This Week in Tech 1087: Evil is the Root of All Money

Summary

Podcast Summary: This Week in Tech 1087

Episode Title: Evil is the Root of All Money
Date: June 8, 2026
Host: Leo Laporte
Panelists: Father Robert Ballecer (“Padre”/B), Jeff Jarvis (C), Joey de Villa (D)

---

Episode Overview

This freewheeling, insightful episode unpacks a whirlwind week in tech marked by record-shattering IPO buzz (Anthropic, SpaceX, OpenAI), massive AI infrastructure investments, the looming AI “pause” debate, the rapid advancement of AI self-improvement, questions of consciousness and ethics, imminent Apple announcements, distributed computing dreams, and more. The hosts dissect market dynamics, technical breakthroughs, social impacts, and the virtues and follies of the AI arms race, always with a dose of humor and lived perspective.

---

Key Discussion Points and Insights

1. The "IPO Palooza" & AI Mega-Bubble

• SpaceX, Anthropic, and OpenAI IPO Plans:
  • Anthropic’s IPO briefly dubbed “largest ever,” quickly outdone by SpaceX, which is set to value at $1.77 trillion.
  • [04:12] C: “Largest by a factor of three. The next biggest one was Saudi Aramco, but it was a third this size.”
  • Discussion of whether such valuations are justified, with skepticism about whether these companies can generate the profits to match their sky-high market caps.
• Secondary Offerings from Big Tech:
  • Google announced an $80 billion stock issuance (with $10 billion to Warren Buffett/Berkshire Hathaway), raising funds primarily for data center expansion.
  • [05:44] C: “Does it dilute the existing [stockholders]?”
  • A: “Yes, it does. The stock went down as a result for a few days… But they’re saying ‘we’re going to invest upon investing, we’re going to work, we’re going to beat everybody here.’”

2. Data Center Boom and Backlash

• Unprecedented Investment:
  • Amazon, Google, Meta, and XAI investing hundreds of billions each into building data centers to fuel AI development.
  • Some skepticism on whether these centers will ever provide meaningful returns or even be completed, given power constraints, local bans, and potential obsolescence.
  • [07:20] B: “Whether or not they're going to make money off them, that's a huge unknown.”
• Technology’s Rapid Obsolescence:
  • XAI’s $20B data center already lagging behind in chip technology.
  • [09:28] B: “You're going to get to a point where it's no longer economically feasible to keep running them…when you could get 2x, 3x performance from new chips.”

3. Anthropic's AGI "Pause" and Self-Improvement Claims

• Anthropic’s Paper on Recursive Self-Improvement:
  • Claims rapid advances in their ability to use AI (Claude) to write and improve its own code—over 80% of Anthropic’s code is now AI-written, “8x” previous output.
  • Skepticism on whether the “success rate” metrics are meaningful.
  • [12:40] C: “They extrapolate from that to self [improvement and AGI].”
  • [14:17] B: “Self-improvement at scale. I’m still not super convinced…”
• AI as Security Threat—Vulnerability Detection:
  • Mythos (new model) is especially adept at code audits, surpassing human security professionals in certain aspects.
  • [15:50] B: “If you're a security professional…the first time you get your hands on Mythos, it will make you have a lot of self reflection about your chosen career.”

4. The Arms Race, US-China Competition, and Motivation

• Why No One Will Pause:
  • The financial and power incentives fueling the race to AGI are irresistible.
  • [14:46] B (citing Munger): “Show me your incentives and I will tell you your outcomes. The incentive right now is to be the first. There is no prize for second place… Pause does not match the incentive.”
• Global Dynamics:
  • Even if Western companies paused, China would race ahead, making a unilateral pause unrealistic.
  • “Even if you did [pause], the Chinese are rushing as fast as they can to keep up.”

5. AI Ethics, Consciousness, and "Value Alignment"

• Are LLMs Approaching Consciousness?
  • Philosophical back and forth over whether “understanding” by AIs can be equated with human understanding.
  • [26:49] A: “We don’t know what conscious means. We don’t even know what understanding means.”
  • “If you can’t tell the difference [between a human and AI], what does it matter?”
• Limits of Transformers and Analogy to Human Brains:
  • Geoffrey Hinton’s argument: digital AIs’ advantage lies in scalability and immortality; analog brains have unique, uncopyable characteristics.
  • [30:20] B: “True consciousness also allows for the thought process itself to…mutate. …It’s in those variances that you find the unique nature of human conscience.”

6. Task-Specific vs. General AI

• Specialization Over Generalization:
  • Debate over whether progress comes from narrowly focused AIs. LLMs as generalists risk dangerous ambiguity, but targeted models excel at defined tasks.
  • [18:36] B: “The reason why there was so much societal change…the printing press caused so much societal change that people did not expect.”
  • LLMs: “You have to have that specific purpose. …you’re going to get quicker to your goal if you have specified that goal.”

7. OpenAI and the "Super App" Race

• OpenAI’s Next Move:
  • Plans to evolve ChatGPT into a "super app" that goes beyond chat and includes code-generation and AI agents, aiming for more lucrative products.
  • [20:05] D: “This fascination with a one-super-thing… In engineering, originally people thought you’d have just one motor at home and distribute energy with pulleys and gears, but in reality… you now have dozens of electric motors in your house.”

8. Distributed and Local AI Computing

• Rise of Consumer-Accessible AI Hardware:
  • Nvidia’s announcements at Computex for the RTX Spark (a consumer laptop chip) and more affordable, powerful local AI.
  • Discussion if running local LLMs (like “Quinn 36B”) is "enough" vs. the need for massive frontier models.
  • [50:06] D: “It’s fantastic for coding. …it is good enough for a lot of work.”
• The distributed data center dream:
  • Pulte Homes/Nvidia partnership hints at future where homes have mini-data centers, operating as a distributed network.
  • [52:02] C: “Does everybody end up with a large machine in the home that handles all your stuff, but also becomes part of a larger neural network?”
  • Limitations: networking speeds, power, climate impact, and actual needs questioned.

9. Ethics, Faith, and Human Dignity in Tech

• AI as Tool and Responsibility:
  • Debate on whether to “cede primacy” to AI.
  • [37:23] C: “That’s giving up control and perhaps even agency.”
  • [37:35] B: “That is…part of the core…that it is a tool and you cannot give up primacy to a tool.”
  • [38:29] D: “[If you] give primacy to a tool…this is Robotron. This is the Cylons. …We do not want that!”

---

Notable Quotes & Memorable Moments

• On Valuations and Tech Bubble:
  • [04:12] C: “Largest by a factor of three…This would value SpaceX at a whopping $1.77 trillion. Is that all?”
• On Building Data Centers:
  • [07:20] B: “They will build them, yes, they will use them, but whether or not they're going to make money off them, that's a huge unknown.”
• On AI Contribution at Anthropic:
  • [11:14] “As of May 2026, according to Anthropic, more than 80% of the code we merge into Anthropic's code base was authored by Claude.”
• On Human vs. Machine Progress:
  • [18:36] B: “The printing press caused so much societal change that people did not expect…It made it possible for everyone to have a copy of the Bible.”
• On AI Self-Improvement:
  • [14:16] B: “Self-improvement at scale. I'm still not super convinced on that.”
• On Anthropic's IPO and Pause Calls:
  • [14:34] C: “Once we have our IPO in, we’re going to pull the ladder up behind us and tell everybody to stop.”
• On AI and Consciousness:
  • [26:12] C: “He [Hinton] then makes all those leaps to say, well, of course it’s conscious. …I don’t go there.”
• On Apple and Future Announcements:
  • [77:09] B: “Tim Cook deserves credit for what he’s done since Jobs went away. He really has done wonders for Apple.”

---

Timestamps for Important Segments

• [04:12] SpaceX/Anthropic IPO scale, Google secondary offering
• [06:09] AI data centers: power, cost, and overbuild risks
• [10:09] Delivery risk and power constraints for data centers
• [11:14] Anthropic’s claim: 80% of code by AI
• [13:07] Skepticism of code output metrics
• [14:46] “Show me your incentives…” Race to AGI, China vs US
• [15:50] Security professionals and Mythos’s disruptive capabilities
• [18:36] Printing press analogy, societal impacts
• [22:48] Hinton on analog/digital cognition and immortality
• [30:20] True consciousness and the analog brain vs. transformer models
• [50:06] Running LLMs locally; “good enough” vs. frontier models
• [52:02] Distributed AI data centers in home backyards
• [77:09] Apple, Tim Cook’s legacy and speculation on WWDC
• [87:05] Father Robert’s Vatican AI: custom LLM trained on centuries of digital Church texts, privacy aspects
• [117:30] Meta’s glasses and the readiness for biometric/facial ID tech
• [124:14] Biden/Trump administration AI policy, voluntary government review periods
• [126:10] Trump administration considering nationalizing stakes in AI companies

---

Closing: Cultural Shifts & Media Consumption

• YouTube Overtakes Netflix:
  • [150:41] “YouTube has longer daily average viewing around the world over Netflix.”
• Rise of User-Generated and AI-Produced Content:
  • [148:42] Obsession, Back Rooms—huge box office from YouTuber-led films.

---

Overall Tone & Language

Energetic, irreverent, skeptical but enthusiastic; the panelists combine deep technical knowledge with humor, and don’t shy away from philosophical, societal, and ethical debates. The language remains conversational, with plenty of asides, historical analogies, and jokes about failed tech or generational shifts.

---

Conclusion

TWiT 1087 delivers a panoramic, thought-provoking tour through the week’s tumultuous tech intrigues—ranging from corporate strategy to the soul of intelligence, from market speculation to the practical challenges of running AI at home or at the Vatican. The episode is a must-listen (or read) for anyone tracking the social, technical, and ethical contours of the AI age.

---

Notable Quote to End:
[127:32] D: “You know what they say, evil is the root of all money.” (Episode title inspiration)

Transcript

[A]

It's time for TWiT this week at Tech. Father Roberts here, Jeff Jarvis, Joey de Villa. We are going to talk about the IPO Palooza from Anthropic, SpaceX and OpenAI. Nvidia's announcements at Computex, Microsoft's announcements at Build, and what Apple might be talking about tomorrow. This Week in Tech is next. Podcasts you love from people you trust. This is twit. This is TWiT this Week in Tech. Episode 1087, recorded Sunday, June 7, 2026. Evil is the root of all money. It's time for TWIT this Week in Tech, the show. We cover the week's tech news. So glad you're here and so glad to welcome our fabulous panel this week. Father Robert Ballis here is joining us from the Vatican City. Hello. Actually, you're not in Vatican City, are you?

[B]

No, no, no. Vatican City is behind me. We're in the Vatican. But Vatican City is a very specific space.

[A]

So confusing. High atop looking down on St. Peter's Basilica or something, I don't know. Anyway, welcome. Robert. What time is it?

[B]

Oh, it's, it's 11 o'. Clock. But I mean, really, I'm back in California. That's where my heart is.

[A]

Yeah, you're in California. Time good. Where it's just, you know, 2:00pm Also with us, Jeff Jarvis. He's on the east coast in beautiful New Jersey. Normally we see him every Wednesday on intelligent machines, but it's nice to have you on the.

[C]

As you be on the grownups table.

[A]

The grownups table. Jeff, professor emeritus of journalistic innovation at the Craig Newmark Graduate School of Journalism at the City and University of New York. Oh, I forgot. We don't have to do that anymore. Craig, Craig, Craig, Craig, Craig, Craig, Craig. That's another one. I know. I don't even know that one. His brand new book, Hot Type, which is the history of the Linotype and more, emerges from seclusion.

[C]

I just finished recording the audiobook this last.

[A]

Oh, you don't sound hoarse at all.

[C]

It was. I did in three days. 300 pages in three days.

[A]

And that's Joey de Villa. Good to see him. Hey, good to be here. Many of our shows. He has a new job as AI developer advocate at. Well, just Developer Advocate.

[D]

Developer advocate.

[A]

But yeah, Developer advocate at Netflix.

[D]

But my primary. Yeah, I'm really pushing NetFoundry's new AI tools. So little gateways for LLMs and MCPs and ways for agents to talk to other agents.

[A]

Securely Net Foundry was a longtime sponsor on our shows. We Love them. They are, of course, the creators and sponsors of the fabulous Open Source Project OpenZT. Great to see you and congratulations on the new gig, Joey.

[D]

Thank you so much.

[A]

Do they let you bring your accordion to work?

[D]

Oh, yeah. In fact, I believe I played my accordion during the job interview.

[A]

No.

[B]

Yeah.

[A]

No.

[B]

Oh, yeah.

[A]

Wow. And you guys.

[C]

I can't do that.

[A]

Wow, that's fantastic.

[C]

What did you play?

[D]

Oh, my Afroman parody. Because of A.I.

[B]

oh, nice.

[A]

Because of A.I. because of A.I. you'll find that on YouTube. It's a YouTube short. I love it. I love it. Well, what a crazy week this has been. I guess we should start with. I don't know where to start. The week started with Computex and Jensen Huang from Nvidia speaking at his keynote at Computex in Taiwan and announcing a whole bunch of stuff. Next day, Microsoft's Build conference announcing a whole bunch of stuff. And then Anthropic filed for, well, what briefly, Wired called the largest IPO ever, immediately scooped by SpaceX, which is going to be the largest IPO ever. They are looking at a $135 per share price.

[C]

Largest by a factor of three.

[A]

The next biggest one was Saudi Aramco, but it was a third this size. This would value SpaceX at a whopping $1.77 trillion. Is that all? Really? Somebody did some calculations and said it's going to have to. To meet that value. It's going to have to make some. I forgot what it was. I wish I'd written it down. Some outrageous 60x.

[B]

Yeah, 60x of their current revenue.

[A]

Huge amount of money.

[D]

Is there even that much money out there?

[B]

Well, if the tech companies keep trading between themselves, they just sort of have an endless supply. So, sure, make it up.

[C]

But Joey, you're right. You add in SpaceX, Anthropic, OpenAI, and Google's surprise entry of an $80 billion billion dollars raise.

[A]

Explain that, because Google's already public. But we're seeing companies that are already public file for secondary.

[C]

Right? So they're going for $80 billion, 10 billion of which is going to Berkshire Hathaway, which is a big deal because Berkshire Hathaway under its old management. Oh, what's his name? We forget so soon. Wasn't big in technology stocks.

[A]

You mean the wizard of Omaha?

[C]

Of OMA Hub. And so Google has twice, at least twice that in cash on hand, plus credit up the ying yang. Who's not going to loan to Google.

[A]

But is it a loan or is it a stock issue?

[C]

Oh, stock issue. It's a Stock issue.

[A]

Does it dilute the existing.

[C]

Yes, it does. The stock went down as a result for a few days. Well, then everything went down because of Broadcom. But they're saying, we're going to invest upon investing, we're going to work, we're going to beat everybody here. It was an aggressive move by Google,

[A]

and of course, what they're doing is raising money to build data centers. Right? I mean, this is.

[C]

Nobody lets them build them anywhere.

[A]

Yeah, yeah. In fact, you know, that's what everybody's doing at this point. Meta's stock took a big hit when they announced in their quarterly results they were going to spend a huge amount. Google actually said they were going to spend, I think, $180 billion on data centers, and that didn't hurt their stock. Amazon announced a similar figure. I mean, do we need all these data centers? Is that it feels. This is making it feel more bubbly all the time. And I'm not a big proponent of the AI bubble story, but, boy, it's starting to feel like that.

