Monologue: Tokenpocalypse Now

0:00This is an iHeart Podcast. Guaranteed human. This is Malcolm Glaubel from Medal of Honor Stories of Courage. This episode is brought to you by Navy Federal Credit Union. At Navy Federal, they want you to know that if you need a positive sign towards achieving your goal of home ownership, this is it. Because even though today's housing market can seem discouraging, the Home Buyers Choice Loan offers members down payment options as low as 0 % with no required private mortgage insurance. Look for your sign today, Navy Federal Credit Union. Navy Federal is insured by NCUA, Equal Housing Lender. Terms and conditions apply.

0:39Loans subject to approval and eligibility requirements. Learn more at NavyFederal.org slash zero down. What's up, y 'all? Summer's got a different tempo. Everything's a little looser, brighter. One plan turns into another. You hear something, you stay a little longer. Next thing you know, you're somewhere you didn't plan to be. It's those in-between moments. That's where the ideas hit. Conversations stretch out, little memories sneak up on you. Sometimes it's just about what's in your hand, that color, that chill. The new tropical butterfly refresher from Starbucks. Guava and passion fruit flavors with mango pineapple flavored pearls.

1:18Yeah, that feels like summer before you even taste it. Funny how one small stop becomes the best part of the day. Start your summer rhythm with Starbucks. Try the new Tropical Butterfly Refresher from Starbucks.

1:36Hey, everyone, it's Cal Penn. I'm inviting you to join the best-sounding book club you've ever heard with my podcast, Earsay, the Audible and iHeart Audiobook Club. Every episode, I nerd out with amazing guests and dive into the best new audiobooks available on Audible. It's the book club for your ears. Listen to Earsay, the Audible and iHeart Audiobook Club on the iHeartRadio app or wherever you get your podcasts.

2:05Unlock the savings at Boost Mobile and save up to$600 a year. I've been scouting these big carriers for a minute now, and I've seen them pull the same play a thousand times. They promise you the world, then hit you with a price hike right when the game gets tight. But Boost Mobile, their$25 a month unlimited wireless plan is the most consistent player on the floor. No contracts, no price hikes. Unlock the savings today at boostmobile.com slash unlock. Based on average annual single line payment of AT &T, Verizon, and T-Mobile customers compared to 12 months on the Boost Mobile Unlimited Wireless plan as of January 2026.

2:34For full offer details, visit boostmobile.com. CallZone Media. Hello and welcome to the second of this week's Better Offline Monologues. I'm your host, Ed Zitron.

2:53It's been a funny few weeks watching the little spurts of air coming out of the AI bubble. It's unclear when it will burst, what will burst it, but really, even what's going on half the time, they just all seem to be running around like a billion Mr. Beans. But I get the sense that everything will accelerate dramatically based on one or two big events, things that just knock the confidence of everyone. Maybe it's an AI company dying, or a funding round not coming together, or maybe it's a hyperscaler cut in CapEx. Goldman Sachs analyst Rich Piratovsky said last week that he believed that everybody was spending simply to remain competitive, and added that the first hyperscaler to signal that it can slow the pace of spending would likely see its share price rewarded.

3:35If I had to bet, I think it's between Meta and Microsoft, the latter of which has made numerous noises about using cheaper AI models like DeepSeek and its co-pilot co-work product. Yeah, and also I should add that Satya Nadella keeps doing interviews where he's like, yeah, no one company should have all that power. We shouldn't rely on one lab. One lab being able to be shut down at any time is bad. It's really funny watching this guy change his tune, considering how a year and a half ago, a year ago, he was rock hard for open AI and six months after that rock harder for Anthropic. It's almost as if he's just doing random shit based on what he thinks might work.

