and the wider theme of men “humbling” their wives The unraveling of Devin and Kaitlyn's marriage, culminating in zero affection and a dramatic late-night argument over boundaries and flirting Belle Jolie's confrontation with Mecca about public comments on her appearance, and her struggle to balance self-respect with loyalty Speculation and reaction to the show's format changes, including Decision Day's new structure and the missing experts Will any of these couples find enough growth and connection to say “yes,” or are the red flags just too many to overcome? How will the absence of the usual support systems on Decision Day shape their final choices? Hear all the relationship drama, raw conversations, and strategic missteps by listening to this deep-dive breakdown of MAFS Season 20's most pivotal episodes. Chapters:00:00 Perfect Match Premise Revealed06:50 Kevin Frazier Replaced at Reunion10:08 Decision Day Format Changes15:05 Couples Retreat Tensions Ignite23:13 Devin Reveals No Kisses Yet27:41 Sean Calls Nikki a "Seattle 10"33:15 Michelle and Cameron Flaunt Strength41:13 Mecca Fails to Apologize Sincerely47:27 Felipe's Retreat Behavior Challenged54:22 Retreat Game Causes Couples' Rift01:03:03 Tori and Felipe's Major Blowup01:13:12 Devin and Kaitlyn Argument Erupts01:25:44 Belle Calls Out Sean's Insecurity01:34:32 Mecca Picks Other Wives01:42:00 Kaitlyn Flirts With Mecca01:44:38 Devin and Kaitlyn Break Up01:52:27 Felipe Justifies Reckless Words02:18:11 Tori Faces Felipe's Bombshell02:23:16 Finale Preview and Predictions Previously on the Love at First Sight Feed:Love at First Sight Recap Archives LISTEN: Subscribe to the Perfect Match RHAPUp podcast feed by visiting https://robhasawebsite.com/feed/mafsWATCH: Watch and subscribe to the podcast on YouTubeSUPPORT: Become a RHAP Patron for bonus content, access to Facebook and Discord groups plus more great perks! Learn more about your ad choices. Visit megaphone.fm/adchoices

Melania 'Doesn't Like' Speculation That Donald Trump and Natalie Harp Have an Affair, White House Insiders Say Amid Fierce DenialsAdvertising Inquiries: https://redcircle.com/brandsPrivacy & Opt-Out: https://redcircle.com/privacy

On "49ers Talk," Matt Maiocco and Jennifer Lee Chan discuss how the 49ers were challenged on both sides of the ball in their joint practice with the Los Angeles Chargers on Tuesday in El Segundo. As Justin Herbert found holes in San Francisco's secondary, Brock Purdy and the 49ers' offense seemed disjointed. On a positive note, coach Kyle Shanahan eased contract concerns about Christian McCaffrey, and rookie Kaelon Black returned to the backfield. Later, Jennifer sits down with cornerback Ephesians Prysock in the sixth edition of Meet the Rookies. -- (1:00) 49ers will move to a new practice facility soon (2:00) Substation will remain conversation topic (6:30) 49ers' secondary couldn't stop Justin Herbert despite improvements (11:00) Kyle Shanahan quiets CMC speculation (14:00) Why 49ers' rash of injuries aren't as bad as they seem (19:00) Jim Harbaugh, 49ers organization share mutual admiration (31:00) 49ers, Chargers bring professional atmosphere to joint practice (35:00) Meet the Rookies: Ephesians Prysock Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

The guys cover the Texas Rangers' recent win and the status of Jacob deGrom's return to the mound. Speculation arises regarding Luka Doncic's future in Los Angeles and the potential for a 2028 reunion with the Mavericks. They also examine the growth of high school football in Texas and critique the latest NFL Top 100 player rankings.

Our guest on the podcast today is Jeff Ptak. He's a longtime Morningstar employee and currently serves as managing director for Morningstar Research Services. He originally joined Morningstar back in 2002 as a senior mutual fund analyst. Jeff was one of the original two co-hosts of this podcast and regularly posts his thoughts on Morningstar.com, Substack, X, and LinkedIn. Ptak's work really focuses on investor outcomes. One of the highlights from the podcast today was when Ptak talked about how such a small number of stocks tend to generate a disproportionate amount of the market returns and what that means for diversified actively managed funds. He also discusses what target-date funds get right for investors, as well as what some of them may be missing. Episode Highlights 00:00:00 Mind the Gap and Investor Behavior 00:08:04 Crypto ETFs and Timing Mistakes 00:11:33 Active Funds and Letting Winners Run 00:23:00 Improving Investor Outcomes Through Lower Costs 00:30:09 Private Markets, SpaceX, and 401(k) Risks 00:34:15 Thematic ETFs, Speculation, and Fun Money 00:38:40 Market Timing Myths and Portfolio Construction 00:42:16 Is Tech's Dominance Sustainable? More From Morningstar Read: Mind the Gap 2026 Leyla Kunimoto: Why Investors in Private Markets Need a Louder Voice Will Danoff: ‘Be Very Careful of Unprofitable Companies' Don Phillips: Encouraging Better Outcomes for Investors   If you have a comment or a guest idea, please email us at TheLongView@Morningstar.com.   Follow Christine Benz (@christine_benz) and Ben Johnson (@MstarBenJohnson) on X, and Christine Benz, Amy Arnott, and Ben Johnson on LinkedIn. Visit Morningstar.com for new research and insights from Christine, Ben, and Amy. Subscribe to Christine's weekly newsletter, Improving Your Finances. If you want more Morningstar podcasts, check out The Morning Filter and Investing Insights. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

In this episode, the hosts discuss various topics including their appreciation for Patreon supporters, reflections on their delivery experiences, and the emotional connections with pets. They delve into rumors surrounding a potential new buyout from UPS, evaluating the implications of such offers and the future of the company in relation to the Teamsters union. The conversation wraps up with speculations about the future and personal reflections on work-life balance. www.patreon.com/aitdpod https://discord.gg/hm8WMUKVF8  takeaways The hosts express gratitude towards their Patreon supporters. Emotional connections with pets can impact delivery drivers. Rumors of a new buyout from UPS are circulating. Evaluating buyout offers is complex and personal. The hosts discuss the implications of buyouts on job security. The future of UPS and the Teamsters union is uncertain. Personal circumstances influence decisions about buyouts. The hosts emphasize the importance of enjoying life outside of work. Speculation about the future of UPS raises concerns. Listeners are encouraged to stay informed about potential changes. Chapters 00:00 Introduction and Shoutouts 02:04 Reflections on Deliveries and Pets 06:15 Rumors of a New Buyout 11:11 Evaluating Buyout Offers 16:04 The Future of UPS and Teamsters 21:02 Speculations and Closing Thoughts Huge shoutout to our TOP RATE LEGENDS Tony, Starla & S_nner THE OPINIONS EXPRESSED OR VIEWS EXPRESSED ON THIS PODCAST ARE THOSE OF THE HOSTS AND GUESTS AND DO NOT NECESSARILY REFLECT ANY DELIVERY COMPANY    