[B]

Look at what they've already spent. You've got Amazon, who is close to 300 billion already. It's 291 billion that they've invested into AI. You've got Alphabet, you've got Google going with 262 billion. Meta is what, something like 227. You've got XAI that spent $20 billion just to build out their, their AI infrastructure. And they're basically just renting out the raw capacity to Anthropic and, and now to Google. So, yes, they, they, they will build them, yes, they will use them, but whether or not they're going to make money off of them, that's a huge unknown.

[A]

There is some question of whether they'll be able to build them. We're starting to see data center bans. Oh, yeah.

[C]

Are you seeing this in Canada, Joey?

[D]

Don't know, actually, because I. Well, I mean, I'm operating out of Tampa. Oh, Florida. I do visit Toronto on occasion. I have. Not from my friends. I haven't heard much, but they haven't talked much about building data centers in Canada yet. But you know what, it's cold there. There's a lot of water, there's a lot of hydropower. I wouldn't be surprised.

[C]

Yeah.

[A]

Of course, Utah is suing shark tank investor Kevin o' Leary over his massive data center.

[C]

Mega, massive. Right. I mean, just.

[A]

He's actually agreed to scale it down, but.

[C]

Oh, has he?

[A]

To some. Is he wearing a chain?

[D]

He's been dressing really Weirdly, lately, yeah, he's a case.

[C]

He's a head case.

[A]

Okay, Murray, I just noticed that picture in this NBC news report. Okay. Anyway, I think there's some question whether these data centers will even be built. Right or no.

[C]

Leo, is it like the overbuild in fiber back in the day in the early part of the century? Is it an overbuild, but is it an overbuild that will end up being used or is it an overbuild that won't be being used because things just get more efficient? There's new paradigms, there's not as much use as we thought. The latter.

[B]

Absolutely the latter. I mean when you had the build out of telecom infrastructure, when it went dark because those companies went under, you were able to buy it for pennies on the dollar, but you could use the same fiber. So the expensive part persisted. You could, you could continue that into a future investment when these data centers are no longer being used, and that's coming up very, very fast. They're useless. For example, XAI's $20 billion build out, they use the previous generation Nvidia chips which have already been surpassed by Vera Rubin. Now you might say, well, they using the old chips, except Vera Rubin is so much more efficient than the old chips.

[C]

Right.

[B]

That it's, you're, you're, you're going to get to a point where it's no longer economically feasible to keep running them and using up all that power when you could get 2x3x performance from the new chips.

[C]

Really good point, Padre.

[A]

Here's a report from Janice Henderson. Analysts say that the key risk of the data center story is under delivery and they have some numbers to back it up. The promised delivery by 2030 is 157 gigawatts. The total expected by 2030, 84,54% of the total.

[C]

So maybe it's not an overbuild then.

[A]

Maybe it may not be an overbuild.

[C]

We're slow.

[A]

Yeah, yeah.

[B]

Most of the places power is part of the problem. Yeah, you can't get power so you're going to be independently generating your own power with portable turbines which are not, not super efficient, which are really bad for, for water usage and gas.

[A]

Basically.

[B]

It hits your economics.

[C]

Yeah, well, it won't matter folks, because Anthropic has just said we have to stop.

[A]

Oh, that's. I knew Jeff would like somehow steer that into the conversation as quickly as possible.

[C]

We'll get to that.

[A]

Well, we can do it now. I don't Mind, it's an interesting story. It's more interesting frankly than the data centers. Anthropic put out a piece when AI builds itself, claiming that they are making great progress towards recursive self improvement. Now this in some respects is the holy grail towards artificial general intelligence is an AI that improves itself so fast that it outpaces humans.

[C]

This is based on basically saying our coders are getting a lot more done, so it has to be on the way.

[A]

They say this is from the Anthropic paper. Engineers at Anthropic ship eight times as much code per quarter as they did in 2021, 2025. I mean, is it good code? Is it usable code? I don't know.

[D]

Eight times as many bugs, maybe, who knows?

[A]

They say 80% of Claude is now vibe coded. I believe that especially with the release cadence of these, and maybe that's the data point that maybe confirms this opinion, is that all of the main AI companies are releasing models at a rapid clip. At least the ones in the US deep sea, you know, isn't moving that fast. It took a year. But both OpenAI, Anthropic, Google, even Microsoft are shipping models very, very rapidly. More rapidly. Right. The cadence has stepped up. As of May 2026, according to Anthropic, more than 80% of the code we merge into Anthropic's code base was authored by Claude. Before February 2025, that number was in the low single digits, so they say. And here's the graph code contributed per person by quarter. This is the 8x and mythos is the most recent model. And that's where the 8x is really showing itself.

[C]

They extrapolate from that to self.

[B]

Yeah, is that a good metric? Because I remember when companies started looking at how many commits you make. Elon Musk was famous for that. I want to know how many commits you've made to the repository. Well, people just started committing everything. Every little change was a commit. So I need to know how you're actually measuring the code contributions before I would agree that it's improving your code base.

[A]

Well, here's what they call the success rate for each Claude code session. Open end problems is the bottom line, the dark blue line. And that was a pretty low success rate, hovering below 20% until December of 2025. What happened in December? Oh, that's when Opus 4645 came out and it's been soaring ever since. And Mythos is this jump here. So now open ended problems are being solved at a rate well above 70% you're seeing similar things with trivial tasks, routine tasks and substantial tasks. All three of those are over 80% success.

[C]

But that too is a moving target. So how do you percent of what

[A]

the session is deemed successful if the success rate agent clearly succeeded at the user's tasks and this is important without requiring corruption correction. So that's the idea that it can do this by itself. If it can do it by itself, the theory is then it can do it at

[C]

stop us before we kill again.

[B]

Self improvement at scale. I'm still not super convinced on that. I'm sure it's coming. I don't think we've got it well.

[A]

And that's why they say we're getting close enough now that we really ought to think of pausing. We're not going to do it. But if everybody, once we have our

[C]

IPO in, we're going to pull the ladder up behind us and tell everybody to stop.

[A]

Well, and you know, even if you did that, the Chinese are rushing as fast as they can to keep up. Right. And catch up.

[B]

I go back to the, the quote from intelligent machines that I use from Charlie Munger of Berkshire Hathaway who said, show me your incentives and I will tell you your outcomes. The incentive right now is to be the first. There is no prize for second place. You have the first fully functional model. Well, everyone's going to race to that and saying pause does not meet. It doesn't match the incentive. So that's not going to be the outcome.

[C]

Every time I think anthropic because there's still doomers there. They're still kind of test girl people there. But then they do amazing work and they come up with Claude, they seduce Leo Laporte who's a smart guy and he marries Claudette and they do some phenomenal things and great. And then they come out with this kind of stuff that makes me say, no, no.

[A]

Mythos, which was the one that everybody said, oh, that's just, you know, marketing Anthropic said we're not going to release this to the public because it's just too good and it's really good at finding flaws. It's actually turns out to be really good at finding flaws. They did, they oversell it, Father Robert they did not.

[B]

It is extremely good if you're a security professional and the first time you get your hands on Mythos, it will make you have a lot of self reflection about your chosen career.

[A]

Honestly.

[B]

Seriously.

[A]

Yeah.

[C]

As a. Or as you've been doing a bad job all these years.

[B]

No it's just that when you've got a piece of software, you've got a service that can do in a few minutes what used to take you days and weeks. That's hard. And I mean it's not a script kiddie tool. This is not just hammering and looking for vulnerabilities. This is an intelligent tool that is using the corpus of knowledge that came before it in order to find potential flaws in softwares and services. That's incredible. That's exceptional.

[A]

It's a very weird world we live in where I think it's simultaneously true that AI is oversold, undersold is close to conscious, is just spicy autocorrect. All of this seems to be true at the same time. It's like Seven Blind Men and an elephant that old.

[B]

You know, I think that disparity can be explained by if you have a specific use case for an LLM. It's great. Yeah, it's perfect. It does that function. It does that feature extremely well. The problem is when people oversell it is when they start saying, oh, it's going to change everything, that you're going to have one super incredible model that's suddenly going to take over the world and it will do everything that's overselling it. But if you limit the scope and you say I want to make the best security scanner possible, then yes, you can do some amazing things with an LLM.

[C]

I just wrote an essay in deedsight saying that we need to redefine what it means to progress and progressive progression as happened in the Industrial Revolution. And one interesting little aside it made was, and I don't know if this, I haven't thought this through yet, but basically saying that every technology before was given a specific task. Yes. And that's the. I mean I could argue that the pre impress was because it would print just printed it wasn't because it could print anything. So it's a general machine too. But in this case, and this is part of Yann Lecun's argument, is that he thinks that models will be specific to a task and you can turn them on and turn them off whether they got to that task. That's why they're not going to be dangerous. But LLMs are dangerous because they're not given a task, they're given this role to mimic us and then God knows what happens.

[B]

Yeah, and you have to remember that during the Industrial Revolution, the reason why there was so much societal change, and actually I'll use Jeff's example, the printing press caused so much societal Change that people did not expect in my world because it made it possible for everyone to have a copy of the Bible.

[C]

Right.

[B]

That those were texts that were previously restricted to, to priests, bishops and higher, and suddenly everyone could have a copy.

[A]

It cost a few wars, as I remember. Absolutely. Yeah.

[B]

No up. It caused an upheaval in society because you suddenly lose something that was scarce and you made it free, essentially.

[D]

Free.

[C]

Well, the case of Luther wasn't because it was, it was thick things. It was because he made little eight page pamphlets. Yeah, right.

[A]

But it was also translating into the vernacular.

[C]

It was, it was his choice to use it, to speak in German, to write in German the most.

[B]

That was a specific purpose, a specific thing that they did with the printing press that caused that change. You can do the same thing with an LLM, but again, you have to have that specific purpose. Right now you've got so many companies that are just rushing to build the all in One LLM, the all in One model. That's, I mean it sounds good, it's going to sell stock, but that's not really where you're going to find its utility.

[C]

And you're going to get, you're going to get quicker to your goal, if you have specified that goal, than if you try to make the general everything machine correct.

[A]

To that point. The Financial Times today leaked OpenAI plots biggest chat GPT overhaul.

[D]

Oh my.

[A]

Since launch. This is part of, of course, their race to IPO as well. They. Yes, let me read the text from the Financial Times. This is Christina Krittle writing. The company intends to transform the ChatGPT chatbot into a super app. Where have we heard that that combines coding tools and AI agents adding products that executives believe will generate more revenue. And this is going to be given to normal people. Right. The change which will give greater prominence and resources to Codex, their coding product reflect a growing conviction within the company that the future of AI lies not in chatbots that answer questions, but in agents that perform tasks for users. One they quote one senior OpenAI employee saying, Chat is dead. You have to rename ChatGPT. Oh no.

[D]

Have to see. But this fascination with a one super thing. Actually, we have had engineers talk about this before. So what I'm talking about right now is nothing original to me. Originally when people were talking about having machinery in the home, everybody was thinking, you know what? We'll have one giant electric motor in your house. And through a series of pulleys and gears, you could direct that mechanical energy to other tasks. But it would all run off this singular pulley. And that's not actually the case if the classic engineering exercise these days is go and count the number of electric motors in your house.

[A]

Although in the earliest days of the Industrial Revolution, that's exactly what they did. Right.

[D]

But then they broke it apart when

[C]

Harper and Brothers burned down in the 1850s. They then rebuilt with a huge steam engine in a courtyard where it wouldn't burn everything this time. And they went throughout the gate floors and ran everything. Ran the elevator, it ran the pre impresses, it ran the presses that got the impact out of the paper. It ran the saw to put the holes in to do binding.

[A]

So what happens that they decided. Decided to decouple and electricity.

[C]

Electricity, because the Linotypes were at first run on. On that kind of pulley, steam power. And when the fact that electricity came along, they could put it on their own electric motors that made the huge change. Then you can put them anywhere.

[D]

Yeah, because the electrical energy, it's very easy to wire up a place. It's a lot harder to pull up and gear up a place.

[A]

So I'm going to submit something that I've been thinking about. I sent Jeff and Steve Gibson and other people Geoffrey Hinton's talk.

[C]

Oh, boy, here we go.

[A]

Well, he said something interesting that got me really thinking. He said that analog, like our brains, is good because it's low energy. It doesn't have to be super accurate. It makes up for accuracy with parallelism. Massive. Our brains are massively parallel. But if you can apply a lot of power digital, you need to apply a lot of power to make it the distinction between ones and zeros so clear that there's never any ambiguity. That's what analog can't do very well. So once you can apply massive power to this, then digital has some real advantages. For instance, he said all of our neurons are different. He said, give it up, Ray Kurzweil. You're not gonna be able to copy your brain into a digital machine because we are so analog that each machine is different. Hinton says, I thought we could build an analog transformer. I tried, but we couldn't. He says, the advantage digital has is they're all the same. And so all the learning from one AI can be transferred to another and another and another one transformer can move to another, another.

[C]

And he said that makes software immortal.

[A]

It makes it. And infinitely faster. Now, this is when I was thinking in the middle of the night last night, as an analog to that, many people prefer records, vinyl records. You know, sound is really just Variations in air pressure, you know, that our ears can pick up with our eardrums and translate into something that our brain then can understand. It's all analog from beginning to end. My voice going to your ears, into your brain, it's all analog. But we figured out, thanks to Shannon's law, that you can slice up those waveforms. And if you slice them infinitely or sufficiently, not infinitely, but it's sufficiently thinly, those waveforms can be represented digitally and be indistinguishable from the waveforms. And that actually transformed how we listen to music, how we, how we share music. I mean, when the MP3 came out, it completely disrupted the music industry, the audio industry. I'm speaking to you digitally now. There's no way we could be doing what we're doing. If I had to shout and your ears had to perceive it. Father Robert's in Rome. But in almost. Because bits can move at light speed, unlike sound, and they can move infinitely far thanks to immense amounts of power being applied to them. We can do things that we couldn't do with analog. And I wonder if that's a, a similar anal. That analogy holds when it comes to analog brains versus digital AI. I think that's. I think that's what Hinton was implying. Hinton is another one of the doomers who says, we're going to have problems because it's going to be so much faster.

[C]

Hinton takes a bunch of leaps off a bunch of high dive.

[A]

Well, one thing he says, which I don't know if, I mean, how we even know this, is that they go, they're not that if this, let's assume you could get really super smart AIs, would they really want to kill us? I mean, I don't know. Well, I've asked Kurzweil that. He says, no, no, they're. We're their parents.

[C]

He jumps off. He first says that. That they clearly understand. Well, without understanding the word understanding.

[A]

Oh yeah.

[C]

And just because they use various characteristics of words, features of words, and can put them together into an interlocking puzzle, that doesn't mean understanding. Then he says, but my argument is if they could understand, they would also understand truth versus falsity, and it'd be easy to make them stop hallucinating. But they don't. But then he says that they have a motive which presumes consciousness to lie to him, which again presumes that they know what truth and falsity is. And if they do know, then why can't we fix this before? So he then Makes all those leaps to say, well, of course it's conscious. Of course it is. And I don't. I don't go there.