4:13You know, like a fucking cargo cult. Anyway, as I've mentioned over the last few months, both Anthropic and OpenAI both started to charge their enterprise customers, companies with over 150 users, for the actual cost of their AI token spend as of the middle of Q1 2026. To explain, when you use a regular chat GPT or Claude account, you burn tokens, each one about three quarters of a word, up to a five hour or weekly limit, depending on what kind of model you're using. So with Anthropic on their Claude subscription, Opus 4.8 and Fable when it was around had a specific limit to themselves. But anyway, just the more powerful the model, powerful being meshed by the companies, of course, the more it burns and the less you can use it.

4:54On a regular subscription, you're able to burn, and I am not kidding,$8 ,000 worth of tokens a month on Claude and$14 ,000 a month on OpenAI's ChatGPT Codex subscription, all for$200 a month. Pretty good deal, right? It's like the discount prices sketch from Tim and Eric that I reference every so often just to see if anyone emails me. And so yeah, after years of being able to burn thousands of dollars of tokens for$200 a month, enterprises are suddenly having to pay the actual costs. The result is multiple companies capping their workers' token spend. As I've reported, both T-Mobile and Brex have done so, and others have reported that Uber, Meta, Walmart, Coinbase, and Cisco have all done so.

5:36Meta notably had a token-maxing leaderboard. Very funny, again, just changing direction based on random signals. None of these people have a plan. None of them. None of these companies have a plan. Nobody integrating AI has a plan. If they did, they have been like, yeah, let's make sure we know how much this is going to cost, or let's make sure we know whether this is good or not, whether we can measure the ROI, you know, anything that would suggest anyone knew what the hell they're doing. But they don't. And 404 Media reports that management consultancy firm and evil corporation Accenture, which is an insane way of saying that that I'm just going to keep, has seen what it calls soaring token spend based on leaked audio recordings with with much of that token spend being driven by non-engineers doing things like converting pdfs into presentation slides i love this shit this is the easiest stuff in the world made slightly easier but probably worse costing way more you could definitely hire an extra but you could just hire a contractor to do this you could hire a bloke at 45 bucks an hour to do that for you he'd be the happiest man alive probably cost you less too i know god this stuff is exhausting.

6:47You ever sit and wonder what you could do with the rest of that money? I could have, I don't know, like, what, like millions of dollars a month? I could have, like, a week's worth of Diet Coke. Anyway, it turns out that telling your workers that AI can do basically anything and to use it as much as possible starts costing you a lot of money when you actually pay what it costs. As I've said before, we're barely three months into enterprises paying the actual cost of AI, and they're already screeching like they're being pecked to death by birds, and that joke is courtesy of Casey Kagawa, who said that to me about a year ago when we talked about the amount of money that Anthropic was allowing people to burn on Claude's subscriptions.

7:24Meanwhile, the threat of open-source models appears to be getting more serious. I've heard from numerous sources that Chinese AI Lab Zipu's GLM 5.2 coding model is competitive with Anthropic's Opus 4.8 at somewhere between a quarter and a sixth of the price, but that seems to vary based on the cost of the task. As far as whether GLM 5.2 is actually competitive goes, it appears that the hype is somewhat real, but you kind of had to check your sourcing. Open source AI coding company Klein ran benchmarks on a natural bug from its repo and found that while GLM used twice as many tokens to fix it, it also did so at a cheaper cost, 41 cents versus Opus 4.8's 81 cents.

8:04Opus was quicker, taking 1.6 minutes versus GLM's 4.7 minutes. But one thing that stood out was that GLM, to quote Klein, cleaned up dead code and verified the build compiled before completing, something that Opus, the more expensive frontier model run by the company that loses billions of dollars, apparently didn't bother to do. It's just one bug, but GLM 5.2 seems to score well on many of the major benchmarks, coming third on cybersecurity firm Semgreps' cyberbenchmarks, with the top two being Semgreps' own harness, so their own thing they put on top of the models running opus 4.8 and gpt 5.5 followed by glm 5.2 which is interesting it's very interesting if i if i was glammy samuel when i'd be looking at this and getting a little worried i'd genuinely be looking at them and even saying oh why don't i train my own but you know that might actually be a bad idea even though it's fully possible because glm 5.2 is a true open weights model it's released under something called the unrestricted MIT license, which means you can, anyone, by the way, company, a regular person, whomever, can self-host and fine-tune it or use it in any way they wish.