The preseason AP Top 25 is out and BYU checks in at No. 14, tied with USC with 839 points. KSL Sports BYU Insider Mitch Harper reacts to the Cougars' highest preseason ranking since the 1985 season, the year after BYU's national championship. It's only the 13th time in program history BYU has appeared in the preseason AP poll and just the second time in the Kalani Sitake era. Four Big 12 teams made the cut — Texas Tech at No. 12, BYU at No. 14, Utah at No. 21, and Houston at No. 23 — and BYU's October 17 home date with No. 4 Notre Dame is already circled in red. Historically, BYU has finished higher than its preseason AP ranking only twice (1981, 2009). Can the Cougars make it a third? Mitch discusses what this ranking means for expectations, the College Football Playoff picture, and how voters are viewing BYU's returning production, headlined by Big 12 Offensive Player of the Year LJ Martin and quarterback Bear Bachmeier. The ongoing Kalani Sitake–BYU football resources and athletic department remains front and center for pockets of Cougar Nation. The speculation that dominated the past two weeks shows no signs of fading, even as the season draws closer. Since Sitake told reporters at picture day, "I'm easy to get along with as long as you take care of our football team," the conversation has only intensified. Sitake's earlier comment that BYU has been "given the challenge of doing the most with the least" added fuel. The hiring of Kentucky's Marc Hill as deputy AD and Santiago's need to post on CougarBoard to debunk firing rumors have kept this story in the backdrop of fall camp. KSL Sports BYU Insider Mitch Harper shares his thoughts and insight into the situation and makes it clear that BYU has to take care of Kalani Sitake. Finally, Mitch recaps an eventful scrimmage as BYU enters week three of fall camp. The Cougars ran 102 plays at LaVell Edwards Stadium on Saturday, and the offense had the upper hand. Bear Bachmeier was in a "great rhythm," according to Sitake, who gave the offense a "slight edge" over the defense.  Subscribe to the Cougar Tracks Podcast to stay up-to-date with all the daily episodes. Cougar Tracks is on YouTube and X every weekday at Noon (MT), and KSL NewsRadio at 6:30 p.m. (MT). Apple: https://podcasts.apple.com/us/podcast/cougar-tracks/id1146971609 YouTube Podcast: https://kslsports.com/category/podcast_results/?sid=2035&n=Cougar%20Tracks Spotify: https://open.spotify.com/show/2NCF1KecDsE2rB1zMuHhUh Download the KSL Sports app Google: https://play.google.com/store/apps/details?id=com.bonneville.kslsports&hl=en_US  iOS: https://apps.apple.com/us/app/ksl-sports/id143593 Mitch Harper is a BYU Insider for KSLsports.com and hosts the Cougar Tracks Podcast daily on KSL Sports YouTube and KSL NewsRadio (SUBSCRIBE). Harper also co-hosts Cougar Sports Saturday (12–3 p.m.) on KSL NewsRadio. Follow Mitch’s coverage of BYU athletics in the Big 12 Conference on X (formerly Twitter) and Instagram: @Mitch_Harper. Want more coverage of BYU sports? Take us with you wherever you go. Download the new and improved KSL Sports app from Utah’s sports leader. Allows you to stream live radio and video, keeping you up-to-date on all your favorite teams.

BB28 Friday Live Feed Update August 14, 2026 Also available on YouTube. Big Brother 28 delivers another unpredictable day as the house wrestles with the fallout from a blindside vote almost happening and a new HoH throwing alliances into chaos. Taran Armstrong brings on Pooya for a live feed update episode, dissecting the latest power shifts, unraveling deals gone wrong, and spotlighting a growing split between the house's men and women. The episode details how Lala's actions as outgoing HoH fueled accusations of bullying and set the stage for drama. The main storyline follows a near vote flip to save Haley, orchestrated by Dee, which nearly put the Icons under major fire for the second week in a row. The actual eviction sees Chuk exit, and the knockout HoH comp ends with Yash winning, marking worst-case scenario for the core vet alliance. As alliances scramble, Yash lays out nomination plans with Kamu and Drew, targeting Haley, Angela, and Dee, motivated by both strategic concerns and a deliberate intent to keep all the male players off the block. Meanwhile, discussions unfold about whether Devens might use the diamond Power of Veto to shake up the nominations, with uncertainty swirling around whether Dee or Angela could actually be in danger. Taran highlights that Dee risks her own game by pushing another blindside vote flip, using up more social capital only a week after a similar move. Possible realignment as Kamu, Yash, and Drew eye a “Bromuda Triangle” alliance (apart from the “Bermuda Triangle”, wary that Devens and Barrett might be too close to the vets. The knockout HoH, which Yash wins, is analyzed for strategic missteps as houseguests fail to target threats, allowing Yash and Barrett to cruise to the end. Haley's social isolation is spotlighted, with house talk centering on whether her methods or her relationships with women may be pushing her into the outsider spot. Speculation grows about whether Dee or Devens will pull out a power to avoid the block, with the possibility of bribes and veto twists in play. With the men sticking together, and Yash promising to keep all of them off the block, the Icons lose their grip on power. So if Haley, Angela, and Dee do see the block, will the vet alliance's luck finally run out, or can their secret powers save them? Tune in for the full Big Brother 28 update and see how the week's wild moves unfold. 00:00 Big Brother 28 LFU Begins 06:00 HOH Lockdown Sparks Bullying Claims 12:00 Vote Flip Plan Backfires 18:00 Blindside Fallout and New HOH 25:00 Knockout Competition Seals Yash's Reign 32:00 Yash Plots Big Brother Nominations 38:00 All-Male Alliance Emerges in House 44:00 Angela, Dee, Haley Face Block 50:00 Diamond Power Veto Uncertainty 56:00 Chaos Looms Before Nominations 01:02:00 Live Feeds Tease More Turmoil Never miss a minute of RHAP's extensive Big Brother coverage! LISTEN: Subscribe to the Big Brother podcast feed WATCH:  Watch and subscribe to the podcast on YouTube SUPPORT:  Become a RHAP Patron for bonus content, access to Facebook and Discord groups plus more great perks!

BB28 Friday Live Feed Update August 14, 2026 Also available on YouTube. Big Brother 28 delivers another unpredictable day as the house wrestles with the fallout from a blindside vote almost happening and a new HoH throwing alliances into chaos. Taran Armstrong brings on Pooya for a live feed update episode, dissecting the latest power shifts, unraveling deals gone wrong, and spotlighting a growing split between the house's men and women. The episode details how Lala's actions as outgoing HoH fueled accusations of bullying and set the stage for drama. The main storyline follows a near vote flip to save Haley, orchestrated by Dee, which nearly put the Icons under major fire for the second week in a row. The actual eviction sees Chuk exit, and the knockout HoH comp ends with Yash winning, marking worst-case scenario for the core vet alliance. As alliances scramble, Yash lays out nomination plans with Kamu and Drew, targeting Haley, Angela, and Dee, motivated by both strategic concerns and a deliberate intent to keep all the male players off the block. Meanwhile, discussions unfold about whether Devens might use the diamond Power of Veto to shake up the nominations, with uncertainty swirling around whether Dee or Angela could actually be in danger. Taran highlights that Dee risks her own game by pushing another blindside vote flip, using up more social capital only a week after a similar move. Possible realignment as Kamu, Yash, and Drew eye a “Bromuda Triangle” alliance (apart from the “Bermuda Triangle”, wary that Devens and Barrett might be too close to the vets. The knockout HoH, which Yash wins, is analyzed for strategic missteps as houseguests fail to target threats, allowing Yash and Barrett to cruise to the end. Haley's social isolation is spotlighted, with house talk centering on whether her methods or her relationships with women may be pushing her into the outsider spot. Speculation grows about whether Dee or Devens will pull out a power to avoid the block, with the possibility of bribes and veto twists in play. With the men sticking together, and Yash promising to keep all of them off the block, the Icons lose their grip on power. So if Haley, Angela, and Dee do see the block, will the vet alliance's luck finally run out, or can their secret powers save them? Tune in for the full Big Brother 28 update and see how the week's wild moves unfold. 00:00 Big Brother 28 LFU Begins 06:00 HOH Lockdown Sparks Bullying Claims 12:00 Vote Flip Plan Backfires 18:00 Blindside Fallout and New HOH 25:00 Knockout Competition Seals Yash's Reign 32:00 Yash Plots Big Brother Nominations 38:00 All-Male Alliance Emerges in House 44:00 Angela, Dee, Haley Face Block 50:00 Diamond Power Veto Uncertainty 56:00 Chaos Looms Before Nominations 01:02:00 Live Feeds Tease More Turmoil Never miss a minute of RHAP's extensive Big Brother coverage! LISTEN: Subscribe to the Big Brother podcast feed WATCH:  Watch and subscribe to the podcast on YouTube SUPPORT:  Become a RHAP Patron for bonus content, access to Facebook and Discord groups plus more great perks!