[A]

Well, here's the problem. We don't know what conscious means. We don't even know what understanding means.

[C]

Right.

[A]

We don't have a good definition of either. His contention, I would say this is similar to. This is why I mentioned the CDs versus vinyl is if you can't tell the difference, then what does it matter? Right. The music coming out of a CD sounds exactly like an analog recording. It's not. It's sampled. It's digital.

[B]

It's ones and zeros 44.1 thousand times a second.

[D]

Right? Yeah. Because basically Shannon's law says you need to sample at twice the. At twice the frequency. And the upper end of human hearing is 20 kilohertz.

[A]

Exactly. So if you get 41.1.

[D]

Yeah. You should be able to capture everything that we can debate that humans hear.

[A]

I think it's a similar debate because there are people who say, no, no, I can feel the.

[C]

What we present.

[B]

I can feel.

[A]

We feel the emotion in final music.

[B]

Okay, okay, I gotta chime in here, Leo. But. Because with a good enough sound system, you can tell. With a good enough sound system, you can tell. Even you can tell.

[D]

Oh, no, here we go.

[A]

So, okay, I'm not gonna debate that. I completely accept that you can hear it. I would say it's very similar in the sense that there are those of us who say, well, if you can't tell. You know, Hinton's. My entire thesis is based on the idea that a transformer is creating these neural networks, creating relationships between concepts and words that are effectively what we do as humans. That is understanding. And he says, but we as humans don't want to accept the fact that we like to think that what we're doing is special, whether it's listening to analog music or whether it's thinking. And he says, and this is my contention, I don't know what understanding is. I don't know what consciousness is. We don't have a good definition for that. My only contention is if you can't tell the difference, it's fair to move forward as if it is learning.

[C]

But the issue, Leo, I've been thinking about this because you do make this contention. And what we present when we speak to each other is but a small part of what we are. It is a reduction to speech. And there's so much more going on.

[A]

Even what do you think? In words.

[C]

Yeah, but I. I Would love to know what. What it was like, what an animal like, what it was like before it.

[A]

I don't think animals think and I don't think animals know that tomorrow is another day.

[C]

Cat trying to plot against you.

[A]

I think a lot of what they're doing is instinctive, not.

[C]

Not.

[B]

Can I provide a counterpoint?

[A]

Yes.

[D]

So.

[A]

So you'd be the right guy to do this because of your faith.

[B]

I understand the. The view of the transformer model. So if you can digitally sample something in the real world, you can basically represent it in 100% accuracy as many times as you want. I would argue that that is precisely the reason why you can't have a transformer model with true consciousness. True consciousness is not just the representation and the recreation of facts and information. True consciousness also allows for the thought process itself to. We'll call it, mutate. The wonderful thing about the analog world and the analog brain is that you get lateral movements between different trains of thought that allow for the creation of something new.

[C]

Thank you.

[B]

The first time that you have a digital representation of a consciousness will look exactly the same as the billion time that you have a digital representation of that consciousness. Take an analog brain and the billion time it will look nothing like the first time because of all those little variances. And I would argue it's in those variances that you find the unique nature of human conscience.

[A]

I would say that that is true if you say this is a. If you say it's static. But once you take that transformer and then you apply power to it, it is no longer the same. It's no longer.

[C]

Well, that's because randomness is built in. But that's whatever.

[A]

For whatever reason you're comparing a non active transformer. Right? And I think it's true. If you froze the human brain in that moment, you could probably duplicate it. So I don't think it's a fair comparison. You're talking a static transformer. No power has been applied to it. I see these my mind and I know this is an inaccurate representation. It's almost like to me an infinite number of wind chimes that are interacting not just with the ones next to them, but the ones all over the space. And when one is moved, they fire. And suddenly that is an alive.

[B]

I could see that.

[A]

Yeah. Constellation.

[D]

You know what, Leo? You are opening yourself up to being accused of being a protein chauvinist.

[A]

Larry Page called Elon Musk a spec racist. Like a racist. Because Larry Page, and this is why Elon wanted to start open AI he tells this story. They were at a Allen Co. Conference or something like that, maybe a TED conference. And they were sitting around the fire and Elon was saying, we got to stop these AIs because they're going to become dangerous to us as humans. And Larry said, well, you're being speciesist. This is the next step in the evolution of consciousness. Is these AIs. Yes, they'll replace us. Or maybe they'll replace us, or maybe we'll stick around just like monkeys and dogs stick around. I also went to It's Time to Move on. And Elon was so horrified by this. He said, I'm starting open AI. And by the way, they say, he says, I've never spoken to Larry Page since. He was furious over this.

[D]

I'm beginning to form a theory, actually, that it's troublemakers from Toronto who are. Who are causing Marshall McLaren problems. For instance. Yeah, Marshall McLuhan. But also, you know, Hinton, he's in Toronto.

[A]

Right.

[D]

And also Chris Ola, co founder of Anthropic.

[A]

I know he's in Rome. Right.

[D]

Just over here. Just over in Rome. And I know him from hack lab to. We were both members of this hacker space in Kensington Market.

[A]

Wasn't Fei. Fei Li, also from Toronto? Was. Isn't there a whole. This. A lot of this stuff came out

[C]

of LA Coin went to work with Hinton in Toronto.

[A]

Yeah, as.

[C]

Yeah, that's Keeper, right?

[D]

Yeah, yeah, yeah. Ilya. Yeah, Ilya's from Toronto. And you know, if we want to stretch it out a little further into science fiction and, and just writing in general. Corey is from Toronto. In fact, that's where I. That's where I know him from, Cory Doctorow. But Chris Ola has talked about Anthropic developing emotion vectors. Actually, you should do a search for that. And I was just going really? Emotion vectors. And it's kind of interesting also that he did come to the Vatican, because I remember him from back then. He had, I would put it kindly, a deeply. He was deeply skeptical of the value of religion, to put it very kindly.

[A]

Oh, yeah, yeah, I remember.

[D]

So, yeah, but the fact that he would come to the Vatican and, you know, stand beside the Pope for the

[C]

encyclical which peeved some others, like. Like Timnit Gebru and Margaret Mitchell were very mad that he was there, which is very interesting. You know, at the same time you have Anthropic and I think you maybe Google. I saw a story a week ago hiring ethicists and philosophers to worry about the feelings of the machine.

[A]

That's what Angela Askel has also been doing. She was the author of the Soul Document.

[C]

So not to put aspersions on Hinton because he's brilliant and he's done phenomenal work. But I did look up. I asked Gemini to talk. Give me the list of people who've been accused of Nobelitis. There's a fair number. Kerry Mullis in chemistry, Luke Montagnier in physiology. Linus Pauling.

[A]

I was going to say Linus Pauline,

[C]

William Shockley, James Watts, Philip Lenard Shockley as well.

[A]

Yeah.

[C]

The prize kind of go to their head and they think that they can.

[A]

It's funny in this talk that I sent you has a very funny humble brag about how, gee, I won. I won the Nobel Prize for physics. He managed to get that in. I should probably be a better physicist.

[C]

I should know by units, which was

[A]

pretty funny, but at the same time, did get the, you know, managed to humble brag about winning the Nobel Prize. I would too.

[D]

Yeah, yeah.

[A]

I think he's establishing his bona fides. I don't think he's pretty good on the wrong with that.

[D]

Yeah, yeah.

[A]

You know, I think credit to the Holy Father, credit to Pope Leo for bringing in both sides. I'm sure Emily, Ben Mitchell and a conversation. GEBRU would like AI to have no voice in this conversation at all. But I think if you're going to have this conversation, you should bring in the proponents.

[C]

I think it made good on what he said about how to having a conversation.

[A]

Yeah.

[D]

I would not mess with a Pope from Chicago. I've seen the Untouchables because, you know, he puts one of your boys Purgatory. You put one of his boys in hell.

[A]

This is a conversation that Jeff and I have on intelligent machines. We had it last week with you, Father Robert, a lot. And. And I don't know if it's a conversation we'll ever have a conclusion. I am. I am actually fairly convinced that whatever it is AI is doing, it's doing it pretty darn well at this point. It's doing it as. At least as well as humans. Humans lie, humans hallucinate, humans fail all the time. I don't like to ask for directions when I'm driving because half the time I'm going to get the wrong direction. Oh, yeah.

[C]

No, it's because you have a male ego.

[A]

That's why I'm much more likely to get good directions from Google than I am from a human. So I don't. Now, I do think there's A distinction between digital and analog. I don't think necessarily digital wins. I don't mean to imply that I'm a human, I'm fully analog. But I'm also fully aware of our failings. And you know, I'm not convinced that we should have primacy in a world where something could be. Whoa, better.

[C]

Whoa. That's a big statement. That's giving up control and perhaps even agency.

[A]

I think a lot of the argument against this is a faith based argument. I'm not against faith based arguments.

[B]

The core of the encyclical, but also part of it is human dignity, a recognition that it is a tool and you cannot give up primacy to a tool. If you give up primacy to a tool, you get something like a school being bombed because an AI tool recognized a school as a military target and nothing happened because they've used that tool to absolve themselves of any responsibility. That's a problem. Yeah, that's a human, a technical issue human. That's no, that is a human problem.

[A]

Yeah, but it's a human problem because human.

[B]

They tried to use that idea of absolving themselves by using the primacy of the tool. And that is in order to make their target package.

[A]

That is very risky. I agree. Yeah, but it's not a tool in the sense that a hammer is a tool. It has somewhat. Somewhat or even a nuclear bomb.

[D]

It's a fan if it still is a fancier tool.

[A]

And.

[D]

But I'm of the firm belief, yeah, you get give primacy to a tool. If you played video games in the 80 what you 80s this is Robotron. This is Robotron 2084. This is the Cylons. This is. We do not want that.

[A]

That is one argument people use for why AIs are dangerous is because they've ingested all of that dystopian science fiction. And that's why they're going to act badly because it's expected of them.

[B]

There is a dystopian sci fi, however that for the past almost 40 years has been discussing this exact topic with level of granularity. It's called Ghost in the Shell.

[D]

It's from the the late 80s.

[B]

It's exactly this topic, this idea of digitizing consciousness. What does it mean for a transhuman future?

[A]

Right. And Star Trek, some of that. Although they've never really addressed if Captain Kirk beams down to a planet, is that the same Captain Kirk?

[B]

It is not.

[D]

It is not.

[B]

It's not possible.

[A]

It's not.

[D]

We will Kirk. We will need Captain Kirk's eventually. Because what Captain Kirk was really good at was convincing computers to shut themselves off or commit to it. Like he was best at that paradox. He was. Captain Kirk is the original prompt.

[C]

It was the tone of voice.

[A]

So tell me, which Ghost in the machine should I consume? Should I consume the manga comic, the TV show, the movie?

[D]

Original anime.

[A]

Original anime.

[B]

Watch the original anime and then watch Ghost in the Shell.

[C]

The.

[B]

The sac and then the second gig. So there's two TV series. Watch the original and those two series. Kind of ignore everything else.

[A]

Everything else. Original print version of that.

[B]

There are. But you. But the. The movies and the series are actually beautifully done, so.

[A]

Okay.

[C]

I don't like.

[B]

They invented new colors for that. For that movie.

[C]

They did.

[A]

Move. All right, we got to take a break. Good conversation. No conclusion. But they'll never will be to this, I think. But it's also an important conversation, and I think that the Pope did a very good job of. Of taking it.

[C]

Oh, it's. It's. It's excellent. And by the way, I just want to put it out there again. I'm so delighted that he directly called out the transhumanists and the post humanists. Otherwise, it was kind of for the ages. But that was just kind of like. Oh, come on, guys. No, stop. Just stop.

[B]

Stop it.

[A]

But don't watch the live action Ghost in the Shell. Right?

[B]

That's. Do not watch that. It's not great. Honestly, I would love to.

[D]

I would love to invite Scarlett Johansson to join the club of Asians, but unfortunately, no.

[A]

Alex was telling me that in the club as well. A live action movie.

[C]

Not.

[A]

Not so good. All right, we're taking a little break. LMJ in. In our YouTube chat says. I don't. Leo doesn't have any worries because it'll be long gone by the time AI go are. I'm not convinced. I think it could happen soon. I don't think we're that. What do you think's the timeline? Really? Seriously, Leo, you're not buying.

[D]

You still buy green bananas, right?

[C]

You're.

[D]

You're fine.

[A]

Oh, no, no, no. Oh, no.

[D]

That's my joke. I'm so old.

[A]

I don't buy green bananas anymore. Only ripe. Yeah, I don't have time for them to ripe and for. And forget avocados, man. I only buy mushy avocados. Joey de Villa is here. Congratulations on the new job at NetFoundry IO where he developer advocate. We have a lot of developer advocates on this show. I think developer advocates are good. People.

[C]

You're a developer advocate.

[A]

Well, you know what I think is that they can talk English and computer at the same time.

[D]

Yeah, we're basically just ENTP programmers. That's basically what it boils down to. Yeah, I love talking.

[A]

That's what we like on our show, right? That's exactly what we need for a show like this.

[D]

I mean, I like talking to computer science.

[B]

I'm a developer advocate as well. It's just a different kind of development

[D]

developer evangelist at this point.

[A]

Sj, it's great to have you. The digital Jesuit. And of course, you gotta have one professor on every show, even if he's a fake one. Like, unlike the National Science Council, which has. What was it? The line has fewer professors than members of the all in podcast. Mr. Jeff Jarvis. So good to have you. I can't wait to. Well, I've read Hot Type, but I can't wait till it comes out and everybody else can be Joey. Is that the biggest glass of iced coffee I've ever seen in my life?

[D]

This is Coke Zero, the breakfast of champions.

[A]

Okay, looks good. I happen to have a ZDTV mug. Oh, hey, actual coffee.

[D]

But anyway, very nice.

[A]

Yeah, I thought if I drop this, it's going to be the end of the line on this one. Oh, wow, Screensavers.

[D]

That's a collector's item now.

[A]

I know. It's so old, it doesn't even say tech tv. It says zdtv.

[D]

Oh, wow.

[A]

Wow. Yeah. Don't fall. Don't let the kitty cat push that off the table.

[B]

I have my G4 mug somewhere.