9:11And yeah, I mean, in theory, any company that wants to use LLMs could either run their own version of it locally, or be on a rig that costs tens of thousands of dollars, or pay an inference provider like Base10 to spin up and train their own version. Now, I must be clear that there's not a ton of evidence that running GLM 5.2 is profitable for the inference provider. In fact, there's really none. I genuinely have looked, nor do we have any significant signs that customers are moving away en masse from OpenAI and Anthropics models. People generally like name brands, and they, even though Anthropics shit is constantly broken, go look at their stability page.

9:45Stability page? Eh, uptime page? I don't know what you call that. Email me if you know what they call that, or I could just look after this. Gonna keep going, though. Nevertheless, people want name brands. People like paying for a company that ostensibly works on this and keeps it updated. How true is that? Who knows? I don't think people are buying LLMs because of particularly sensible reasons anyway. But what might change things would be Microsoft, Amazon, or Google offering GLM 5.2 via their Foundry, Bedrock, or Vertex AI platforms. And even then, it's hard to tell if customers are excited about AI or just paying OpenAI and Anthropic to use buzzy models.

10:19Really isn't clear. We should know by now. All that being said, now does feel like the perfect time for a shift towards open source models if only because everyone is crapping themselves about costs everyone is really freaking out and i think that this is going to be the time when we start separating the wheat from the chaff the people that actually like this stuff versus the people who just feel good because they're doing ai and they don't know what else to do with their time and they like forcing people to use it i'll get to who actually does in a minute

10:58Support for today's episode comes from Square, the business platform that helps sellers become neighborhood favorites. Whether you're gearing up for a busy season or just trying to keep up with everyday demand, Square keeps your business running smoothly, from payments and POS to online orders, inventory, staff, and more, all in one place so you can focus on your customers, not your to-do list. I like Square a lot as a customer, and from talking to various business owners who use it, everybody seems really happy with how easy and intuitive everything is. Business owners don't have to juggle a bunch of tools and their hardware and software are designed to be easy to use so day-to-day options feel a lot simpler with three clear plans depending on the tools you need sellers who use square see nine percent higher sales on average square helps you run your business more smoothly bringing payments operations and insights together in one place so you're ready for whatever's next right now listeners can get up to two hundred dollars off Square hardware when you sign up at square.com slash go slash better offline.

11:53That's S-Q-U-A-R-E dot com slash go slash better offline. Get started with Square and build a setup that works the way you do. Unlock the savings at Boost Mobile and save up to$600 a year. I've been scouting these big carriers for a minute now, and I've seen them pull the same play a thousand times. They promise you the world, then hit you with a price hike right when the game gets tight. But Boost Mobile, Their$25 a month unlimited wireless plan is the most consistent player on the floor. No contracts, no price hikes. Unlock the savings today at BoostMobile.com slash unlock. Based on average annual single line payment of AT &T, Verizon, and T-Mobile customers compared to 12 months on the Boost Mobile unlimited wireless plan as of January 2026.

12:33For full offer details, visit BoostMobile.com. Wait, this is a soda? Yeah. And it has protein? 10 grams. No sugar? Zero. And it actually tastes good? It's Sky Pop. Sky Pop Protein Soda delivers the refreshing taste you want from a real soda. Crisp and delicious with 10 grams of complete protein, zero sugar, and just 45 calories. So you're not choosing between great taste and real benefits. You're getting both in every sip. Sky Pop Protein Soda. Reach for the sky. Get your Sky Pop Protein Soda now at Target or Walmart. Hey everyone, it's Cal Penn, host of Earsay, the Audible and iHeart Audiobook Club.