Crypto adoption is accelerating, but does that mean token prices follow? This week, we unpack growing institutional interest in crypto as brokerages and fintechs push onchain, even as CLARITY stalls and memecoin activity dominates. We explore SEC and CFTC catalysts, Fomo versus Pump.fun, the durability of meme-driven markets, token value accrual, and crypto's financialization problem. Enjoy! TIMESTAMPS: 00:00 Intro 02:17 Has Crypto Found A Bottom? 08:22 Can Regulation Replace Clarity? 14:42 Fomo Vs Pump Fun 22:07 How Founders Survive The Reckoning 26:31 Is Crypto Worth The Venture Bet? 34:39 Is Erebor Worth $8 Billion? 41:40 Is Collector Crypt Just Gambling? 48:03 Everything Is Becoming Financialized 54:23 Why Santi Is Long AI 57:26 What Benchmark Gets Right FOLLOW THE SHOW › Empire – https://x.com/theempirepod › Jason – https://x.com/jasonyanowitz › Santi – https://x.com/santiagoroel › Rob – https://x.com/HadickM › Telegram – https://t.me/+CaCYvTOB4Eg1OWJh › Blockworks – https://x.com/Blockworks EVENTS › Join us at Digital Asset Summit 2026 Asia October 7th & Digital Asset 2026 London November 10-11th https://blockworks.com/events DISCLAIMER Nothing said. onEmpire is a recommendation to buy or sell securities or tokens. This podcast is for informational purposes only. Any views expressed are opinions, not financial advice. Hosts and guests may hold positions in the companies, funds, or projects discussed.

Grace opens the show with the news that Karoline Leavitt will be leaving her position as Press Secretary, which has sparked speculation over who will replace her.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

On today's podcast conversation, Chris Cillizza and Matt discuss:— Massive polling misses in Michigan and Wisconsin. One reason this happened?: Under-sampling middle-aged Black voters and rural Democrats— Clear evidence that DSA and DSA-aligned candidates are underperforming with Black voters — the same coalition that blocked Bernie Sanders— Why 2026 is likely a referendum on Trump and the economy, but 2028 could see the “socialism” label stick and drive voters back to Republicans— Trump's recent endorsement flops (Darlene Graham runoff, Mike Lindell, Tennessee) are signs his iron grip on the GOP is loosening (or are they???)— Jeff Flake's arc from pure conservative to endorsing a Democrat Jon Ossoff, and the death of the anti-Trump lane in GOP primaries— Hunter Biden's Tucker Carlson interview, fathers-and-sons dynamics, and the complicated human side of political families— Trump's secret catering-truck plane switch amid Iran threat concerns: security necessity vs. bad optics and press deception— Speculation on the next White House press secretary (Kari Lake, Spencer Pratt, or Scott Jennings?)— And MUCH more!Subscribe to Matt Lewis on Substack: https://mattklewis.substack.com/Support Matt Lewis at Patreon: https://www.patreon.com/mattlewisFacebook: https://www.facebook.com/MattLewisDCTwitter: https://twitter.com/mattklewisInstagram: https://www.instagram.com/mattlewisreels/YouTube: https://www.youtube.com/channel/UCVhSMpjOzydlnxm5TDcYn0A– Who is Matt Lewis? –Matt K. Lewis is a political commentator and the author of Filthy Rich Politicians.Buy Matt's books: FILTHY RICH POLITICIANS: https://www.amazon.com/Filthy-Rich-Politicians-Creatures-Ruling-Class/dp/1546004416TOO DUMB TO FAIL: https://www.amazon.com/Too-Dumb-Fail-Revolution-Conservative/dp/0316383937Copyright © 2026, BBL & BWL, LLC

Spielberg's film Disclosure Day takes on the narrative of the recently released government UFO files. Speculation claims that what Spielberg aimed for was to shake up the Christian faith by his documentation of possibility of aliens from other planets. These extraterrestrials woud prove to Christians that biblical Christianity is false. But that's not what Spielberg necessarily said. We examine the premise of the movie as well as what the disclosed government files tell us. Comparing the two we see where facts end and speculative fiction begins.

Michael Howell, CEO of CrossBorder Capital, an investment advisory firm, and author of Capital Wars, returns to explain why the global liquidity cycle peaked in late Q3/early Q4 of last year — and what that means for the rest of 2026. His core argument: money is fungible but finite, and a booming real economy is now pulling liquidity out of financial assets, which compresses P/E multiples even as earnings look fine. That puts us in what he calls the speculation phase: rising bond yields, strong commodities, pressured crypto, and low-quality equity returns where index gains mask widespread underperformance. He also pushes back hard on the popular "debasement trade" explanation for gold, arguing the real driver is the People's Bank of China injecting liquidity to devalue the yuan internally while holding it steady externally — with Chinese retail locked out of crypto and the Shanghai Gold Exchange now setting the marginal price. On the bond side, he lays out how the Treasury is quietly monetizing through front-end issuance and buybacks — private-sector QE under Treasury direction — a strategy that works until it doesn't, with Japan's move from 50bps to nearly 3% as the cautionary tale. His bottom line: range-bound Wall Street, no bonds, gold and silver on weakness, and watch commodities for the first sign the boom is ending.Thank you to our partners Augusta Precious Metals — To learn more, visit https://juliabuysgold.com/ or text “Julia" to 35052Monetary Metals - learn more at https://www.monetary-metals.com/julia/Links:  Website: http://www.crossbordercapital.com/ Twitter/X https://x.com/crossbordercapSubstack: https://capitalwars.substack.com/ Book: https://www.amazon.com/Capital-Wars-Rise-Global-Liquidity/dp/30303929020:00 The call: range-bound market, own gold0:20 Welcome back, Michael Howell1:19 Two pools of money: markets vs. the real economy2:30 The liquidity cycle has peaked3:20 What this phase looks like4:48 Why a booming economy is bad for stocks5:22 The P/E multiple is where liquidity shows up6:34 Late cycle, explained7:38 Augusta Precious Metals9:29 Global liquidity vs. the world business cycle10:45 Atlanta Fed nowcast near 6%11:54 The K-shaped economy is global12:45 Monetary inflation vs. Main Street inflation14:45 Speculation now, turbulence next15:15 The cycle map17:55 Monetary Metals19:49 Gold: it isn't the debasement trade20:30 It's China: PBOC liquidity22:15 Why gold and not crypto23:14 Inside the PBOC balance sheet25:00 Yuan gold and the 27,000 line26:15 Bond yields track nominal GDP27:40 NGDP at 7-8% vs. a 4.7% ten-year28:18 Treasury QE: funding at the front end30:20 Who's actually buying the debt?30:51 The beach ball under water32:35 The two-year note leads the Fed34:30 The 2022 analogue36:00 Why MOVE matters more than VIX37:08 Treasury buybacks and the volatility cap38:30 Margin debt and the 2026 range call39:31 Parting thoughts: commodities as the warning40:30 Gold, silver, and the ratio to watch

Christopher Luxon has called for an 'urgent in-person meeting' on Wednesday at 9:30am. Any of his MPs who aren't in Wellington will need to travel back given it's a recess week at parliament. Citing media reports and conversations about increased speculation about his leadership, Luxon called it an 'unfair distraction'. Former National Party MP Maurice Williamson spoke to Melissa Chan-Green

With speculation over his leadership swirling, National's Christopher Luxon is calling all National MPs to the capital for an urgent in-person caucus meeting tomorrow. He's just come off a fortnight defending his comments to Rotorua small businesses, his Foreign Minister telling a Chinese-born MP to go back where he came from, and announcing on live radio he wants a referendum on MMP - without having told his caucus. Political reporter Russell Palmer has the story.