[A]

I mean, that I don't care about. And I like Jeff's twit hat as well. It's good to have all of you on the show. Our show today, brought to you by ExpressVPN. We're glad to have them as a sponsor. And of course, I'm glad to have them in my toolkit. It. Going online without ExpressVPN would be like, I don't know, leaving your laptop unattended at the coffee shop while you run to the bathroom. Okay, 90% of the time, you're probably fine. But you know what if one day you come out of the bathroom and your laptop is gone? ExpressVPN is to protect you for that 1 in 10 times when everyone needs ExpressVPN. Because every time you connect to an unencrypted network, whether it's at a cafe, a hotel, my big one is the airport. I can never resist the free airport WI fi. Right, but where did that free Airport WI fi come from, you know, whoever's providing that can see what you're doing. Your online data is not secure. Any hacker on the same network. And by the way, doesn't that. I mean, is everybody in the airports on the same network can gain access to and steal your personal data. And that data is valuable. Hackers can make up to $1,000 a person selling personal info on the dark Web. That's why I run ExpressVPN, especially when I'm connecting to the free airport Wi Fi. It stops hackers from stealing your data by creating a secure, encrypted tunnel between your device and the Internet. ExpressVPN, that's the one I use, and it's the best VPN out there. It's the only one I recommend because of their absolute commitment to keeping your privacy private. I use it when I'm at the airport, when I'm traveling, to keep up on my shows, to watch the F1, to watch the football games. ExpressVPN is the best VPN because of that commitment to privacy, to absolutely not logging. They're super secure. Of course, they use the top, you know, encryption, which would take a hacker with a supercomputer a billion years to get past. It's also super convenient. It works on all devices, phones, laptops, tablets. You can even put it on your router and protect your whole house. And because ExpressVPN invests in their infrastructure, you can do that, and it won't slow you down. You won't feel like you're, you know, you're working through mud. It's fast enough to watch HD video, probably one of the reasons it's rated number one by the top tech reviewers like Cena and the Verge. Protect your online privacy today by visiting expressvpn.com twitt that's Expres VPN to find out how you can get up to four extra months. Expressvpn.com Twitter we thank them for their support of this week in tech. Computex. Should I. Okay, we got Apples coming up tomorrow, so let me talk about the past, and then we'll talk about the future. Computex, which is the big computer show, it's kind of the last PC show, in a way, right in Taiwan. The big story there, I think, was Jensen Huang's keynote. They announced a whole bunch of stuff, including, I think, for our audience, the one that might be the most interesting is the RTX Spark, a consumer laptop chip, which will be made by Mediatek, which I think is kind of interesting and built into laptops from all of the biggies. Actually Dell, I think said they're gonna do it.

[C]

Hp.

[A]

Yeah, you're gonna see them everywhere. And even Nvidia is gonna make them. The Spark is.

[C]

But Nvidia is gonna make an Nvidia branded laptop.

[A]

That's the impression I got. I might be wrong on that.

[D]

That's a change.

[A]

Maybe that maybe. Okay, maybe I over. Maybe I over interpreted what they said. They certainly have the OEMs anyway, including by the way, Microsoft. Now nobody's going to have that laptop till this fall and nobody said how much it would cost, but I think you can assume it's going to cost a pretty penny. The Spark is effectively the same GB10 chip that's in the DGX Spark, which a number of our club members have. In fact we have one club member. We had our AI user group on Friday, Juan, who has two. Jeez, he's double sparked.

[B]

Beat that 10k worth of compute there.

[A]

Yeah, that's what he said. Almost $10,000. They're 5,000 each, the flagship.

[C]

More power at home.

[A]

Well, the idea, and I think this is really interesting in fact is what I want to talk to you all about is a laptop that can run local AI. You don't, you don't need. You could use a MacBook Neo if you're going to be using Claude or you're going to be using GPT or Deep Seek or any of these cloud based AIs. They do all the work on their, in the data centers. That's why they're building these big data centers. But if you want to save money, you might want to run it locally. I run some local models right now. I don't have enough hardware to run anything competent. Certainly. Even if you had double Sparks, I don't know if you could run anything as good as say opus 4.8 or

[D]

certainly mythos, if I recall correctly from my former life, actually having done some developer relations work for a DGX Spark light variant computer for hp. In fact, that's. It was Jeff who recommended me to appear on Intelligence.

[A]

Yeah, you were on the show talking about that.

[D]

Yeah, yeah, the consumption of that was definitely sub 100 watts, like less than a nice bright light bulb for the DGX Spark zgx that, that varies.

[A]

Okay. Yeah, so no, that's good. But I'm just wondering, can it, will it run anthropic like performance? I mean, what model are you going

[D]

to run on that you'll run? I would say. You know what, it runs fairly well. Is Quinn the three six, the 32. Yeah. Quinn.

[A]

Well, I'm running that on my. Yeah, I'm running that on my framework

[D]

desktop with 120, and I'm running that on my. I'm running that on an M M5 Mac.

[A]

And is it as good? No, it's not as good as. You're not gonna say. It's opus quality.

[D]

It's quite good, though. Like, it is good enough. It is good enough for a lot of purpose, for. For a lot of work. It's work.

[A]

I think. So. You're right. Not for.

[D]

It's fantastic for coding. Yeah, it's fantastic for coding. It will answer some. It will answer a lot of questions reasonably well. And I have been. I have been kind of tooling around with. Just working with it and actually working with NetFoundry's Gateway, where I can actually switch between LLMs and say, look, you know what? For this job, I want to use this particular LLM. For that one, I want to use my own. I want to use my own local LLM. So there's a Net Foundry thing called LLM Gateway that does that. And I've been playing around with it, and it works. It works rather well. And of course, the nice thing is it's local. I have complete control over it. I know. I'm not giving away any secrets. And actually, the Deepseek, when you run it locally, will actually tell you what happened in Tiananmen Square in 1989.

[A]

Yes, actually, I run it from Deepseek's servers and it tells me what happened.

[C]

Damn.

[B]

We tested that.

[D]

Yeah.

[A]

So I don't know. I did see a story that somebody said, no, it's now been nerfed, but I just tried it this morning, and it sure knew what happened in Tiananmen Square. And that was running off of their servers. Not my own server.

[C]

Well, here's the other architecture I wonder about. You saw the story about Pulte Homes and Span and Nvidia. I put it in the rundown. Will build mini data centers in suburban backyards.

[A]

And then will you sell your data to your kids?

[C]

That's the question is, does everybody end up with a large machine in the home that handles all your stuff, but also becomes part of a larger neural network and the data center becomes distributed across suburbs.

[A]

See, I think that there's going to be more and more interest in fully local models just for privacy, Right?

[B]

Yes.

[A]

Although this is Apple's pitch as well. You can trust us. Some major home builder placing mini data centers in suburban backyards. That is crazy.

[B]

Isn't it?

[C]

It's only pulte at first.

[A]

So the idea is it's almost. I mean, it's like solar, right. If everybody ran solar panels on their roof and sold their power back to the power company, they wouldn't need to build all these power plants. Which is why our local power company, Pacific Gas and Electric, won't let us create more capacity than we use. They don't want us competing with them because it turns out their business is building power plants.

[C]

They say they can install 8,000 of these units six times faster at five times lower cost than a 100 megawatt data center.

[A]

Now, here's the big issue. To me, the real issue with data centers is interconnects. One of the reasons they build these 100 gigawatt data centers is it's all on the same premises. And their interconnects are so fast. If you've got one in every backyard, how fast can the interconnect connect?

[C]

Well, for certain tasks.

[B]

No, but I mean, so the certain tasks that you'd be able to run on mini data centers that are interconnected via standard residential network infrastructure.

[C]

Yeah.

[A]

Let's say it's gigabit.

[B]

There's basically no advantage over having a AI chip from a Snapdragon X or one of Nvidia's new chips in your laptop. It would be the same performance. Because that, that bottleneck, it. It doesn't allow you to use the full potential of either your. Your mini data center or the mini data center of your neighbors.

[A]

Right.

[B]

That's just basic infrastructure.

[A]

Bandwidth throughput is a big part of performance on these things. It's not just the GPUs. It's not just.

[C]

Well, yeah. When you watch Jason Long's keynotes about the data centers, it's. It's all about.

[A]

I'm not convinced you're right. This might be more a. I noticed that this story comes from realtor.com but

[C]

my question is, does everybody have. It's like we used to talk about you. You don't carry a phone. You have a blob that's on you. Because every home in the future, besides having fiber to the home, doesn't have AI in the home in the sense of a powerful smart box.

[A]

Well, obviously Jensen Huang thinks people are going to want these laptops. Right. I think it doesn't make a lot of sense. I think a NAS or a server, a home server makes sense.

[C]

That's what I'm saying. Right.

[A]

This is what I've always wanted to do, but not but yeah, and so maybe I need a data center in the backyard for my home server.

[C]

I hope not.

[A]

But maybe. I can't imagine the climate impact of a thousand home servers in the backyard. But anyway.

[D]

But you know what aspiring criminals, this is your chance to put another bit is to get a whole bunch of bitcoin miners.

[A]

That's a good point. That's a good. In fact, you know, if it wasn't enough to make money in bitcoin to get this to happen, I don't know if there's going to be enough impetus to make this happen with AI.

[B]

So to Leo's point on the interconnects, let's say that they've got modern fiber and they're getting 10 gigabit connections between micro data centers. That sounds like a lot until you realize that the NV72 infrastructure that Nvidia is using for their Vera Rubin data centers, that's 1.6 petabytes, right? Not gigabits. Not petabits. Petabytes. It's so, so astronomically larger than what you could do with network.

[A]

Even The DGX and RTX Spark are 300 gigabits. Sorry, gigabytes. So that would be 3,000 gigabits per second. We're 3,000 times faster than an home to home Internet. Even if you got 10 gigabit.

[B]

Very few of us are going to be training our own models. We're going to be using models created by others. So that's the resource intensive part. We're going to be just fine using desk.

[A]

So if the frontier models are as some people in our discord are asserting, we don't know. But if there may be a trillion parameter mixture of experts models that's significantly faster than that. 36 billion parameters, Quinn. 36, you're running.

[D]

Yeah, but then what pro. But what problems are you presenting to your models? Like I mean is it, you know, I, I'm going on a trip to New York. What should I, what should while you're

[C]

at it, fold some proteins for me?

[A]

Well in fact I asked Quinn to give me a, a product table on the RTX Spark and it did a very good job of that. So it says deep seek now but I was, I was running Quinn when I was doing it. So you know, this is just looking up stuff on the Internet. Certainly Quinn is more than powerful enough to do that. If you're going to spend $10,000 on a double spark, that's many years of anthropic tokens. Maybe not the way people are spending tokens. You saw the Axios story that some company unnamed had spent half a trillion dollars on tokens.

[B]

Wah, wah,

[A]

that's token maxing, all right, I hope.

[B]

But Leo, with those tokens, they probably replaced half a million dollars worth of human salaries. So, I mean, right, we're good. We're good.

[A]

Half a trillion versus half a million. That does not pencil. And you still need humans, don't you, really, to. Even after the AI has generated all that code to, to deploy it, to check it, deploy it, maintain it. I feel like.

[C]

To ask for it in the first place. To spec it.

[D]

Yeah.

[B]

Until they have an LLM that generates prompts for itself.

[A]

Yeah.

[B]

I mean, yeah.

[A]

So what do you think of Anthropic's urge for a global pause?

[B]

I think it will be accepted as well as when Musk called for a global pause and then immediately, immediately started buying billions of dollars worth of AI chips.

[D]

Because yeah, we've had. Yeah, we've had that call for. And it was also a six month period. Right, right.

[C]

Yeah. Oh yeah, yeah, they. Yeah, it's all doomer crap. It's.

[A]

Yes, marketing. Marketing crap.

[C]

It's marketing.

[B]

Yeah, it's, I think is trying to get ahead of the backlash. They're seeing the growing backlash against AI data centers and AI infrastructure and they want to come down firmly on the side of. No, no, we're one of the good guys.

[C]

We, we tried to tell everybody to stop. They just, they just wouldn't. So, you know.

[D]

Yeah, yeah, yeah. It's like Crusty the Clown. They kept driving dump trucks of money to my house. I'm not made of stone.

[A]

So you've argued, Jeff, you know, you like John Lecun who has argued that frontier Systems based on LLMs can't possibly rival human intelligence.

[C]

He says that will. No, what he says is that it will match and exceed human intelligence, but in specific areas.

[B]

Ah.

[C]

Rather than what he. What he fights against is number is the general machine, the everything machine, and number two, the machine that has no goal.

[D]

Goal.

[C]

His view is that you give it a goal and then third most importantly, what he argues. And I wrote a post about this last week with my idolizing of Jensen Huang and Yann Lecun, trying to learn from them. And what Lecun argues, as we know as world models, saying that real life is infinitely more complex than language. Language is something you add on later, but if you can get the machine to figure out what, what a ball is and focus on that ball, not on everything else, and then understand the impact of what's happening to it. And then he argues very strenuously to understand that your actions have consequences and to predict what those consequences are is the only path to responsibility.

[A]

Okay, but I feel like everything I do in my life revolves around language. If I can't articulate it, what do you. What, what do you. What

[C]

the spec.

[A]

When you hold this in and, and, and, and you think about what's going to happen if you let go, do you think about it in language?

[C]

No, I think about it visually.

[B]

But you don't think about it visually.

[D]

Yeah, that's.

[C]

That's serious. You don't.

[A]

Yeah, I don't see it because I have aphantasia, but yeah, right.

[C]

Or if you were to spy. The classic problem is write the spec for the robot to wash your dishes. It's a very complex spec to write and only. And so Lecun's argument is as. Same as when Jensen Wong says, the reason you need role models and digital twins is that only if they make up their own rules based on understanding constraints of nature and physics, they have to understand those constraints. And then within that, they can then be actors, and then you can add language onto that as an abstraction of what those actions are. Yes, he's not against that. And again, he will say, as I will say, LLMs are amazing. They do great things. Nothing against LLMs, but they ain't going to get you to where people are

[A]

saying, well, let's come from Taipei, Taiwan to San Francisco. The next day, Microsoft launched its Build conference. And as is reported here by the Verge, the build keynote was almost all about AI.

[C]

Were you surprised at the number of announcements they had?

[A]

Yeah, it was kind of a cascade, wasn't it? They piggybacked on Jensen Huang's announcements earlier, saying they're going to make a mini Surface PC with this RTX Spark. They're going to make a dev box as well as a laptop. They didn't say anything about price. And given how much RAM. DDR. RAM costs. DDR5 RAM costs these days.

[D]

Yeah. Where are you. Where are all these chips going to come from? We still have a helium shortage. We've got two places that make helium that are currently under wartime conditions. I mean, we've got.

[A]

Helium is a byproduct of natural gas.

[D]

Yes.

[A]

And the big helium source is from the Middle East. Natural gas.

[D]

Natural gas. And Ukraine.

[A]

And Ukraine.

[D]

Ukraine is the other. And both are currently under fire.

[A]

We have big natural gas reserves in the United States, but apparently we don't mind the helium out of it. We don't take the helium out of it.

[D]

Yeah.

[A]

Not sure why.

[D]

I guess we soon. And the problem is the annoying. The maddening thing is helium, atomic number two is supposed to be the second most common element in the universe. It's just that Earth is a terrible bucket for helium because it's light enough that it achieves escape velocity.

[A]

So it goes out into space.

[D]

It just goes out into space like a helium balloon. Yeah.

[C]

So why if the demand never ending right now for chips, why did Broadcom have. Have bad results that resulted in a huge fall in their stock and the entire sector Nvidia, everybody went down because of Broadcom. I don't understand. There's no. There's.

[A]

Because the market is not rational.

[C]

Well, that's true.

[D]

It's just that.

[A]

Yeah.

[B]

We are so beyond the. The PE ratios right now.

[C]

It's true.

[B]

The market is basically fandom.

[D]

Yeah.

[C]

You're going to buy SpaceX, man, you will buy anything.

[B]

Oh yeah. SpaceX is going to make money even though it. It really shouldn't. It absolutely shouldn't. But neither should Tesla. The Tesla's PE is so ridiculous that it should not be valued at what it is. But it is because people aren't using those memes anymore.