13:11This week on the podcast, I'm sitting down with Divergent author Veronica Roth to talk about her sprawling new novel, Seek the Traitor's Son. It's a sci-fi fantasy epic about two protagonists on opposite sides of a war and a prophecy neither of them wanted. My first book was Divergent. And when that came out, like, because it was so popular, I think it attracted, like, mostly positivity, but the negativity, I sucked in like a sponge. And I think it was like critiques of things I liked when I was like, you know, I was 23 and I wrote this book and it had all my like dorky little cheesy or maybe unrealistic loves in it.

13:49And I started to feel a lot of shame about those things. And so for the rest of my career, I steered away from those little things that like make you feel pleasure when you read. But I also was like saying no to these parts of myself that I then was like, screw it. So that's this book. Listen to Earsay, the Audible and iHeart Audiobook Club on the iHeart Radio app or wherever you get your podcasts.

14:20Open Router's rankings show a dramatic move away from Frontier Labs. The top 10 most popular models are all open source, say for Opus 4.7 at 6, Opus 4.8 at 8, and Sonnet 4.6 at 10, as well as a mysterious new model called Owl Alpha, which is currently free to use, which is probably why it's trending. To be clear, Anthropix models still dominate the top models by task across the board, and that's based on the spend on the task. I hate to give him any credit, though. I really don't like doing this, but Scott Galloway recently made the point that China could thrive by AI dumping high quality open source models into the ecosystem as a means of destabilizing the major model providers.

14:59And I hate to say this was happening, it's like a fucking broken clock I guess. Zipu and Minimax are both partially government funded by the Chinese government, and also both woefully unprofitable, but neither do so at the scale of Anthropica or OpenAI, and while both do the kayfabe discussion of AGI and autonomous compute and all that bullshit, they seem far more focused on creating open-weight models to compete with them. And just for the differentiation here, open-source means you can share some of it, open-weights means you actually can diddle with the model itself. That's a flattening, someone's going to email, they're going to be mad at me, I don't care.

15:33Sorry, like, it gets the point across, alright? Just write on the subreddit if you're mad. Anyway, in truth, I think the future of LLMs will end up being locally run models on expensive specialist hardware. The underlying economics of cloud-based GPU compute do not make sense for anybody. You need to make sure the GPUs are saturated, you have to buy in advance, otherwise you won't, like, you just won't be able to get the capacity, and if you have less people than you expect, you will lose money for sure, and if you have more people than you expect, you have to get more compute, which means you'll also lose money.

16:04So as long as you can Goldilocks this bullshit, maybe you'll eke out the world's worst gross margins in history, or you'll just lose money, which is what everyone's doing. But I think there may be a point at which Zipu, Minimax, or even American model developers start focusing more on services-driven custom deployments, and become boring, ugly, license-driven monstrosities like Oracle. And that's the thing, I've talked to quite a few people recently, and they are all saying the same thing. Even NVIDIA is moving into this realm, and they're selling these giant 100 grand machines. I think that's kind of interesting.

16:42And like I said, that GPU compute, it requires a certain critical mass of customer demand that I don't believe will exist once the cargo culture of the AI bubble passes, because GPU commitments are done a few years in advance, and any drop or lull kills the margins of even a cheap-to-run model. Remember, Base 10, they're an inference provider. They just raised$1.5 billion. They're not profitable. I mean, Zipu isn't profitable, and they do the same thing OpenAI does with their stuff, where they say, oh yeah, yeah, our cost of revenue is lower than our revenue. However, when you add the sales and marketing, it's actually unprofitable as well.

17:18Oh, also, we lose hundreds of millions of dollars because of training. Other than that, we're super profitable. Yeah, it's a really fucking stupid industry. And look, even if GLM 5.2 can do things at half the cost of Opus 4.8, that's still half of what in some cases is hundreds of thousands or millions of dollars a month. I've heard tell from multiple sources that that hardware demand is growing and that customers at big companies are starting to look into those high-end hardware solutions. And I think that that is the only future for this stuff. And I think it would reduce the cost significantly for the end user, probably keep the shit off the internet, or at least massively marginalize the damage that LLM-based code can do by focusing on people that actually give enough of a shit to buy the actual machines.