Marjorie Taylor Greene Shuts Down Tucker Carlson 2028 Presidential Speculation: 'He's Not Running'Advertising Inquiries: https://redcircle.com/brandsPrivacy & Opt-Out: https://redcircle.com/privacy

A senior National MP's throwing cold water on the idea the Prime Minister will be voted out of his job as National party leader today. Mark Mitchell says there's been a lot of speculation from media commentators about what will happen, but there's no confirmation. He told Mike Hosking everyone's getting ahead of themselves. Mitchell says he sat with Luxon until late last night to make sure he's in the "strongest possible position today and supported for the caucus meeting". He said he'd seen media reports claiming Chris Penk had been hitting the phones to shore up support for Erica Stanford, but stressed there would be no discussions until the 9.30am meeting. Affirming his loyalty to Luxon as leader, Mitchell described himself as "very old-fashioned" and said he took "great offence" to people who work outside that operandus. "If you've got someone that's not been loyal to the leader inside the caucus, you can't do it publicly. You actually front up and sit down and say, 'These are what the issues are'," he said. LISTEN ABOVE See omnystudio.com/listener for privacy information.

The Deputy Prime Minister won't make a prediction on the outcome of today's National leadership challenge, but says a Prime Minister leading a coalition that's ahead in the polls should be supported. Christopher Luxon called an emergency caucus meeting for this morning in response to the growing speculation over his ability to lead the party into the election. David Seymour told Mike Hosking he's not going to sit there and bag Luxon, who is doing the best he can, given difficult circumstances. He says the National party leader has foibles just like everyone else, but he's managed to unite his own party and the Coalition. Seymour said the leadership challenge was “not helpful” and was a distraction from policy – specifically replacing the RMA, which he claimed was "the only thing that really matters" for New Zealand's future. LISTEN ABOVE See omnystudio.com/listener for privacy information.

A former National MP believes Christopher Luxon's decision to call today's emergency caucus meeting was a mistake. The Prime Minister announced the meeting on social media yesterday, saying the speculation about his leadership needs addressing. With Parliament in recess, MPs have had to fly in from around the country. Mapp told Mike Hosking Luxon could have dismissed the speculation as the work of a few dissident MPs. He says Luxon's constructed the stage for his own execution. LISTEN ABOVE See omnystudio.com/listener for privacy information.

**Clay Edwards Show with Creston Burch – Episode Description** Big updates on the Nolan Wells case dominate this episode. Sources indicate the Wells family and Ben Crump have not turned over Nolan's cell phone to the Jackson County DA's office despite requests, and they have declined or ignored an invitation to sit down and review the state crime lab autopsy and toxicology reports. The toxicology is said to show high alcohol levels, no cocaine, and likely THC. The autopsy is expected to list drowning or undetermined as the cause, with the manner of death remaining undetermined — a finding that does not equal homicide. Discussion covers the condition of the body after roughly 36 hours at sea and the ongoing narrative push around deleted messages and foul play. Also covered: neighbors in Pascagoula have sued Judge Ashley Cole and Dr. Hudson over a permitted fence that allegedly blocks their harbor view, adding another layer of pressure on the family amid the broader case. Locally, the death of Taseya Fortune, found hanging behind an abandoned house on Road of Remembrance in Jackson, draws scrutiny of lynching claims. A person of interest living with her had already been arrested on an unrelated charge, with reports of a prior quarrel. Speculation about white supremacist involvement is examined against the emerging facts. Additional segments touch on a widely circulated but false claim that a Yemeni immigration attorney was appointed municipal judge in Ridgeland (denied by both the individual and the city), plus ongoing WNBA drama involving Sophie Cunningham and former NBA players seeking to enter the women's league. Straightforward discussion of the facts as they stand, the incentives driving public narratives, and why some stories refuse to die quietly.

#ScrubHopTalk Ep. 296 - The boys get deep into their speculation bag regarding the Your Mom's House Podcast and we find out some interesting information about our down below. We see a video of a lady trying to enjoy a pastry on the waterfront only to have her time ruined by a seagull, and we tumble down a rabbit hole about waterfronts. JDirty shows off a misprinted copy of "The Amazing Jeckel Brothers", but that too leads us down a pretty disturbing path about his past and some unknown stories. @troxy_cotton @scrubhopking @bigtrox303 #ScrubHop #imfuckingColumbo #GiveMeAWaterfront #itwasnteveninmybackyardScrub Hop Talk is a weekly show with JDirty, Big Trox, and Troxy Cotton. The boys bring you their take on life and pop culture, reacting to crazy videos, and showcasing a different song from their catalog every week. Brand new episodes air here at YouTube.com/ScrubHop every Sunday night at 5pm Pacific time.Please comment, like, and subscribe!For more information, visit ScrubHop.com to learn all about the music and join the movement.Big Trox's hat selection this week is brought to you by Static-X.Visit Howard's 3D Prints for all your 3D printing needs!https://www.instagram.com/howards3dprintsThis week's song:JDirty - "Until Next Time" https://open.spotify.com/track/2KPXjEpK5mwaliImpTxmOI?si=ee98e5a1e9924ca4Buy the merch at:http://ScrubHopShop.bigcartel.comFollow the socials at:@ScrubHop on EVERYTHING!JDirty:http://scrubhop.com/jdirtyhttp://instagram.com/scrubhopkinghttp://twitter.com/jdirty303http://facebook.com/JDirty303Big Trox:http://scrubhop.com/bigtroxhttp://instagram.com/bigtrox303Troxy Cotton:http://scrubhop.com/troxycottonhttp://instagram.com/troxy_cottonhttp://twitter.com/TroxyCottonhttp://facebook.com/TroxyCottonCO

The best value add plan is not always the one you start right away. Chris Lento explains why waiting for the market can produce better long term results. Description Chris Lento, Managing Partner of EM Capital, joins Dave Dubeau to discuss how multifamily investing has changed over the past few years. Instead of relying on rapid rent growth or market appreciation, Chris explains why today's environment rewards strong operations, careful expense control, and disciplined decision making. The conversation covers how EM Capital manages nearly 1,200 apartment units across the Southeast, how they motivate onsite teams with owner funded incentives, why investor communication matters during difficult markets, and how they are exploring AI to improve capital raising, acquisitions, and operations. Key Topics Building systems for better asset management Waiting for market support before renovating units Managing investor expectations during longer hold periods Finding off market opportunities through broker relationships Using AI to improve operations and decision making Guest Information Chris Lento Managing Partner, EM Capital Website: www.emcapitalgroup.com LinkedIn: Available through the EM Capital website. Call to Action Learn more about Chris Lento and EM Capital by visiting www.emcapitalgroup.com. Connect with Chris on LinkedIn through the website.

Tucker Carlson unveils 10-point manifesto amid Presidential bid speculationAdvertising Inquiries: https://redcircle.com/brandsPrivacy & Opt-Out: https://redcircle.com/privacy

Hello and welcome back to the Oasis Podcast. Today's episode is with Richard Bowes @rbmusicwriterSupport the show patreon.com/oasispodEmail oasispod@gmail.comStay young!