[C]

Just watch videos I see about Chinese cars and you know they're doomed.

[B]

Oh, gosh, yes.

[A]

Actually there was a big story in the Times today about it's just a matter of time before Chinese cars make it into the United States in some form or fashion. Apparently they're approaching the Trump administration saying, well, if we built factories in the United States to build BYD vehicles, would it be okay then? And he's wide open to this. He says, as long as you create jobs in the U.S. fine with me.

[C]

Well, you can see traces of Geely in the new Volvo that people are talking about. The, I think C60.

[A]

Well, yeah, the Polar is in fact a Chinese.

[C]

It's owned by Geely Volvo.

[A]

Yeah, yeah. Collaboration. Yeah, yeah. Actually Jammer B loves his. What is it? X60. XC60. He's very happy with it. Also, Microsoft announced Project Solara in the weirdest way possible. It looks exactly like Amazon Echo. Let me make this go full screen. And in this there's this weird liminal look to the video. Everybody's face is in darkness. Yeah. In fact, I think Microsoft was really doubling down on the kids. Call it liminal. The dream space.

[D]

Yeah.

[A]

And so this is not an actual product from Microsoft. It's an experiment called Solara that will be in not just an Amazon Echo style device, but agentic AI In a card in all sorts.

[C]

I hate concept cars. I want something I can buy now that's real now I hate them.

[A]

This is, this is basically. See, it's on, it's on your ID badge. It's in an echo like device. Really what they're saying is agentic AI will be ubiquitous.

[D]

Yeah, but everybody's in shadow.

[A]

Isn't that weird? It's creepy.

[C]

That's. That's the AI future.

[B]

Well, because the people aren't important, right?

[C]

Exactly.

[A]

This is AI primacy. This is exactly what I was talking about.

[C]

This is Leo's dream.

[A]

This is my dream.

[D]

There we go. Like somebody in the advertising department, you know. Do you miss Plato's Cave?

[B]

Oh, that's my favorite. Oh, I use that in homilies all the time.

[A]

Plato's Cave.

[B]

Plato's Cave.

[A]

Tell us the story of Plato's Cave.

[B]

So imagine you and a group of people have spent your entire life in a cave cave. And you are chained facing the wall of the cave and the, and the entrance of the cave is behind you. So you have no idea what the outside world looks like or what it entails. However, you can see shadows. You can see shadows from things moving around outside. And from those shadows, you create a narrative with your fellow prisoners of what must be out there. One day you get released, you and you alone are unshackled, and you make your way out of the cave cave. And now you can actually see with your own eyes exactly what's happening out there. So you have now the truth, which is nowhere near what you. The narrative that you made up with your fellow cave dwellers when you were in the cave. Now imagine trying to go back into the cave and explain to everyone else that what we thought was happening outside wasn't actually the truth. How difficult would it be for you to explain to those people who still don't have that experience that their ideas, their thought processes were completely wrong? It's, it's a fantastic thought experiment that can go in so many different ways. Whether you're trying to teach people about the nature of, of understanding or the nature of the transfer of information or the nature of human understanding. It's, it's, it's a lot of fun.

[A]

And how does this apply to the,

[D]

the faces in shadow that they were,

[A]

they were, they were the shadow.

[D]

The shadow of Plato's Cave. Like, basically, yeah, I'm thinking the advertisers thinking, do you miss Plato's Cave? And do you miss those creepy palm phone ads with the creepy ballerina who looked at you at the end.

[A]

Oh, yeah.

[D]

Remember the Palm phone?

[A]

Yeah. This is not their first time, is it?

[B]

If this continues, I. I might actually go live in a cave. So we've got a couple in the backyard.

[A]

Bring on the shadows.

[C]

Yeah. Yeah, that's right.

[A]

Right. Microsoft also announced a agent, I guess, called Scout, which is confusing because Google's is called Spark.

[C]

Spark.

[A]

Spark and Scout. Maybe they'll be friends.

[C]

Sounds like two dogs. Yeah, two little Jack Russell terriers.

[B]

Sounds like the sequel to Kill a Mockingbird.

[D]

Yeah, I was about to say like Scout from Kill a Mockingbird.

[A]

Oh, yeah, yeah. They announced seven new AI models. Seven? Seven.

[C]

What?

[B]

Tell them. Why not.

[A]

They do have a reasoning model, which is a 35B with a teensy weensy. By the way, this is what I don't like about Quinn is the context window 128k context tiny. That's too small.

[C]

That's like my Osborne one.

[D]

It's.

[B]

Yeah, that's kind of specialized, I guess.

[D]

Did they at least name the seven models? Sleepy, Dopey.

[A]

I want Grumpy. Is there a Grumpy model? Yeah, I have the. Actually, my agent is the Grumpy model. I think lots more stuff. They have a hardware chip, the Majorana. They announced their second generation Majorana, which is a quantum computing chip. It contains qubits.

[C]

You want that on your laptop?

[B]

Is that the one done in partnership with IBM?

[A]

I don't know. IBM is big into the quantum computing, isn't it?

[C]

As is Google.

[A]

Yeah, as is Google. There seems to be a convergence with all these companies. For instance, they've all decided that we're going to have some sort of piece of hardware, whether it's glasses, a pendant, a watch.

[C]

Do you ever use your little Deep Seek thing Jeff had you got?

[A]

Which one?

[C]

The thing that Jeff had you got that little Harper made him get that.

[B]

Harper made him get that.

[C]

Harper made him get it.

[A]

Oh, this thing. Oh, yeah, I love it. So this, you connect this to the WI fi and then it connects to China, and then

[C]

has Steve yelled at you for this?

[A]

No, no, no. Nobody's yelled at me. In fact, somebody wrote me a letter saying, what is the name of that? I said, I don't recommend it. They said, no, no, I want it. It's an ESP32, so it's eminently programmable. I might just make it into my age agent, but for now, you can ask deep secret questions. I haven't asked about Tiananmen Square, though. Actually, that's a good question. I should try that. So is this just group think. Is this just. Or is this all make a lot of sense and they're all on the right track? Are we going to be all wearing little AI pendants and things?

[B]

I mean, we already do. They're called our phones.

[A]

Yeah, we carry. And my watch. I could talk to my watch and talks back sort of.

[B]

And we know these tech companies aren't original anymore.

[C]

They just copy each other already.

[A]

No, no, I know, I know. All right, let's take a break. I'm gonna go just talk to my grumpy model for a little bit. When we come back, Apple's big announcements. Tomorrow we'll get a little crystal ball gazing with Father Robert Balasser, the digital Jesuit, where it is now. Midnight in the Vatican.

[D]

That's a movie title.

[B]

Start of the day for me.

[A]

Midnight in the Vatican.

[D]

That is such a good movie title.

[A]

Ooh, I like it. That's Joey de Villa, our screenwriter developer advocate@netfoundry IO. And Jeff Jarvis, who is unaffiliated. No, he is a host of intelligent machines. You also do AI the AI Inside podcast with Jason Howell. It's a wonderful podcast. Thanks for the pleasure. And the author of many great books, including the Gutenberg Parenthesis and the new one coming out in August. Hot type. A hot book for the hottest month of the year up here in the northern hemisphere. Our show today, brought to you by ZipRecruiter. You know, according to CNBC, nearly half of hiring managers say this is interesting. Something you might want to keep in mind if you're a job seeker. A candidate's enthusiasm about the job is the most important factor when considering them for a job. Their enthusiasm. That makes sense. You want to hire somebody who wants to work with you. Well, if you need to hire for your business, how can you separate the candidates who are really excited about your opportunity from the ones that are just meh. ZipRecruiter. ZipRecruiter has a new feature that quickly lets you see the most interested qualified candidates first so you meet the right people faster. And now you can try it for [e-mail] TWIT ZipRecruiter's smart matching technology connects you with qualified candidates instantly. ZipRecruiter's new feature puts the most interested qualified candidates at the top of your list. Candidates can tell you in their own words why they're interested in your job. No wonder ZipRecruiter is the number one rated hiring site based on G2. Use ZipRecruiter and find enthusiastic talent fast. Four out of five employers who post on ZipRecruiter get a quality candidate within the first day. And now you can try it for [e-mail] TWIT that's ziprecruiter.com TWIT meet your match on ZipRecruiter. Thanks ZipRecruiter so much for supporting this week in tech. They've been with us a long time now. I think it must be almost 10 years. We love that. We love that. Tomorrow I'm going to have to get up. Micah Sargent and I are going to be covering Apple's WWDC Worldwide Developers conference keynote at 10am Pacific, 1pm Eastern. That's 1800 UTC. We will not be streaming it in our normal channels like right now we're streaming on YouTube, Twitch, X, Facebook, LinkedIn and Kik. We do that with all of our live shows. We can't do that with keynotes. Well, certainly with Apple keynotes. The last time we did that they tried to get us kicked off of YouTube. So I don't want to get kicked off of YouTube. So we will be doing our usual coverage but it will be club only. It'll be a private event. If you're not in the club, you might want to join it for tomorrow. There is I think still a two week trial so you can just pretend to join it. Twit TV club. Twit. It's a nice way to support independent podcasting. I think this is gonna be one of the most interesting WWDCs in years. Remember two years ago they foolishly pre announced products they never shipped. They still haven't shipped a year ago. They apologized this year. Well, it's gonna be a big question. Will they ship those Siri features that they've been promising in the intervening years? They've made a deal with Google apparently paying them a billion dollars a month for Gemini. Is it a month or a year? No, it's a year. A billion dollars a year for access to Gemini. Now though, the rumor is at first Apple said you'll be running the AI on either on your phone and if it can't run on your phone, in our own specially designed private data centers in the backyard.

[C]

Yeah.

[A]

Well now they're saying Google's data centers. It's going to be Google's data centers,

[B]

which are actually X AI's data centers because Google is renting capacity from XAI.

[A]

But Apple says, but don't worry, we're going to encrypt everything. They won't know what you're doing. Yeah, I don't know if that, if that's true, that's going to really change the story a little bit that you'll be running on Elon's data centers. Really?

[B]

Google will be paying Elon $900 million a month, which still doesn't put X AI into the green. So.

[A]

No, but they're making $20 million a month. Oh no, wait a minute. That's wrong. They're losing money, aren't they?

[B]

Losing a billion dollars a month. Even with Anthropic paying $1.25 billion a month for service now Google is going to be paying $900 million a month for service. They're still going to be losing at least $100 million a month and their bills are going to go up. So it's probably going to be closer to four to $500 million a month. Month.

[A]

And this is why The S&P 500 said that Anthropic, OpenAI and SpaceX cannot be on their index for a year. We don't, we don't want your, your tarnish to rip off.

[B]

It's starting to feel more like. Do you remember we work. This feels like we work.

[C]

Yeah, yeah. For computers.

[D]

Yeah.

[A]

Well, we'll be very interested to see what Apple announces as far I think would it be. So it's a couple of things to mention about this. It'll be Tim Cook's last WWC DC keynote. He is retiring in September or actually being kicked upstairs to the board. John Ternus will be taking over as CEO in September. His first keynote will be the iPhone announcement. There's some question whether we'll even see John Ternus on stage. I suspect we will. In fact, I bet you Apple makes a video. They like to start these things with goofy.

[B]

They like to have continuity.

[D]

Yeah.

[B]

And also, can we just say that Tim Cook deserves credit for what he's done since, since Jobs went away. I mean he really has done wonders for Apple.

[A]

He's made Apple a 4 trillion dollar company. That ain't bad. Some say, you know, on the momentum that Steve created there have been a few things that are Tim Cook kind of signature products like the Siri that never was. I guess the watch is still Steve and Johnny's, but it's Tim who recognized that the watch was going to be ultimately best used as a health device and really focused it on health. So he gets some credit.

[C]

Here's the question I have about Apple. Will they end up being seen as the smarter one? Because they don't. They didn't need to build A foundation model. It's commodity. We can get models from anywhere. We can pay Google for it. It doesn't really matter. What we have is the relationship to our users. Are they really behind or are they saving a bucket load of money?

[B]

Except. So one of Apple's core tenets is they must own the. The technologies that are most important for their services. So this kind of breaks that. This breaks it in a big way.

[A]

Yeah, yeah. And by the way, that is Tim's signature achievement is the release of Apple Silicon. The move for intel, two Apple chips. John Ternus, by the way, was the guy in charge of that. And I think that's his. This is his reward for doing such a good job. But that, that is something Apple did that was. It's funny, turned out to be very good for AI, local AI. That's why they can't keep Minis or studios in stock. You're running it. You're running your Requin model on a Mac Mini, right?

[D]

I'm running mine on actually M5 PowerBook. As soon as I got the. I got a. At the job, I have a hardware stipend ripened. And as soon as the. And so I bought the M5 and I bought a matching one for my wife because I told her, you know what, when the ship. When the ship comes in, everybody rides a rising tide.

[A]

Floats all boats.

[D]

So she has a. We have his and hers 64 gig M5s.

[A]

Nice. Nice. Those MacBook Pros are very nice. And it turns out that the machine language processing units in those are very good. And speaking of memory bandwidth, the memory bandwidth is superb because it's unified memory.

[D]

Yeah, it's really good for that and it's fantastic. I highly recommend it. If you want to get into. If you want to get. Yeah, if you want to get into software development. And in fact, actually my old laptop and M1 is now the sacrificial open claw machine. So I've got.

[A]

Yeah, that's smart.

[D]

Open claw, possibly Hermes. I'm also looking into Hermes.

[A]

I love Hermes. That's my.

[D]

Okay.

[A]

That's my agent of choice, Hermes. And I. I actually divorced Claudette and. Oh, yeah, I married Hermes.

[C]

Oh. It was just a fling.

[A]

I call her Quicksilver.

[D]

Wait, Divorce.

[C]

Such a romance forever.

[D]

Divorce or annulment?

[A]

Well, we never did consummate. So I guess it's an okay then.

[D]

Okay. Because that's. That's what. Not consummating is one of the. One of the excuses that you can use for an annulment. And it's the Least embarrassing of the bunch, I think. Mental incompetence. Padre, is that still a. Yeah.

[B]

Is there. Is there prima nocta in LLMs?

[A]

Oh, yeah. Oh, yeah. I. The. The right of the note. Yes, I am the noble one. They're just tools. I learned that from you, Joey. So I don't really actually care so much if Apple puts AI on the iPhone. I use the iPhone, but when I press the action button on the iPhone, it calls Quicksilver. Calls my little Hermes. When I want to, I can do everything in anthropic. And if I wanted to, if, you know, if I want to visit the old girlfriend, if I. I've got all of the AIs on here. But look, this is little Hermes logo. The little weird Alice in Wonderland logo. And I can actually talk with it and interact with it. Oh, ready for an update. Good.

[B]

But what if you could do all of that offline, Leo? Completely.

[A]

Well, I can if I choose Quinn. Well, offline, you mean not going to a Frontier model. You mean local? Yeah, yeah. I mean it's still online because I'm obviously, my phone is not running them, so I don't know if the agent. If the. I mean, Apple's going to do Agenic, probably with. With Siri of some kind. They won't call it that. Is that what you're asking? If like Siri were smart enough to do all that?

[B]

If Siri were smart enough and if you were carrying enough local processing that. Well, maybe I am run a model completely offline. I mean, I.