18:06And I don't know, I feel like an actual serious fiscal commitment makes companies take things a little bit more seriously. Unless, of course, you're Microsoft, Google, and Amazon, and Meta buying GPUs, at which case, who gives a shit? But I mean, if this takes off, it's curtains for Zusha. It's a bad time for OpenAI and Anthropic, unless they, too, start doing these hardware-driven things. This is just a guess. This is just a guess they would ever think about that. But I think it's possible. and I think it's really the only economically viable end for this unless there are just going to be companies that lose billions of dollars a year for no apparent reason I just don't think that's going to happen it's not enough money to do it and god damn please go and listen to the open AI will not get bailed out AI is not too big to fail podcast I did I'm tired of getting please don't please stop emailing me I love hearing from you I really I would love to hear from you or more other than the people who email me saying i think the gpus are all for surveillance i think it's all that i think it's a big government that's called oracle oracle's first client was the cia oracle has a giant government gpu array they have several of them jesus christ that's already happening and they're not doing surveillance on that they did that with maven they did it with the promise software jesus fucking christ sorry i understand that i sound agitated here.

19:25It's just that if the only choice you can make when looking at something is to say, well, bad thing will definitely happen, that's not actually an intellectual pursuit. It's just crapping yourself. It's just saying, well, everything's bad, so everything stays bad. I actually want to bring you a message of hope. I don't think these companies get bailed out. I don't think they get anything. I think that they get allowed to die. I think partner assets might be scooped up by Amazon, Google, and Microsoft. I think the government might take a chunk out of them. Perhaps they get some sort of bridge financing, but I actually don't know if the government would even extend that because AI is very unpopular with regular people.

20:04Regular people do not like this and they hate data centers and they have intelligently connected data centers to these horrible AI companies. And so in the end, I tell you, please stop emailing me the surveillance thing. It's not true. You're wrong. You sound like someone that wants to be paranoid, but doesn't want to do the reading well enough to be paranoid. If you want to be paranoid, go read the Promise software. Go look at Oracle. Don't look at the actual history of these companies if you want to be freaked out by something. But in truth, the AI industry is just a directionless egregore of capital that bumbles about and is only kept alive by a belief system that's mostly growth-focused and doesn't really understand how to run a business anymore, or even build innovation or technology.

20:51And it's blatantly obvious that LLMs are, at best, a classic risk-versus-reward software tool for the real sickos that know what they're getting into. But the problem is it's been sold as this mass-market panacea for basically any problem. And I'll say that my thesis remains undeterred. i believe a great deal of ai usage is cargo culture driven by executives that don't do any work and people that feel pressured into adopting ai by their peers of the media or of course people that spend 12 to 18 hours a day on twitter who just follow the algorithm i call it the dunces casino because you just crank the thing and a bunch of insane people come out and all of these people come out and they're like if you don't use agents you're going to die they're going to kill your wife they're going to eat your dog if you don't use loops boris cherny's going to blow up your house ah twitter is driving people insane using ai but not because of grok or anything but because it just fire hoses nutters into your feed who will constantly make you feel terrible for not picking up the next big ai thing it's this insane it's i keep saying cargo cult because that's what it is it's people just going around being like yeah you know this is the big thing now we all like that, right?

22:06We love it. We love it. It's good. If we all follow this, the great prophecy will come true and everyone will make money somewhere, somehow, I think, to sickness.

22:24Support for today's episode comes from Square, the business platform that helps sellers become neighborhood favorites. Whether you're gearing up for a busy season or just trying to keep up with everyday demand, Square keeps your business running smoothly, from payments and POS to online orders, inventory, staff and more, all in one place so you can focus on your customers, not your to-do list. I like Square a lot as a customer and from talking to various business owners who use it, everybody seems really happy with how easy and intuitive everything is. Business owners don't have to juggle a bunch of tools and their hardware and software are designed to be easy to use, so day-to-day options feel a lot simpler.