Another week of strong market gains served as a reminder that long-term fundamentals can quickly regain center stage once short-term volatility begins to fade. Markets rebounded sharply this week as investors looked beyond July's volatility and refocused on corporate fundamentals. The Dow Jones Industrial Average gained 3.0%, the S&P 500 rose 3.6%, and the Nasdaq climbed 5.2%. Year to date, the Dow is now up 12.4%, the S&P 500 has gained 13.3%, and the Nasdaq leads with a 14.8% return. The Money Wise guys discuss how July's deleveraging appears to have largely run its course, allowing buyers to return as earnings continue to exceed expectations. They also review the latest employment data, upcoming inflation reports, and why markets remain sensitive to Federal Reserve policy while continuing to emphasize the importance of separating short-term headlines from long-term fundamentals. The remainder of the program focuses on investor education, beginning with the importance of teaching younger investors how to build wealth through consistent saving rather than speculation. The guys stress that long-term investing is built on patience, dollar-cost averaging, and diversification - not leveraged products or "get rich quick" strategies. The second hour then turns to an extended discussion of equity-indexed annuities, explaining how these products are structured, why investors should understand participation rates, interest-rate caps, surrender periods, commissions, and liquidity restrictions before purchasing them, and why it's important to carefully evaluate marketing claims before making long-term financial decisions. Throughout the discussion, the recurring message remains the same: successful investing comes from discipline, education, and understanding exactly what you own. Avoiding Speculation One of the most valuable investing lessons doesn't involve finding the next hot stock or timing the market perfectly—it's understanding the power of consistency. Whether you're just beginning your career or helping the next generation start investing, regularly contributing to retirement accounts, taking advantage of employer matching contributions, and allowing compounding to work over time can have a tremendous impact. While speculative products and leveraged investments often receive the most attention, long-term wealth has historically been built through disciplined saving, broad diversification, and patience. For most investors, the tortoise still beats the hare.  In the second hour, the Money Wise guys delve further into their discussion on Equity Index Annuities. You don't want to miss the details! Tune in for the full discussion on your favorite podcast provider or at davidsoncap.com, where you can also learn more about the Money Wise guys or take advantage of a portfolio review and analysis with Davidson Capital Management.

In this episode of Scene N Nerd, Sarah and Will dive deep into the much-anticipated Spider-Man: Brand New Day. They share their thoughts on Tom Holland's evolution as Spider-Man, reflecting on how we've witnessed his journey from a teenager to a young adult grappling with the implications of his identity after the events of No Way Home. The duo discusses Zendaya's significant character development in this film, emphasizing her expanded role compared to previous installments. They also touch on the dynamic between Tom Holland and Jon Bernthal, noting the delightful comedic chemistry that adds a new layer to Frank Castle's character. Listeners can expect an insightful exploration of the film's themes of grief and loss, as Peter Parker navigates his relationships and the weight of his responsibilities. They analyze the film's approach to the Rogues Gallery and how it balances intimate character moments with larger comic book action, culminating in a well-structured narrative that resonates on both personal and heroic levels. The episode also hints at future developments in the Spider-Verse, speculating on potential sequels and character arcs, particularly for Jean Grey and her connection to Peter. Timestamps: 0:00 — Introduction to Spider-Man: Brand New Day 1:30 — Tom Holland's growth as Spider-Man 5:00 — Zendaya's expanded role 10:00 — Chemistry between Tom Holland and Jon Bernthal 15:00 — Themes of grief and loss 20:00 — The Rogues Gallery and character dynamics 25:00 — Speculations on future Spider-Verse films 30:00 — Closing thoughts and wrap-up Follow Sarah on social media at @sjbelmont and Will at @willmpolk. Follow Scene N Nerd on X, Blue Sky, Facebook, Instagram, and Threads, and visit SceneNNerdPodcast.com. Rate, follow, and leave a comment on Apple Podcasts, Spotify, YouTube, or wherever you listen to podcasts.

Sarah Fields joins Harmony with a major update on the Nolan Wells case. Friends of Wells are preparing to sue Ben Crump and online voices who accused them in his death. Fields unpacks the claims, the planned legal push, and why this fight is moving from social media into the courtroom. A clear-eyed conversation on truth, reputation, and accountability when viral accusations outrun the facts. ➢ Show Outline 00:00 Introduction: Ben Crump's "Marketing Circus" 03:11 Sarah Fields on Combating Misinformation 04:43 Nolan Wells Case Compared to Karmelo Anthony Case 08:14 Timeline of July 4th Events 14:20 Threats Against the Boys and Families 20:11 Discussion on the Nature of Threats 25:57 Gibson Go Fundraiser for the Boys 30:47 Ben Crump Calls Defamation Case a "Distraction" 33:02 Speculation on Nolan's Relationship with Parents 36:43 Major Update: Edward Andrew Paul Joins Defense 39:11 Edward Andrew Paul's Background and Intent 41:12 Impact of Ben Crump on Community and Friends ➢ Follow Our GUESTS https://www.youtube.com/@TheShemekaMichelle  ➢ Subscribe to Jason's other channel https://www.youtube.com/JasonWhitlock?sub_confirmation=1 https://www.youtube.com/@JasonWhitlockHarmony?sub_confirmation=1 https://www.youtube.com/@JasonWhitlockBYOG?sub_confirmation=1 https://www.youtube.com/@JasonWhitlockClips?sub_confirmation=1 ➢ Connect with Jason on Social Media:  https://x.com/JasonWhitlock  https://www.instagram.com/realjasonwhitlock/ https://www.facebook.com/jasonwhitlock ➢ Send Jason an Email FearlessBlazeShow@gmail.com ➢ Support The Blaze Visit https://TheBlaze.com. Explore the all-new ad-free experience and see for yourself how we're standing up against suppression and prioritizing independent journalism. Support Conservative Voices! Subscribe to BlazeTV at https://www.fearlessmission.com and get $20 off your yearly subscription. Learn more about your ad choices. Visit megaphone.fm/adchoices

En 1637, la Hollande découvre la première bulle spéculative de l'histoire : la tulipomanie. Dans l'euphorie du Siècle d'or, certains bulbes de tulipe s'échangent plus cher qu'une maison… avant que les prix ne s'effondrent du jour au lendemain.Plongez dans l'histoire des grands personnages et des évènements marquants qui ont façonné notre monde ! Avec enthousiasme et talent, Franck Ferrand vous révèle les coulisses de l'histoire avec un grand H, entre mystères, secrets et épisodes méconnus : un cadeau pour les amoureux du passé, de la préhistoire à l'histoire contemporaine.Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.

Rodriguez Vazquez, et al. v. Bostock, et al., No. 25-6842 (9th Cir. July 30, 2026)class action; no mandatory detention for EWIs; Hurtado; seeking admission; INA § 235(a)(2)(B); Laken Riley Act; applicant for admission entry; canon against superfluity  Cirrus Rojas v. Olson, et al., No. 25-3127 (7th Cir. July 30, 2026)mootness; deemed; no mandatory detention for EWIs; Hurtado; seeking admission; INA § 235(a)(2)(B); applicant for admission entry; canon of constitutional avoidance; plain text Matter of A-L-S-, 29 I&N Dec. 794 (BIA 2026)bond; flight risk; speculative relief; clear error review; long time residence and employment in U.S. as negative factors for bond  Matter of L-L-R-, 29 I&N Dec. 799 (BIA 2026)good moral character; possibly alien smuggling even if dropped off at the border; Al Otro Lado; INA § 101(f)(3);  Matter of A-W-M-K-, 29 I&N Dec. 805 (BIA 2026)bond; flight risk; USCIS decision as evidence; speculative relief; suspected human right violations; Afghanistan  United States v. Lopez, No. 24-3268 (9th Cir. July 28, 2026) Cal. Pen. Code § 273.5; crime of violence; Borden; Gomez; mens rea and use of force; recklessness; general intent crimes; assault; battery; use of “willful” in statute not determinative Perez-Hernandez v. Blanche, No. 25-3592 (6th Cir.  July 28, 2026)motion to suppress; egregious constitutional violate; race-based stop; Miranda warnings; intra familial dispute in Guatemala; nexus; relocation for CAT Kim v. Blanche, No. 24-2042 (1st Cir. July 30, 2026)sua sponte motion to reopen; conviction vacatur; citation to vacatur statute alone sufficient to establish procedural or substantive defect; Super. Ct. R. Crim P. 11Kurzban Kurzban Tetzeli and Pratt P.A.Immigration, serious injury, and business lawyers serving clients in Florida, California, and all over the world for over 40 years.eimmigration"Immigration law software you'll love to use."get.eimmigration.com/IRP Gonzales & Gonzales Immigration BondsP: (833) 409-9200immigrationbond.com Stafi"Remote staffing solutions for businesses of all sizes"Click me!Want to become a patron?Show the Podcast some loooovvveeeCONTACT INFORMATION:Email: kgregg@kktplaw.comFacebook: @immigrationreviewInstagram: @immigrationreviewTwitter: @immreviewAbout your hostCase notesRecent criminal-immigration article (p.18)Featured in San Diego VoyagerSupport the show