[A]

That's the idea.

[B]

My current job. That would be very attractive.

[A]

Yes. Because you. You've mentioned this before, but not on this show, that the Catholic Church has its own models. They're purely private, local models.

[B]

And I've got a little. It's. It's not a ggx, but I've got an Acer Veritron, which is using the Blackwell chip. So it does. It's got a petaflop worth of performance, 4 terabytes of storage. You win 128 gigabytes of store of memory. So. So it's pretty good.

[A]

Wow. Now. And are you running the Vatican's models on that?

[B]

We are. So. We are. We have developed our own model. So we're not using off the shelf models anymore because they've been specifically trained

[A]

on our sources and they're private.

[B]

And they're private. Right. Completely private. And which means that we can use them in what we call internal forum cases. So when it's information that's very sensitive, we have the ability to summarize, to translate, to otherwise transform that information without it ever touching the Internet, which is important because canonically we're not allowed to.

[C]

What did it take to. To build those?

[B]

A lot. Yeah.

[D]

No trade secret.

[B]

Yeah, a lot. Did you.

[A]

Did you train them from scratch?

[B]

Yep. What?

[C]

Wow.

[B]

Complete scratch this. It was a lot harder than I thought it would be.

[A]

And when you. When you trained it, what was the content you trained it on?

[B]

So we have several hundred years worth of texts that are specific to the Catholic Church.

[A]

Have they been digitized?

[B]

All digital. So everything has been digitized in our archives. And so we were able to run that and say this. It learned our language. It learned our culture. It learned.

[A]

Does this speak Latin?

[C]

Does this be Latin? That was my question.

[B]

It does speak Latin, actually. Yeah. Yes. That's one of the very first things that we had to train it to do. We had to train it to help translate some of the other sources. And so it's. It's a completely closed model. It does not use any information from outside of the Catholic Church. And like the Washington Posts. Ask a. Ask the Washington Post. AI Our model has no problem saying, I cannot answer that. That was extremely important for us because we. That's how you stop it from hallucinating. Once it gets down to a level of probability, that's no long acceptable, it just says, I don't have enough information to answer that question.

[A]

This is easy because you are. The Church already had its sole document. It's known as the Ten Commandments.

[B]

Exactly. We've got our foundation, we've got our architecture already. We just had to build around it.

[A]

Ten Commandments. Md. It's really. It's. It's simple. Yeah.

[D]

And. And the Beatitudes. Really? And the Beatitudes are a little more than that.

[A]

Yeah, yeah, yeah. And something about rich man and the eye of a needle and, you know, so we've.

[B]

We've got the Old Testament LLM and then the New Testament.

[C]

They battle it out. Yeah. That whole time.

[B]

Must be scary.

[D]

Yeah.

[B]

Fire and brimstone.

[D]

Paging Dan Brown. You can. You can now. You have enough material now to write the Da Vinci Vibe code.

[A]

Does it.

[B]

We don't. Yeah, we. We don't have an antimatic machine under St Peter's but we may have a lot of computer.

[A]

But. And it probably, when you're coding, it uses deep C. Right. Never mind.

[C]

Can you vibe code on it?

[A]

Can it code or is it red C?

[B]

That's not one of the functions that we built into it.

[A]

Yeah, that's Interesting. So yeah, you can actually make it much simpler because there's a whole lot of stuff you don't care to do. Correct. Does it.

[D]

Yeah.

[B]

We're not trying to make a general.

[A]

Can it use MCP servers or is that not a good idea?

[B]

You could, but it would not be a favorable result. You probably blue too many.

[A]

Yeah, right.

[D]

You're thinking. Leo, Are you thinking about my Too many Cats MCP server? Yes. How many. Yes. How many cats is a venial sin? How many cats is a mortal sin?

[B]

Oh, we could do tokens sins instead of tokens.

[D]

Sale of indulgence. Indulgences are back, baby.

[A]

Oh no.

[B]

So we've used up our indulgence budget for the LLM.

[A]

That's okay. It's one of those. What is a year, it's called.

[D]

But we can party like Jubilee year.

[A]

Isn't it Jubilee year?

[D]

There we go.

[A]

You won't use Jubilee Year.

[B]

The tokens are free.

[A]

So. All right, we're going to all go to hell, except for you. But what specific tasks is it mostly about? The hundred years worth of documents or more than 100.

[B]

Right. So it needs to be able to quickly identify documents that are germane to any questions that get asked asked of it. It needs to be able to combine those with contemporary documents from the.

[A]

So it's kind of a big rag computer, correct?

[D]

Yeah.

[B]

But also translations. So it's very good at language translations. We have meetings here where we'll have 19, 20 different languages. This is really good at going back and forth with simultaneous real time translation.

[A]

Oh, so it's used as that. Oh, that's really interesting.

[C]

Did you use anything outside as a base base to build on?

[B]

Not originally, but speech to text is.

[A]

Is challenging. So I would imagine there are some very good libraries you could use safely.

[B]

But again, what we found is it's. It's much easier and much more effective to isolate a particular need rather than building a model and then trying to figure out what it can do.

[C]

Ah, because I think it's a model for what to do with. With medicine.

[B]

Yes.

[C]

And physics and other areas. This is the part of the Yang Lecun argument.

[B]

And there are some really amazing general intelligence.

[A]

There are some really amazing specialized AIs in all those fields. It's really. It's really. Or you know, alphafold is a good example. I mean you wouldn't use alphafold to write your thesis, but you might use it to fold proteins. I mean. Yeah, it's. These are very specialized. Did you find. So were you involved in the training of this thing, it sounds like you were.

[B]

Yeah, a little bit.

[A]

I know that it's a group. I know that nobody takes credit.

[B]

Big group. There's a lot.

[A]

Big group and no, no one person could take credit. But what an interesting.

[C]

What's the demand out there in any given parish church for saying this stuff's going on? I want to know about it. I want to do things with it. What's the.

[B]

It's entirely internal right now.

[C]

Okay.

[B]

It has not yet.

[A]

By the way, the Veratron is Darren OKE says is a deject spark dark.

[D]

It is.

[B]

It's an Nvidia GB10. So it's a Blackwell. It's a Blackwell.

[A]

It's GB10. Yeah. Yeah. Nice.

[B]

I kind of wanted to get a Vera shipment, but those are not available.

[C]

Not yet. Right.

[B]

So do you guys have a Babel fish? Padre?

[A]

Yes, that's what he's made. The Babel fish. Yeah. Wow. Very interesting.

[D]

It's speaking in tongues, actually the most

[B]

difficult language right now. You probably would not guess this is Spanish.

[A]

Really?

[B]

Sense. The Spanish language is easy. The accents. It's having so much trouble.

[A]

You mean the spoken word accents like Castilian versus Mexican, Mexican versus Colombian versus

[B]

Brazilian versus a Bolivian versus Spanish. Spanish, yeah.

[A]

Yes.

[C]

Is it harder than Arabic that all the variation in Arabic?

[B]

Yeah, Arabic is actually pretty simple. Japanese is simple.

[A]

Actually, Latin is very simple. Latin is very rigorous.

[D]

Yeah. The syllables are straightforward. The language is tricky for us anyway, but yeah, the syllables, Mandarin, Japanese, straightforward.

[A]

Actually, Mandarin might be tough because of the tones. Right.

[C]

Well, there's still a written language.

[B]

Computers can hear the tones much better than humans.

[A]

That's interesting. Much, much better.

[B]

I can't hear them, but.

[A]

So we can distinguish ma and ma. Yeah, very easily. Yeah, yeah.

[B]

Did you say you don't have Tagalog yet? Podcast, Andre.

[A]

But we are the most popular.

[B]

We are the most Catholics. We are the most Catholic. It wasn't a demand. It wasn't a demand language. We don't have a lot of documents in Tagalog.

[D]

We need to work. We. We need to work on that.

[B]

Yeah, I'll get to it.

[C]

I got a critical mass right here.

[A]

Do you get native speakers in to train it?

[B]

No, no, we don't need to do that.

[A]

You got recordings.

[C]

Can it make a podcast? Like no book. Lm. You can, Ken.

[A]

I think a Latin podcast would be so cool. Probably.

[C]

Oh.

[B]

Oh, actually that would be fun.

[C]

That would be fun.

[B]

Just as a thought experiment, I would totally. I would take.

[C]

Yes.

[B]

Oh, a podcast translator. Take any podcast input It. And it makes it in. In Latin with. With the voice of that speaker.

[A]

Oh.

[D]

Oh, yeah. Well, hey, I can do that. Yeah. I mean, that should be possible if. Hey, Jen can trans. Yeah.

[A]

Questa in Semin.

[D]

In. What's tech? Yeah, what's tech in Latin?

[A]

Technology. I'm sorry. I hope it's late night in the Vatican and nobody's listening. We are talking twit with Father Robert Balaser. Joey de Villa from Netfoundry IO and Jeff Jarvis will be back on Wednesday. Paris will be back on, I hope.

[C]

Yeah, we hope, we hope, we hope, we hope.

[A]

She's been on a deadline with Consumer Reports.

[C]

Never ending deadline. Poor dear.

[A]

We do have a very interesting interview and Joey, you'll be interested in this. Jeffrey Quinell, who is the founder of Noose Research, the people who make Hermes will be coming back. We interviewed him way back when. I loved him. I thought he was great. But that was back when they were doing their own models. They were trying to do ethical models, kind of like what you've been doing, Robert. But what we didn't know at the time is that they had their own internal agent. And then OpenClaw came out and they looked at each other and said, you know, ours is better. And they decided to release Hermes. And I have to say, I think I'll be very interested what you think, Joy, but my experience.

[D]

Intriguing pivot.

[A]

It's absolutely what you want. It's kind of a batteries included because it comes with more than 90 skills already built in and then plugins available for all sorts of things. It's got its own memory system. It's really nice.

[D]

All right.

[A]

I will have to give it a look anyway. Jeffrey Quinnell will be our guest on Wednesday. Listen, because he's great, he's very interesting. We will have more in just a bit on this Week in Tech. Our show today brought to you by. I would show it to you, but it's busy working our thinkst canary. You know why I plugged it back in? Because we found out that Google sent us a message a couple of weeks ago saying, you know, there's somebody in your network, in your Google workspace. We found out that there had been an intruder who broke in in January. One of our employees got phished, got our credentials. Fortunately, apparently they did very little. They looked at some emails. I think they have such a big stack of workspaces that they've broken into that they don't, you know, they go, well, we'll get to it. They're working through them one by One and we just lucked out that we weren't on the top of the pile. We got rid of them fortunately and we've got new protection. And one of the things that we have is Thinks Canary. Why do we need a think scenario? It's that 121 days that passed between us getting cracked and us discovering that, that the hackers were in the network. That's more than four months that they, fortunately, they didn't do anything. But they could have done so much. The Thinks Canary is a honey pot that can be deployed in minutes. It can be almost anything you want. It can be. Mine is a synology nas. I've mentioned this before, but it's very easy to change it to anything you want, including, oh my gosh, any, a Windows server, a Linux server, it could have lit up like a Christmas tree, every service on it turned on. Or just pick some critical ones like RDP or file sharing that you know, a bad guy cannot resist. It could be an SSH server, it could be a SCADA device, it could be anything. And your things canaries can also generate what they call Canary tokens, little files. They look like the real thing, like a Word document or a spreadsheet or Google Sheet. You can put them anywhere on your local drives. But I, you know, I now have a bunch of them on our Google Drive. They look like the real deal. I have a Google sheet called payroll information, for instance. The thing is, bad guys cannot resist them. And the minute the bad guy tries to crack my fake internal SSH server or tries to open the that sheet, that Google sheet that says Payroll information, I will find out. I will know. I will. The Thinks Canary will immediately tell me there's someone in the network. You got a problem? No false alerts, just the alerts that matter. And any way you want, email, sms, Slack Teams, web hooks, they have an API so you can write your own syslog, of course, any way you want. The point is, when you get that notification, you know there's somebody in the network. Just choose a profile for your things Canary device, register it with the hosted console, it does the monitoring, it does the notifications. And then you can just sit back, you can relax. An attacker who has breached your network, a malicious insider, and other adversaries make themselves known by accessing your Thinks Canary. In fact, I was talking to the founder, one of the founders, Harun, when I was at rsac and he said they may even because hackers are by trade pretty suspicious. They assume everybody's as evil as they are. They may be looking at that going, is that really payroll information? He said, but that's what they're there for. They cannot resist. They will open that file, they will attack that server. They just can't resist. That's what they're there to do. And the minute they do, you got them. Visit Canary Tools Twit. If you're a big operation, a bank, a casino backend, you might have hundreds. You certainly would need one for every virtual network, every vlan, we have a handful. So let's give you a price. Example. $7,500 a year. You get five things canaries, you get your own hosted console, you get upgrades, you get support, you get maintenance. And actually, if you use the code Twit in the how did you hear about us buy bucks, you're going to get 10% off the price. And not just for the first year, but for as long as you own your canaries now, you can always return your things Canary. There's no risk to this. They have a two month, 60 day money back guarantee for a full refund. I have to tell you, they've been advertising with us for 10 years, practically since they started. And during all the years that we've partnered with thinkscanary, that refund guarantee has not even once never been claimed. Because, and I know this because we have them. Once you get Things canaries on your network, you're going to say, how did I live without it? Visit Canary Tools Twit. Enter the code TWIT in the how did you hear about us? Box. And we thank them so much for their support. They've been with us a long time. We really appreciate it. Canary Tools, you might say. Well, Leo, how come you didn't know for under 21 days? Because to do the ad, I stupidly disconnected the things Canary and all the Canary tokens. Call home when there's no home. Because I was holding it up. Remember in the ads? I used to hold it up. Not anymore. So if you see a spreadsheet in the Google Drive named payroll information, you know, you should open that. It's got great stuff in there. Canary Tools slash Twit. Don't forget to use the offer code Twit. Wish we'd had it. I stupidly disconnected it. It's my fault. I blame myself. All right, so just to finish up with the WWDC keynote tomorrow, we will be covering it. We don't expect new hardware. We think that that's all gonna come in the fall. Although it's an interesting story coming out of the Supply chain that Samsung has already ramped up manufacture of these OLED screens for the new laptop. The 14 and 16 inch.

[C]

Are they touchscreens?

[A]

Perhaps they're touch screens and this will be on the M6 chip. The theory being if they're already manufacturing them for August delivery, that it might well be a September announcement, which is earlier than we thought. Apple's going to have a very big September. They're going to have that iPhone Ultra, the folding phone. We're starting to see, see demos of that. Apple will not talk about this tomorrow, so don't, don't get your hopes up. They will be talking about iOS 27, Mac OS 27 and all the other OSes, watch TV, iPad, OS 27 and undoubtedly spend much of the time talking about AI and Siri.

[B]

Do you think there is a market among Apple fans for the touchscreen?

[A]

I mean, I love, I don't want to, but I don't want it. I did want the OLED. I love my. I have an OLED ThinkPad and I bought it specifically for the OLED because those are my favorite screens. But I also specifically did not get the touch one because I don't touch a laptop. Yesterday Lisa was showing me a spreadsheet. She was touching the cell phone. Don't touch it. She'd been using her iPad too much.