22:57With three clear plans depending on the tools you need, sellers who use Square see 9 % higher sales on average. Square helps you run your business more smoothly, bringing payments, operations, and insights together in one place, so you're ready for whatever's next. Right now, listeners can get up to$200 off Square hardware when you sign up at square.com slash go slash better offline. That's s-q-u-a-r-e dot com slash go slash better offline. Get started with Square and build a setup that works the way you do.

23:56compared to 12 months on the Boost Mobile Unlimited Wireless plan as of January 2026. For full offer details, visit BoostMobile.com. Wait, this is a soda? Yeah. And it has protein? 10 grams. No sugar? Zero. And it actually tastes good? It's Sky Pop. Sky Pop Protein Soda delivers the refreshing taste you want from a real soda. Crisp and delicious with 10 grams of complete protein, zero sugar, and just 45 calories. So you're not choosing between great taste and real benefits. You're getting both in every sip. Skypot Protein Soda. Reach for the sky. Get your Skypot Protein Soda now at Target or Walmart.

24:32Hey everyone, it's Cal Penn. I'm the host of Earsay, the Audible and iHeart Audiobook Club. This week on the podcast, I'm sitting down with Ray Porter, the narrator of Andy Weir's audiobook Project Hail Mary, massive sci-fi adventure about survival and science and what happens when you wake up alone very far from earth. I really had to make a decision because I caught myself getting that frog in my throat and starting to get teary as I'm narrating some of these sections. And it's like, okay, yo, yo, yo, is this indulgent? And I really thought about it. I was like, no, at this point, it would kind of be betraying the trust the author and the listener have in telling this story if I don't go through it.

25:16But there's places in this book that deeply emotionally affected me and i left it on the mic that's great because it served the story people will say like oh my god i cried at the end it's like yeah dude me too listen to earsay the audible and iheart audiobook club on the iheart radio app or wherever you get your podcasts

25:39it's a genuine sickness so i went a little tuck across and i what's going on why are we using LLMs. What are they going to do to us? What are they going to do to your family? But anyway, back to the podcast, back to the monologue. I apologize for ad-libbing. I must be clear as well, I do love hearing from you because I keep getting emails from people that actually use AI, and I got one just before I started recording this from a bloke who said, and I quote, he's used it to deliver good work and seen it tried to submit absolute cramp. And that really is the AI industry. Hey, you know, sometime it works, and when it does, it's alright, and when it doesn't, it's dogshit.

26:17Yay! Great, I'm so glad we sunk a trillion plus dollars into this. Look, these are specialist tools that are being forced to solve generalist problems, and they're probabilistic models that fail to do deterministic things, and they're really just an eternal financial black hole for anyone trying to scale them into a Google search or a meta-sized business and it's never going to work like that uh just before this as well i saw that broadcom and open ai announced their jalapeno chip this giant dinner place sized motherfucker dinner place dinner plate i'm not fixing that and it's funny because everyone everyone's doing the cargo cult shit around it they're all like oh well it's gonna bring down inference costs if they did that that would fix everything if if jalapeno fixes everything then everything will work out there's no proof that that's the case actually here is the thing if you want if you want to like here's a specific request if you have an information subscription at some point in the last year i know there was an information story that suggested that broadcom's open ai chip would have modest gains if you bring that to me i will fuck i will venmo you five dollars if you can bring me that story seriously i need that story because i'm pretty sure that if broadcom's chip was going to change everything for open ai open ai would sink every fucking dollar they had into it open ai would dispense with nvidia instead it's like oh yeah we're getting microsoft to pay for some of it and it'll be there in 2027 maybe uh i don't know sounds like something you pre-ordered on kickstarter and you got it and it kind of sucks it turns out the video was ai generated which i did just see with this thing where someone had done a kickstarter where it's uh yeah you can build things using real principles and you pour your own cement but it's really small and the whole thing is ai generated fucking hell this industry really is miserable isn't it all the people every other era other than of course crypto and the metaverse when something cool happened even in the augmented reality era there were still people doing goofy funny shit there were still people doing really interesting things with computer vision there were people doing interesting things with depth sensing.