BB28 Sunday Nominations Ep Recap, Week 4 Also available on YouTube. Big Brother 28 sees the house split as alliances battle for control and trust becomes more fragile than ever. Rob Cesternino and Taran Armstrong recap all the action and drama from a messy HOH reign, joined by Sasha Joseph for insight and fresh takes. This week, Haley wins HOH and immediately dives into brewing tensions, making snacks and personal feuds a focal point of house chatter. Speculation over secret powers and the BB Time Capsule sets fans into a frenzy, especially as Drew’s odd behavior and coin talk throws viewers off the scent. The episode covers how Dee secretly secures and considers her $5,000 bribe power, why she keeps it from Devens, and how power uncertainty leads to house paranoia. Meanwhile, petty rivalries over snacks, comments, and past votes create plenty of fireworks, with Melody and Taylor in the crosshairs and Drew's fate as “pawn” or target hotly discussed. The social game is just as wild, with conversations and diary rooms full of alliance maneuvering, lies, and mean-girl accusations. Sasha highlights how online speculation over Drew and the “coin” power sent fans on a wild goose chase, only for it to fizzle out. Taran explains why Dee hides the true nature of her bribe power, given its limited influence. Rob and Sasha laugh about how houseguests use snacks and petty squabbles to justify nominations and targets. Taran Armstrong points out how edit visibility might predict the next BB Time Capsule recipient and how Melody lands herself at the center of attention due to her unfiltered reactions. Alliances remain shaky and every move is scrutinized, as power players size up who's next to win a game-changing power or become the next easy target. Did Haley play this HOH right, or did she just put a bigger target on her own back? Listen in for all the strategy, snark, and analysis on Big Brother 28. 00:00 Big Brother 28 Weekend Drama 06:16 Drew's Coin Power Conspiracies 10:44 Dee's Secret $5K Bribe Power 16:36 BB Time Capsule Favors Returning Icons 21:27 Haley's HOH Reign Brings Tension 26:23 Feuds, Snacks, and Nominations 32:15 Melody vs. Haley: Mean Girl Clash 40:55 Lyric's Position After Veto Win 46:22 Replacement Nominee and House Rifts 52:13 Drew's Fate and House Dynamics 58:47 Under-the-Radar Players and Next Moves Never miss a minute of RHAP's extensive Big Brother coverage! LISTEN: Subscribe to the Big Brother podcast feed WATCH:  Watch and subscribe to the podcast on YouTube SUPPORT:  Become a RHAP Patron for bonus content, access to Facebook and Discord groups plus more great perks!

BB28 Sunday Nominations Ep Recap, Week 4 Also available on YouTube. Big Brother 28 sees the house split as alliances battle for control and trust becomes more fragile than ever. Rob Cesternino and Taran Armstrong recap all the action and drama from a messy HOH reign, joined by Sasha Joseph for insight and fresh takes. This week, Haley wins HOH and immediately dives into brewing tensions, making snacks and personal feuds a focal point of house chatter. Speculation over secret powers and the BB Time Capsule sets fans into a frenzy, especially as Drew’s odd behavior and coin talk throws viewers off the scent. The episode covers how Dee secretly secures and considers her $5,000 bribe power, why she keeps it from Devens, and how power uncertainty leads to house paranoia. Meanwhile, petty rivalries over snacks, comments, and past votes create plenty of fireworks, with Melody and Taylor in the crosshairs and Drew's fate as “pawn” or target hotly discussed. The social game is just as wild, with conversations and diary rooms full of alliance maneuvering, lies, and mean-girl accusations. Sasha highlights how online speculation over Drew and the “coin” power sent fans on a wild goose chase, only for it to fizzle out. Taran explains why Dee hides the true nature of her bribe power, given its limited influence. Rob and Sasha laugh about how houseguests use snacks and petty squabbles to justify nominations and targets. Taran Armstrong points out how edit visibility might predict the next BB Time Capsule recipient and how Melody lands herself at the center of attention due to her unfiltered reactions. Alliances remain shaky and every move is scrutinized, as power players size up who's next to win a game-changing power or become the next easy target. Did Haley play this HOH right, or did she just put a bigger target on her own back? Listen in for all the strategy, snark, and analysis on Big Brother 28. 00:00 Big Brother 28 Weekend Drama 06:16 Drew's Coin Power Conspiracies 10:44 Dee's Secret $5K Bribe Power 16:36 BB Time Capsule Favors Returning Icons 21:27 Haley's HOH Reign Brings Tension 26:23 Feuds, Snacks, and Nominations 32:15 Melody vs. Haley: Mean Girl Clash 40:55 Lyric's Position After Veto Win 46:22 Replacement Nominee and House Rifts 52:13 Drew's Fate and House Dynamics 58:47 Under-the-Radar Players and Next Moves Never miss a minute of RHAP's extensive Big Brother coverage! LISTEN: Subscribe to the Big Brother podcast feed WATCH:  Watch and subscribe to the podcast on YouTube SUPPORT:  Become a RHAP Patron for bonus content, access to Facebook and Discord groups plus more great perks!

Send us Fan MailWelcome to another load D23 prep show! In this episode we discuss 2 of the big presentations taking place at the Honda Center, the Entertainment Showcase and Parks & Experiences. We go through our hopes and predictions from each of the studios and parks which makes for a great discussion. We hope you enjoy the show!Join us in our completely free Discord https://discord.gg/4nAvKTgcRnCheck out all of our amazing sponsors!Getaway Todayhttps://www.getawaytoday.com/?referrerid=8636If you want to book a Disney Vacation, please use our friends at Getaway Today. Also, if you call 855-GET-AWAY and mention Walt's Apartment, you will get a special dose of magic Where In The Park The Podcast-“Discover the history behind the details of Disney parks and more on the Where In The Park podcast”https://whereinthepark.comCheck Out Sunken City Designs - from the mind of Louis Medinahttps://sunkencitydesigns.bigcartel.com

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

Sports director Tom Ackerman joins the show in-studio to analyze deadline deals across Major League Baseball. Ackerman breaks down how the Los Angeles Dodgers leverage market-leading financial flexibility and deep farm systems to land top talent, while outlining the St. Louis Cardinals' long-term strategy of stockpiling high-level minor league prospects. The discussion also covers potential roster moves involving Lars Nootbaar and Dustin May, local support for women's professional sports, and St. Louis City SC's playoff push following peak television viewership during the World Cup. Hashtags: #TomAckerman #MLBTradeDeadline #CardinalsBaseball #Dodgers #CitySC

Send us Fan MailAn armed citizen refused to sit back and watch.Aaron and Trent examine the deadly shooting outside an In-N-Out in Twin Falls, Idaho, where a rifle-wielding attacker reportedly killed two people and wounded seven others. They break down the footage, the unanswered questions surrounding the shooter's motive, and how an armed citizen returning fire may have prevented an even larger massacre.This episode goes beyond the usual gun-control argument. It's about carrying legally, getting legitimate training, recognizing danger, using whatever options are available, and accepting one ugly reality: denial does not protect you when violence begins.Their message is blunt—being armed without being trained isn't enough, and waiting for somebody else to save you may cost lives.Links: OnesReady.comChapters: 00:00 - Outside The Wire 00:18 - Shooting in Twin Falls 01:01 - Two Dead, Seven Wounded 02:20 - Police Respond to the Chaos 03:29 - Motive, Politics, and Speculation 04:57 - An Armed Citizen Fights Back 06:56 - What If Nobody Was Armed? 09:26 - Return Fire 10:21 - Your Vehicle Is an Option 11:28 - Denial Kills You Twice 13:59 - The Duty to Be Ready 14:34 - Carrying and the Second Amendment 16:52 - The Final TakeawaySupport the showJoin this channel to get access to perks: HEREBuzzsprout Subscription page:  HERERegister for our Operator Training Summit:  OperatorTrainingSummit.comFind an Air Force Recruiter: AirForce.comCollabs:Ones Ready - OnesReady.com 18A Fitness - Promo Code:  ONESREADY ATACLete - Follow the URL (no promo code):  ATACLeteDanger Close Apparel - Promo Code:  ONESREADYDFND Apparel...