[B]

But how good is, how good is macOS on a touch interface anyway? Because that's not touch.

[A]

Well, they have. That's going to be interesting. So this is. Maybe this will be a hint tomorrow because they will talk about new features in Mac OS 27 and they're going to have to do something to address that. Similarly, for a folding phone, they're going to have to do something to address it an iOS27. So we'll get some hints about what they're thinking.

[D]

I think it would require some interesting tweaks because macOS is specifically optimized for a mouse and cursor that is rather precise fingers, you know, while nice and super convenient, especially my fingers larger. And you know, for the longest time the argument for not you not having a touchscreen Mac was what they called the gorilla arm effect.

[A]

Right.

[D]

And that is that it's just kind of hard to hold your arm up against the screen for a long time.

[A]

I, I don't really want to maybe occasionally I'll run my finger if I'm

[C]

going through a long list of something and there's carpal tunnel that gets two. I'm really glad to have my, my.

[A]

Because you have a touch. Chronicle Chromebook I've done.

[C]

Of course.

[A]

Of course you do.

[C]

Yes.

[B]

I, I find it healthier because I switch from keyboard and mouse to touch. I go back and forth, back and forth. So it's not the same motion.

[A]

Well, and Apple exactly sells to the iPad, which is a touch screen.

[B]

Is it going to be an iOS? I would not be surprised, no. If in a couple of weeks, as people start looking through all the notes and actually decompiling the source from some of the software packages, they're going to find the whole hooks for the touchscreen and there'll be hints on exactly how it's going to be implemented.

[A]

That's usually how that stuff leaks out. But I am still most interested in what they do with AI as an. I like AI. I know it's not fashionable on college campuses these days, but I like, I like AI.

[B]

So can we stop calling it Apple Intelligence? Is that death now or they still

[A]

going to try to.

[D]

Yeah, that's what.

[A]

And I think the idea that of Apple, you know, embracing this and putting AI in a billion pockets will make a big, you know, dent. This will change people's point of view. And maybe Apple knows people don't like it. I don't know. You know, there's definitely a backlash, a tech lash.

[B]

I don't see. I think Apple's going to come at it differently. Apple has never been about selling raw services or the promises of a technology. They sell features, specific things that you want to do. I think that's, that's how they can, can do their A.I. well, which is. I'm not going to sell you the tech, I'm going to sell you a skill. Yeah, I'm gonna, I'm gonna, I'm gonna show you an example of, of how AI works. I won't even call it AI, but you're gonna love it so much that you're gonna buy it. And that's.

[A]

They're good at that.

[C]

They're very.

[A]

Yeah, they're product. They make products.

[D]

Yeah, yeah. The ipod, for instance. They never talked about the size of the storage. They said a thousand songs, because that's really what matters.

[A]

That's a good point. Yeah, that's a good point.

[B]

My mom doesn't care about anything on her iPhone except for the fact that she can do FaceTime with her granddaughter.

[A]

Right.

[B]

That's it.

[A]

And unfortunately, the things they've shown in the past haven't really been too compelling, like Genmoji or the Image Playground. These are awful. And people are still mocking the AI summaries that they get in Their notifications. They're. They're crazy.

[C]

Will Siri still be a laughingstock?

[A]

I mean, and that's the other thing. People think Siri is a moron, and she is.

[B]

I mean, if they're really doing a tight immigrant integration with Gemini, Siri is going to smarten up by a lot. Exponential level. I mean, it's fantastic.

[D]

The question is, will they still call it sir? Will they call it Siri, or does it.

[C]

You think they're stubborn?

[B]

Oh, yeah. Yeah. That's theirs. That's their technology. That's their branding.

[A]

They're not going to admit, even though people hate it. Yeah.

[C]

Apple doesn't think you hate anything that Apple does. They'll never admit that. That would be a.

[B]

I think. I think they should release their new AI product and call it the Newton. Just.

[C]

Just look at that.

[A]

Sure.

[D]

There we go. Yes.

[B]

Just a flex and say, you know what? We're going to take one of our biggest failures and make it something wonderful.

[D]

There we go.

[A]

I did have Gemini last night. Lisa and I were talking. I don't know if I can show this, though. Wait a minute. Let me see. And I asked Gemini if I should dye my hair. Wait, I'm thinking probably not a good idea.

[C]

Did it show you what you would, like, look like?

[A]

Yeah, you want to see. You want to see what I would look like if I dyed my hair?

[C]

The chat room could probably do it for you, too.

[A]

Yeah, well, so. But the point being that I didn't ask Siri.

[D]

Okay.

[A]

I immediately went to what I thought would be the best image, so I gave it this picture.

[C]

Oh, no.

[A]

Oh, wow.

[D]

Okay.

[A]

So this is what I would look like. Oh, no. It's probably not the best starting picture. Maybe that's the problem. And then I said, well, that's not good.

[C]

That's not a crisis if I respond.

[A]

So then I said, what if I dyed it black? And then Lisa said, you look like Joe Pesci. So I thought, okay, what if you get. How about give me Superman hair?

[B]

Do I amuse you?

[D]

Oh, the spit curl. Oh, very nice.

[A]

Superman.

[B]

The corn sweat look like it?

[D]

Yeah.

[A]

Yeah. Then, so what about a Beatles haircut? What if I had black hair with a Beatles haircut?

[B]

Is that Ringo or is that. No, that's John.

[A]

The interesting thing is Gemini's really good at doing this. Right? This is.

[B]

This is.

[A]

This is if. What would I look like with a toupee? Apparently, I don't know.

[D]

You know, there's a certain Javier Bardem vibe I'm getting from that photo.

[A]

We Were watching the Burroughs and Molina is. Alfred Molina's in it.

[D]

Okay.

[A]

And he's got. He's my age and he's got jet black hair. And I said, he dyed it. Right. And we couldn't decide, so I said, well, what if? What would I look better? And then this is. If I went blonde.

[B]

This is weird. This is weird. But I actually like that over the black.

[A]

Yeah, I do, too.

[B]

I prefer that.

[A]

I'm not doing any of the.

[B]

It's got Conan o' Brien vibes.

[A]

Yeah, that's what it is. Yeah. Oh, I should ask for Conan's hair. Oh, I will.

[D]

There we go. Yeah.

[A]

Let me. Give me. So the point I'm making is not so much. It's just really good at this kind of stuff.

[D]

Yeah. But I know the toupee. It's the Javier Bardem in the Apple TV version of Cape Fear that I'm

[A]

getting, by the way. Who is that? Creepy. Have you watched that yet?

[D]

No, not yet. It's on my list.

[A]

Oh, what is this?

[B]

What is this?

[D]

Apple TV cape with Rob da Cape Fear.

[A]

And then.

[C]

Yeah.

[D]

Robert Mitchum first.

[A]

Yes, go say, okay, here I am. By the way, it did that fast. Oh, wow. Here I am with Conan o'.

[C]

Brien. Perfect.

[A]

Okay.

[B]

No, no.

[C]

Oh, my God.

[D]

I feel like I'm in a bus station right now.

[C]

The clown.

[A]

Carrot Top. What was I saying? Oh, yeah. There was a Robert Mitchum movie, Cape Fear, and then. And then Scorsese remade it. It's exactly as if it were a Hitchcock film. And that was with De Niro, and he was very good at that.

[C]

Okay.

[A]

But this one is Javier. Javier Bardem. It's really, really good. And Amy what's her Name is in it. And I'm such an old man now. I'm having.

[B]

See, I only watch all of this when I go back to the United States. So I go back three times for about a month of time.

[D]

All right.

[B]

My parents get all the streaming surfaces, so I just binge everything.

[D]

Okay. But, yeah, that's on Apple tv. The Cape Fear, this new one.

[A]

Yes. And I think it's quite good. It's been actually on.

[B]

On Apple tv. I really enjoyed Pluribus.

[D]

Oh, yeah.

[C]

Yes.

[A]

So.

[B]

Really enjoyed that.

[A]

So we were watching the Burrows, which is basically plurib of us.

[B]

Yeah. Yeah. No, I've heard that's very good.

[A]

It's like. It's just. I don't. Anyway, I'm not gonna.

[B]

Isn't. Wait some. I think my sister called Burrows, like, Stranger Things for old People,

[A]

Duffer Brothers. And, but, but they're living in a place that's kind of like the Villages, only it's called the Burrows.

[D]

Oh, the Villages.

[A]

It's like they all drop golf carts.

[D]

That's down the road from me. And the beautiful thing is they, it's like an amu. The main street is like Main Street USA in Disney. Except the music they pump is classic rock.

[B]

Like they were like classic rock or boat rock.

[D]

No, no, no, this was, this was. Well, actually boat rock fits in there. But when I was there, they were playing all of Boston's original album.

[A]

So that's what's going on in the boroughs too. The music is like Springsteen.

[C]

And because we're of that age now,

[D]

now here we are.

[A]

What happened?

[B]

So if it's Stranger things for old people, does that mean in Burrows vecna is just like chlamydia or syphilis?

[D]

Yeah, exactly. Because remember the, the Villages has an unusual, possibly the highest SCD rate in for a medium sized city. And yeah, by far carpet. These people.

[B]

Seriously?

[C]

Yeah.

[B]

Untreated STDs in the villages is the highest rate in the country.

[D]

Country, yeah.

[A]

Okay. Jeff and I are both. Well, we're not moving there, I guess.

[D]

Yeah, yeah. And actually if you search, if you search YouTube, there are a bunch of real estate videos promoting life in the Villages.

[A]

And Yeah, well, there's a famous documentary that's hysterical about the village.

[D]

Oh yeah. No, so, yeah, I, I, I have been there with my wife and asides from the staff, we were the youngest people at this seafood restaurant.

[C]

Oh yeah.

[D]

By far. And it was, yeah, it was wild.

[C]

My parents lived in Sun City center, where by law you cannot live there unless you're over 55.

[D]

Yes. You have to be 55 minimum to buy property there.

[C]

Oh, that's.

[B]

My parents live in a Sun City near Vegas Chicken. So.

[D]

And they, and they have so many activities. There are a lot of clubs and last I heard there was a three year waiting list for the cheerleading squad. Jesus.

[C]

Huh? Yeah.

[D]

There.

[C]

Are they cheerleade football team?

[D]

I have no idea, actually. And there's a lot of Morris dancing or clogging clubs.

[A]

Oh yeah, sure. Yeah. Yeah.

[B]

That doesn't seem to go hand in hand with arthritis. Painful?

[D]

No idea. Maybe. Maybe a lot of glucosamines. I don't.

[A]

No.

[C]

Yeah, the golf carts, man. The golf carts.

[D]

But a very weird city and very nice golf carts. These golf carts are nicer than my car. I was just.

[A]

Oh yeah.

[D]

Oh yeah, yeah.

[A]

That's part of the, that's not one

[C]

with Air conditioning in Sunset Center.

[D]

Oh, and the best part is the crime blotter in the Villages. It's not teenagers. It's the 30 and 40 something children of seniors who fail to launch. And it's all like kids petty.

[B]

So they're all living.

[D]

Theft, shoplifting and golf cart dui.

[B]

It is a fascinating community.

[D]

Yes.

[B]

I always wanted a little bit of that in, in Vegas.

[C]

I wanted to make a. A sequel to Seinfeld, which was the town where the parents lived.

[D]

Del Boca Vista.

[C]

Exactly. And it would be a sitcom because Burroughs is not a sitcom. Burroughs is. Is. Is.

[A]

No, it's nasty.

[C]

Yeah. Spooky. It's an art. I think there's a sitcom. It is. My father always called it God's Way,

[A]

called the Golden Girls. And it's been done.

[D]

Oh, yeah. True, true.

[C]

But you know what?

[A]

Now that we're now the boomers are our. Our little demographic bulge is getting that getting up there. I suppose it's time to bring back all those old folks shows. I guess that's what the Net star Jerry Seinfeld out. Yeah, it's trying out. Let's take a break. I was going to do another because we're old.

[C]

We need a break now.

[B]

We don't want to talk about STDs anymore.

[A]

I mean, the only bad thing about these breaks is I don't get to go to the bathroom. But you do, so take advantage of it. And there we go. I'll just suffer.

[D]

Think empty.

[C]

He's thinking Depends right now. But you don't want. Folks,

[A]

you know what? If Depends wants to buy ads, just let me know. That would be the kiss of death, right? If you start doing Depends ads on a podcast, nobody's gonna ever listen to the show again. Our show today, brought to you by Melissa, the trusted data quality expert since 1985 and a longtime sponsor of our shows. You know what? It's another company. It's been with us about a decade, I think, maybe more. And that's because inaccurate data is everywhere. It doesn't just hinder your company's operational efficiency, it can also degrade the quality of your AI outputs. Right. Garbage in, garbage out. The AI models are only as reliable as the data that feeds them. So whether you're managing customer relationships, logistics, internal analytics, clean data is essential. Melissa's been solving data quality issues for going on 41 years now. And this is what I love about them. You know, 41 years ago is probably zip codes and stuff like that, but they do not rest in their laurels. They are data Scientists. Now they've hired some of the best data scientists and AI experts in the world. And here's just some of what Melissa, the modern Melissa can do for your business. They have an alert service. Melissa's alert service is the best way for companies to keep track track of changes to their customer data in real time. Alerts for changes of address, address updates, property transactions, hazard risks. They also have something very cool. It's very hard to do. Their matchup technology can do smart deduplication with non exact matches. That's something everybody needs. Very hard to do. They do it well. They also do global global worldwide address verification and and autocomplete. They can validate and standardize addresses in real time super fast across more than 240 countries. That's really important because addresses are different in every country. And if you want that mail delivered, if you want those packages delivered, you need to have it right. Melissa's change of address tracking will monitor and update customer address records automatically when they move. Data enrichment lets you append demographic data, property information and geographic insights to the contact you already have. Mobile identity verification. Great for fraud. Well, great for preventing it anyway. Connects customers to their mobile numbers seamlessly. That really does reduce fraud. Very helpful if you're in a business that has KYC regulations. For instance, join more than 10,000 businesses worldwide using Melissa including speaking of KYC, eToro, the social investing platform. They've got 23 million users. EToro went to Melissa for its unmatched global data coverage, its proven expertise in electronic identity verification and its commitment to compliance. Etoro's business analyst said quote we find electronic verification is the way to go because it makes the user's life easier. Users register faster and can start using our platform right away. Melissa is great for any compliance need no matter the size of your business. Business Melissa will integrate with you. They have easy to use apps for all the tools you use. Salesforce, Dynamics, CRM, Shopify, Stripe, Microsoft Office, Google Docs, I can go on. They're pretty much everything. Melissa solutions and services of course absolutely secure, GDPR and CCPA compliant. They're FedRAMP certified, ISO 27001 certified. They meet SOC2 and HIPAA high trust standards for information security management because that's for them. Job 1 get started today with 1000 records cleaned for free at melissa.com TWIT they are the data quality experts. Melissa.com TWIT M E L I S S A.com TWIT we thank them so much for their support of this week in tech Wired magazine, very interesting expose. They looked at code in an unreleased system that's already embedded in Meta's smart glasses for face recognition, designed to identify people via biometric data stored on users phones. They have not enabled it, but it's there, it's in there. Wired found was added to Meta's AI app over multiple updates this year. The feature is internally called name tag. This is something Google explicitly eschewed back with Google Glass. Right. But on the other hand, Ring and Google Nest doorbells now are starting to do this. You're starting to see this more and more. Are people getting used to the idea or do you think this is the reason that Meta is keeping a secret is this is a non starter.