Brick Lombardi is in the house for another electrifying session of Tundra FM. As the sun sets, the mood turns introspective, delving into the raw emotions surrounding last season's heartbreaking playoff exit. Tonight, the music sets the stage for a deep dive into the blame game that followed. Key discussion points: - The fallout from the 21-3 halftime lead in the wildcard game. - Fanbase reaction: grief replaced by a quest for blame. - Speculation on dropped passes and offensive line struggles. - The contentious debate: players versus coaching decisions. Tune in for more thought-provoking discussions and the best music on the airwaves. Subscribe, rate, and review Tundra FM to stay connected with the Packernet Network! This episode is brought to you by PrizePicks! Use code PACKDADDY to get started with America's #1 fantasy sports app. https://prizepicks.onelink.me/LME0/PACKDADDY To advertise on this podcast please email: ad-sales@libsyn.com Or go to: https://advertising.libsyn.com/packernetpodcast Check out everything I'm building across the Packers and NFL world: NFL Draft Grades: https://nfldraftgrades.com/ Hashmarks: https://hashmarks.io/

We are ALMOST done with WoBs. I know, that is the real fantasy story, but we do finish the December spoiler stream. Wow! Then we proceed to JordanCon. We have quite a few Stormlight back half WoBs which we just go off the deep in with, which maybe you'll enjoy. Today we have Eric (Chaos), David (Windrunner), Evgeni (Argent), Grace (thegatorgirl), and Bonnie (Cosmeregirl)! Thumbnail is a promo image for War for Roshar by Amirul Hhf: https://bsky.app/profile/brotherwisegames.bsky.social/post/3mrxd7lngf724 0:00:00 Introductions 0:02:57 Connections of children born in the Emberdark 0:05:32 Stormlight Back Half Prologues 0:16:05 Stormlight Back Half Protagonists 0:33:45 Nightblood in a silver sheath 0:41:47 Student of mine 0:42:40 Copper effects with the Emberdark device 1:10:26 Fires of December Cosmere Significance 1:11:40 Which of Hoid's Travail's is His Fav to Tell? 1:13:59 Lesser dragons 1:20:57 What's coming down the pipeline that no one has asked about? 1:27:07 Investiture's association after being used 1:51:51 Voidlight for Surgebinding 2:03:53 Cryptics in reflected light, photos 2:16:30 Who's That Cosmere Character If you like our content, support us on Patreon: https://www.patreon.com/17thshard Purchase merch here! https://store.17thshard.com/ For discussion, theories, games, and news, come to https://www.17thshard.com Come talk with us and the community on the 17th Shard Discord: https://discord.gg/17thshard Want to learn more about the cosmere and more? The Coppermind Wiki is where it's at: https://coppermind.net Read all Words of Brandon on Arcanum: https://wob.coppermind.net Subscribe to Shardcast: http://feeds.soundcloud.com/users/soundcloud:users:102123174/sounds.rss Send your Who's That Cosmere Characters to wtcc@17thshard.com

On today's show, we're reviewing a new book by Andrew Ross Sorkin called 1929: Inside the Greatest Crash in Wall Street History, and How It Shattered a Nation.Sorkin is best known for Too Big to Fail, his account of the 2008 financial crisis. In this book, he goes back nearly eighty years earlier to examine the most famous market collapse in American history.Most people know the basic outline. The stock market rose dramatically during the Roaring Twenties. Speculation took hold. The market crashed in October of 1929, and the Great Depression followed.But knowing the outline is not the same as understanding what happened.Sorkin's strength is narrative. He takes a complicated financial event and tells it through the people who experienced it. Bankers, traders, politicians, regulators, journalists, and ordinary investors appear not as distant historical figures, but as human beings operating under pressure.The book's central lesson is not simply that markets can fall. Everyone already knows that.The deeper lesson is that intelligent, experienced people can see warning signs and still fail to act.Why?Because the incentives of the moment are often stronger than the consequences of the future.-------**Real Estate Espresso Podcast:** Spotify: [The Real Estate Espresso Podcast](https://open.spotify.com/show/3GvtwRmTq4r3es8cbw8jW0?si=c75ea506a6694ef1)   iTunes: [The Real Estate Espresso Podcast](https://podcasts.apple.com/ca/podcast/the-real-estate-espresso-podcast/id1340482613)   Website: [www.victorjm.com](http://www.victorjm.com)   LinkedIn: [Victor Menasce](http://www.linkedin.com/in/vmenasce)   YouTube: [The Real Estate Espresso Podcast](http://www.youtube.com/@victorjmenasce6734)   Facebook: [www.facebook.com/realestateespresso](http://www.facebook.com/realestateespresso)   Email: [podcast@victorjm.com](mailto:podcast@victorjm.com)  **Y Street Capital:** Website: [www.ystreetcapital.com](http://www.ystreetcapital.com)   Facebook: [www.facebook.com/YStreetCapital](https://www.facebook.com/YStreetCapital)   Instagram: [@ystreetcapital](http://www.instagram.com/ystreetcapital)  

Brick Lombardi is in the house for another electrifying session of Tundra FM. As the sun sets, the mood turns introspective, delving into the raw emotions surrounding last season's heartbreaking playoff exit. Tonight, the music sets the stage for a deep dive into the blame game that followed. Key discussion points: - The fallout from the 21-3 halftime lead in the wildcard game. - Fanbase reaction: grief replaced by a quest for blame. - Speculation on dropped passes and offensive line struggles. - The contentious debate: players versus coaching decisions. Tune in for more thought-provoking discussions and the best music on the airwaves. Subscribe, rate, and review Tundra FM to stay connected with the Packernet Network! This episode is brought to you by PrizePicks! Use code PACKDADDY to get started with America's #1 fantasy sports app. https://prizepicks.onelink.me/LME0/PACKDADDY To advertise on this podcast please email: ad-sales@libsyn.com Or go to: https://advertising.libsyn.com/packernetpodcast Check out everything I'm building across the Packers and NFL world: NFL Draft Grades: https://nfldraftgrades.com/ Hashmarks: https://hashmarks.io/

This week, Kat & Carol are talkin' shop on all things pop culture! We discuss the (suspected) drama surrounding the Girls Gotta Eat hosts, the Love Island USA season 8 finale and our islanders outside the villa, and Gracie's new album Daughter from Hell (and if we think it's shady to Taylor Swift.) Then we get an update on what our good friend Liv Schmidt is up to these days (spoiler: it's bad!) and Alix Earle's latest "plastic surgery" rumors and speculation. We end by discussing the negativity and drama surrounding Emilie Kiser and why we aren't here for it. Tune in and join the spiral!Email us at hello@prettymuchfine.com Follow us on Instagram: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.instagram.com/prettymuchfinepod⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Follow us on TikTok: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.tiktok.com/@prettymuchfinepod⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Subscribe to our YouTube channel: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@prettymuchfine2194⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.prettymuchfine.com/

3rd hour of the G-Bag Nation: Sounds of the Sports Day ft Dick Fangio; Reckless Maxx Crosby speculation; Around the Romo and more Cowboys practice observations full 2159 Thu, 30 Jul 2026 23:23:17 +0000 3ciCT995IvMpHvfpBtUmPK49gG0H6M1n sports GBag Nation sports 3rd hour of the G-Bag Nation: Sounds of the Sports Day ft Dick Fangio; Reckless Maxx Crosby speculation; Around the Romo and more Cowboys practice observations GBAG Nation sets the afternoon sports pace for Dallas-Fort Worth with an energetic, roundtable approach that speaks directly to the heart of North Texas. Featuring Gavin Dawson, Super Bowl winning scout Bryan Broaddus, Eric Chiofalo, Zach Wolchuk and Lucious Alexander, the show combines insider-level knowledge, strong debate, and the confident swagger of the Metroplex, plus plenty of laughs and the kind of friendly ribbing you'd expect from a group of best friends. Your drive home is filled with in-depth coverage of the Cowboys, Rangers, Mavericks and Stars. GBAG Nation also tracks college football across Texas along with the biggest national sports headlines, translating them through a distinctly local lens. The GBAG Nation has some of the best contacts in DFW. They pull back the curtain and give you information that no one else can. This is where informed analysis meets bold opinion, with humor and camaraderie that keep it fun and real. © 2026 Audacy, Inc.

Reckless Maxx Crosby speculation & other Cowboys Camp Takeaways full 692 Thu, 30 Jul 2026 23:36:55 +0000 QfGIsTzVUFNKn5UZk0H0c4vjZKGO22YQ nfl,dallas cowboys,sports GBag Nation nfl,dallas cowboys,sports Reckless Maxx Crosby speculation & other Cowboys Camp Takeaways GBAG Nation sets the afternoon sports pace for Dallas-Fort Worth with an energetic, roundtable approach that speaks directly to the heart of North Texas. Featuring Gavin Dawson, Super Bowl winning scout Bryan Broaddus, Eric Chiofalo, Zach Wolchuk and Lucious Alexander, the show combines insider-level knowledge, strong debate, and the confident swagger of the Metroplex, plus plenty of laughs and the kind of friendly ribbing you'd expect from a group of best friends. Your drive home is filled with in-depth coverage of the Cowboys, Rangers, Mavericks and Stars. GBAG Nation also tracks college football across Texas along with the biggest national sports headlines, translating them through a distinctly local lens. The GBAG Nation has some of the best contacts in DFW. They pull back the curtain and give you information that no one else can. This is where informed analysis meets bold opinion, with humor and camaraderie that keep it fun and real. © 2026 Audacy, Inc. Sports https://player.amperwavepodcast

Matty A. and Ryan Breedwell dive into a packed week for the financial markets, starting with predictions for the upcoming FOMC rate decision and the impact of the ongoing Iran conflict on global oil prices. They explore how inflation and geopolitical tensions are keeping the S&P 500 range-bound, while highlighting crucial earnings reports from major AI and semiconductor companies like SanDisk, Seagate, and Nvidia.The hosts also analyze the recent spike in United States real estate foreclosures, breaking down why record-high homeowner equity and supply shortages mean a housing crash is highly unlikely. Finally, the conversation shifts to digital assets, discussing the Crypto Clarity Act, the regulatory threat to meme coins, and how tokenization could soon reshape institutional finance.KEY TOPICS DISCUSSEDFOMC rate hike probabilities and Citadel's surprise hike prediction.Impact of the Iran conflict on global oil prices and WTI trends.Semiconductor stock pullbacks and AI data storage investments.Q2 tech earnings expectations for Meta, Apple, and Microsoft.Analysis of rising United States real estate foreclosures compared to 2019.Record homeowner equity and the national housing supply shortage.The Crypto Clarity Act and the future of real world asset tokenization.Regulatory crackdowns on meme coin markets and platforms like PumpFun.KEY TAKEAWAYSA surprise FOMC rate hike is highly unlikely given current market conditions, despite some hawkish institutional forecasts.Geopolitical energy shocks are being digested faster by the market, with oil prices retreating sharply after recent spikes.Semiconductor and memory storage companies present strong buy opportunities as they continue to beat earnings despite broader tech sector pullbacks.The current real estate market is insulated from a crash due to a massive 11 trillion dollars in tappable equity and pervasive sub-6 percent mortgage rates.The impending Crypto Clarity Act will likely eliminate unregulated meme coin exchanges while attracting trillions in institutional capital to legitimate tokenization projects.CONNECT & TAKE ACTIONImagos Income Fund: Text "INCOME" or "DEALS" to 844-447-1555 to learn more about Matty A's private debt fund targeting 10% fixed returns paid out monthly.

On this Tuesday edition of 2 Pros & A Cup Of Joe, Jonas Knox, Brady Quinn, & LaVar Arrington speculate more on the Michael Lombardi situation in North Carolina as the GM was put on leave. Plus, the guys discuss LeBron living in NYC and traveling to Philly, we have another DWI edition of ICYMI, and more!See omnystudio.com/listener for privacy information.

Summary In this episode, Shelby shares her exciting experience of securing last-minute Southwest travel deals and her plans for upcoming trips. She also discusses her thoughts on Taylor Swift's wedding secrecy and her current TV show favorites. In this episode, we dive into the fascinating world of reality TV, gardening, Disney resort policies, and personal fitness journeys. Most shockingly - we cannot believe Disney's Grand Floridian Resort has CANCELED the annual Gingerbread House! Join us for a lively chat filled with insights, tips, and personal stories that keep the conversation engaging and informative.     Key Topics   Last-minute ticket acquisition strategies Travel deal hacks and airline credits Speculation about Taylor Swift's secret wedding TV show recommendations: Love Island and Off Campus Privacy and security at high-profile events Reality TV casting and production insights Gardening tips for zucchinis and pea pods Disney resort restrictions and alternative ideas Personal fitness updates including Pilates and treadmill issues Innovative ideas for staying cool at Disney parks      Takeaways   Casting directors look for 'psycho' traits for good TV Pinching zucchini flowers can boost yield Resort restrictions are evolving, consider alternative locations Mixing pudding with protein powder creates better ice cream Frozen water bottles in pockets are a park hack for staying cool

Spain's 1-0 victory over Argentina stops them winning successive World Cup finals and their captain Lionel Messi from becoming the first to lift the trophy twice. Speculation is rife that, at 39, it may be his last World Cup tournament. Also: We have a special report on how the soaring value of gold has prompted a surge in illegal mining in west Africa. After a ferry sank off the coast of Guyana, 60 passengers are still unaccounted for. The MV Barima was en route from the capital, Georgetown, to Port Kaituma when the disaster happened. British authorities are waiting for US courts to sanction the extradition of the social media influencers, Andrew and Tristan Tate, who were arrested in Miami on new charges in the UK, including rape. They've always denied any wrongdoing. Why going bare is the latest in thing in nail salons, but does this new trend betray class and racial prejudices? And Rikke Dam speaks to the BBC after winning the women's marathon ... at the North Pole.The Global News Podcast brings you the breaking news you need to hear, as it happens. Listen for the latest headlines and current affairs from around the world. Politics, economics, climate, business, technology, health – we cover it all with expert analysis and insight. Get the news that matters, delivered twice a day on weekdays and daily at weekends, plus special bonus episodes reacting to urgent breaking stories. Follow or subscribe now and never miss a moment. Get in touch: globalpodcast@bbc.co.ukPhoto: Spain's Lamine Yamal lifts the World Cup trophy alongside teammates as they celebrate winning the World Cup. Credit: IMAGN IMAGES via Reuters/James Lang