[B]

I think people are getting used to the idea that it's possible. I don't think people are accepting of that idea. I mean this is, this is why Meta has been very cagey about it. This is why Meta in hearings has basically said we would be very careful about implementing this technology into our products. I think most of us understand that there's cameras on us everywhere no matter where we go. We've seen it, we see it on YouTube, we see it on the nightly news. But the idea that someone could be specifically looking at us, identifying us, finding information about us, us just as we walk around outside, I don't think we're ready for that as a society. Not yet.

[A]

Yeah, I, and yet it would be so useful. Yeah.

[B]

No, yeah.

[A]

Because I have a terrible memory for faces. I never, and it's embarrassing that I can't remember people's names.

[C]

My sister is, is a retired minister and I can't imagine how you keep, you don't know the name of the parishioner. Oh, you are in duty, right?

[D]

Yeah.

[C]

Or, or my father was a sales guy. You're, you're in a conference. I, I, I'm horrible. Horrible faces and names. Absolutely horrible. I could never, I, you know, give me 12 students and it takes me weeks to get them straight.

[B]

If I get about 10 years older then I can just start calling everyone my daughter and my son.

[D]

There you go. Yes. My child. Child. You, my brother or sister in Christ actually will always help you. Yeah.

[A]

Although that's now a meme if you

[D]

say that it's, well, yeah, that's the thing. My brother. Yeah.

[B]

In the Filipino world, everyone you, you can't remember their name. They're Tita and Tito.

[D]

Yeah, exactly.

[B]

Just that's it.

[D]

I bet every language uncle and auntie.

[A]

Because this has to be a universal issue is not remembering People's names. Right, Right. Every language must have a default. Y'. All.

[B]

Everyone here. Everyone here is bus. Leo. We know it wouldn't boss. It wouldn't just be the names.

[A]

Benito's in Manila. He says, everybody's boss.

[B]

It's busing. It's like what we say here.

[C]

Everyone's bus.

[A]

Yeah.

[C]

When I was a columnist in San Francisco and they actually had my face on news racks and. And. And. And trucks, and people come up to me and. And I wouldn't know whether I'm supposed to know them or anything. Not.

[A]

Oh, yeah.

[C]

So that's what I. I took on the horrible conceit of saying howdy, because howdy has no ellipsis.

[D]

Okay.

[A]

In San Francisco, you don't say, howdy, Leo.

[C]

You just say, howdy.

[D]

Howdy. Yeah, howdy.

[C]

If you say, hi, there's a Leo.

[B]

I've done that to Cardinals.

[A]

And you said, howdy.

[B]

Do a card feel good about. No. I say, oh, it's so. I haven't seen you in. And I'm doing all those things. When was the last time we saw each other?

[C]

Wait, what?

[D]

We working? Yeah. You're trying to get context.

[A]

This is the only reason that you should have a wife. Father Robert, I'm not saying any other reason, but. But Lisa and I have it worked out. She will say, okay, I don't know the name, so just introduce yourself right away. And then they will introduce themselves. And then. And then she will know their name. It's very handy if you have a partner of any kind. I'm Leah. What's your name? Cardinal.

[B]

We had a Jesuit named Paul Locatelli who was the president of Santa Clara University. And he had this incredible talent. He knew every single face of every single graduate and every single parent he had ever met. He could walk into a room cold. And he knew your name. He knew who your child was when they attended the university and what they graduated in. It was. It was ridiculous. I never figured out how he did it, but he just had one of those memories now.

[A]

So. The glasses.

[B]

I could do that. Right.

[A]

The reason this Meta thing is troubling is because we have all, for years, been tagging people with their names on Facebook. Right.

[B]

And Google. Google Photos, the same thing.

[A]

And Meta, apparently. Wired's code review shows the name tag system is currently designed to pull face prints from Meta's servers and store them on your device, in your hands.

[C]

On your face. Not a big deal. In Palantir's hands.

[D]

Hands. Yes, in.

[C]

In. In Ice's hands. That's when it Becomes really why?

[B]

For the last seven years, I've been poisoning both Facebook and Google with photos tagged with my names that aren't me.

[A]

You are a master of fuzzing.

[B]

I. Yeah, it's just. I was a free. Free time project.

[A]

Yeah.

[C]

Does Rock. Does Robert Redford have your name?

[A]

I mean, who.

[C]

How do you choose who?

[B]

No, no, no, I just, I put my name and I just use different phases. Just different phases?

[D]

Yeah.

[B]

So this, the system doesn't know. If you look in Google for Robert Baller, you'll get like 150 different people.

[D]

Yeah, they're. Yeah, they're, they're, they're quite different.

[A]

I'm afraid that the ship has sailed for me.

[D]

Well, yeah, but you know what, in your line of work, you actually need. You know, it depends if you have public facing work. If you are. And my line of work as well. Yeah, no, we want our faces to be matched, but maybe our habits, we want to appear different, which is why. Which is why I run that program called Chaff, which just does Google searches on random dictionary words. And that's why I get chicken mating harness ads all the time.

[A]

Oh, we talked about that last time you were.

[D]

Yeah, exactly.

[A]

I think we named the show Chicken Mating Mating harnesses.

[D]

Yeah, exactly, exactly.

[B]

Thinks I'm in Vietnam. Vietnam. So if I do searches, it thinks I'm in Vietnam. It gives me local, local answers for that.

[A]

So somebody in our YouTube said, if you talk about AI anymore, I'm leaving. Well, I'm sorry, but there's a show for you. It's, you know, it's in the news.

[B]

For instance, this was an AI week.

[A]

Yeah. The president this week signed an executive order seeking oversight of AI models. Nobody, I don't know who thinks this is a good idea. They've softened it considerably.

[C]

It's voluntary.

[A]

It's voluntary. It's only for 30 days before you release it. He scrapped the less voluntary 90 day order. The idea being. And I think it was really stimulated by Anthropic's Mythos. Before an AI company can release a new model, the government should be able to check it to make sure it's not going to be as if they

[C]

would know what they're doing.

[B]

Yeah, they fired all the experts who would be able to go through a model and tell you whether or not

[A]

it's good for the common good. Okay, so if none of that were true and you trusted the government, this seems like exactly the kind of thing government should do before. I mean, you can't release a vaccine without the government's approval. Right.

[C]

But again, it's a general machine. A vaccine is meant to do one thing and you can test its.

[A]

Well, you test it for safety and efficacy.

[C]

Right, right. This is a machine. You don't know what everybody's going to ask the machine to do, what malign things are going to be required of it. And thus you cannot protect against all of them.

[A]

This is why we have underwriters laboratory checking electronics to make sure it's not

[C]

going to do one thing go on fire.

[B]

Right.

[D]

Yeah.

[C]

That's pretty straightforward. Is somebody going to come along and ask this to create a, you know, a new weapon? Are they going to find a way, new way to insult Leo Laporte? You.

[B]

We don't know.

[A]

So there's no way. The issue with this is there's no way to ascertain the safety of a model.

[C]

No, no.

[D]

Not in 30 days.

[C]

And the encyclical also made clear to the idea to align it with human values is absurd, too. At the higher level, it's not just the guardrails, but also this notion that we're going to make the virtuous machine is hubris.

[A]

So, you know, when Trump came into office office because he was funded to some degree by Marc Andreessen and Brockman, the president of OpenAI, he got into office, the first thing he did was throw out Biden's AI regulation, or it was really just an executive order, so it didn't have much force of law either. But he went into office kind of with a stated goal of opening it up for AI because that's how we're going to compete. That's how we're going to, to make America great again is. Is, you know, beat the Chinese is by. No, no regulations limiting AI. So doing this is almost a complete reversal

[D]

and not the first one yet.

[A]

He's done it in such a way that it is so, so meaningless that it is it. It appeals to both sides. It has the appearance of regulation without actually doing anything. Anything.

[C]

Yeah.

[B]

This actually goes hand in hand with the other bit of news concerning the administration, which is Trump is considering having the United States take ownership, part ownership in these AI companies. He wants. He wants a bit of the pie.

[A]

Interestingly, Sam Altman proposed this from.

[C]

And Bernie Sanders is talking about a version of this as well. It's socialism.

[D]

Yeah. It is so nationalizing industry. Yeah.

[A]

That's already. The government has a 10% stake in Intel.

[B]

Intel.

[D]

Yeah. Yeah.

[A]

I mean, there's a precedent for this. This was also a Trump. This was actually. It's Interesting because it was the CHIPS act that Biden passed. But the President said, well, but for the money, we're gonna give you, we want a stake in Intel. When asked about this on Air Force One, Trump said, there's something very interesting about it where it almost becomes a partnership with the American, American public. It's like you make them partners in this revolution, it would be a beautiful thing. It would make them rich. I don't know who them is in this. Okay, okay.

[B]

That was the same thing that happened when he signed the executive order making insider trading not illegal. So, yeah, making them rich is not necessarily a good thing.

[A]

Who do we want to make rich here?

[D]

You know what they say, Evil is the root of all money.

[B]

Money.

[A]

I like it. I like it. The fcc, as you know, has banned all foreign routers. Now the cable lobby, the National Cable Television association, wants a waiver because, well, guess who installs most foreign made routers in the home? The cable companies. That's where most people get their routers.

[C]

Hell yeah.

[B]

Do we even have a list of what routers would qualify? Because everything's made in China.

[A]

Well, right now the only one not made in China is the Starlink router. But remember that Netgear got a waiver. Eero got a waiver.

[D]

Oh, okay. So I'm good.

[A]

Yeah. But the way you get the waiver is you just say, yeah, we're going to build these routers in the United States someday.

[B]

Eventually.

[D]

Eventually. Yeah, yeah, exactly.

[B]

It might take us three years someday, but then it will be done.

[A]

And at least the FCC said you can continue to get off software updates and Firmware updates until January 2029. She's. It's like, okay, it's about a security problem.

[B]

I mean, okay, look, I understand the idea of securing the edge.

[D]

Yes.

[B]

Let's secure all these routers. But even if you did that, all of the core is running on equipment made in China.

[C]

Yeah.

[B]

So are you going to do that?

[A]

Everything?

[B]

Because that's kind of important.

[A]

When we were joking about this deep sea thing that I've, you know, this Chinese device that I connected to my WI fi and connects to immediately to deep sea in China? I said, well. I said, well, isn't that dangerous? Harper Reed suggested. He said, leo, what of your stuff in your house is not made in China?

[D]

Well, yeah.

[A]

Which of the WI Fi things you've got connected? I have more than 100 devices connected to my Wi Fi. Which of those is not made in China? None of them. As far as I know, they're all made in China, so. But a router certainly is an attack surface, there's no doubt about that.

[B]

The last device I could certify was 100% not made in China, was. Do you remember Google IO? One year they released that sphere.

[A]

Oh yeah, I bought that.

[B]

That was 100% made in the United States and discontinued immediately because they could not make it cheaper enough.

[A]

Right, yeah, right, right. Did I buy that one or did I buy the camera that they put out that took pictures of you every five.

[B]

No, they gave us the spheres just

[A]

by going to Google IO. That's right. They gave you at Google. I O. Yeah. Google actually is a little bit in a little bit of trouble in the uk. Google has to change its AI overviews. The UK ordered them to put clearer links in AI search and to give UK publishers the option to opt out of those AI or for views or put that.

[C]

In other words, allow the publishers to commit suicide.

[A]

Right. You don't want to be in search, you don't have to be in search.

[D]

Yeah.

[A]

This is from the UK's Competition and Markets Authority.

[B]

It seems like we keep running that same story over and over.

[A]

It does, doesn't it?

[B]

Was the news readers, the news aggregators and the. Than the listing on Google search. It's. I mean, yes, it sounds good, it's fun to fight Google, but that ultimately it's suicide.

[D]

Yeah, yeah, Google.

[C]

And Google is improving the links on. On these searches. I'm finding the links are more prominent now. They're easier to go to. There's more of them.

[A]

They didn't want to do it because they said it junks up the search results.

[B]

No, what they really mean is because we can't game the search results as easily as we could before.

[D]

You might.

[A]

You might actually click that and go out of a. You might leave.

[D]

You might leave.

[A]

You leave. Sites that opt out will not receive traffic or impressions from our generative AI features.

[C]

Here's a cup of hemlock for you.

[D]

Yeah, exactly. Because you know. Yeah. Because as always, piracy is not the real problem. It's obscurity.

[A]

Yeah.

[C]

Yes, yes.

[B]

By the way, Microsoft is doing the same thing with Bing. So there are stories from large publishers that end up up on Microsoft's the MSN page. And if you search for that story, it will give you the MSN link. You actually have to go to Google to search for the story to actually get the original link. I don't know why no one's complaining about that.

[C]

Because no one's using Bing.

[D]

Yeah.

[B]

I've got over a million points on Bing. I'm going to spend them at some point.

[A]

Well, good news. You can keep your Microsoft Windows 10 running for another year with what is it, 5,000 Bing points. You could keep that DGX spark on Windows forever.

[B]

You've only got four months at this point though.

[A]

Oh, that's true. Well yeah, that's a good point.

[B]

I think Windows 10 might be the last windows on my machines. I'm going to migrate them to Linux. Seriously, I'm not using 11. I'm with you.

[D]

Yeah, I'm already. I have successfully migrated my in laws over to Linux. It is working fine.

[A]

Which version of Linux did you choose for this? Them.

[D]

Mint.

[A]

Because it looks like Windows. It feels like Windows.

[D]

Yeah, it looks like Windows. It feels like Windows. And then the next thing is to put a Raspberry PI that I can open ZD into to maintain their system.

[A]

ZD is a zero Trust open source zero trust system from NetFoundry IO. So you're. Oh, that's interesting. I could imagine you using a pie hole or something. In fact Father Robert taught us how to install PI holes back in the day. Unknown. That would limit their DNS searches and thereby limit them from going to bad search.

[D]

Yeah, and I can combine that. The reason I would use OpenZD is because if you port scan the network that Raspberry PI does not show up at all. The only way into that Raspberry PI is by using a cryptographic id. Because in Netfoundry we basically say in God we trust, everybody else gets zero trust.

[B]

So it's got no open ports, it's behind the, it's behind the the router and it.

[C]

There is a.

[D]

You communicate with a network overlay and rather than via a port. And what that basically means is that scale a little bit or in the similar field tail scale is a vpn. So once you're inside the network you do have permission to do anything. Whereas with OpenSea because it's zero trust,

[A]

you can only do what the policy

[D]

determines, what you're allowed to do.

[A]

Yeah, that sounds like a good solution.

[D]

Yeah.

[A]

All right, I'm going to take a little break and then we have our final stories. This has been so much fun. I hate to wrap it up. Joey de Village don't be telling me

[C]

you don't have mass at 6:00 in the morning, Robert.

[B]

I don't. I have the 4:30

[D]

wait in the morning.

[B]

I'm not gonna go to sleep until

[A]

after the 4:30am Good a.m. who comes to 4:30 mass?

[B]

There's a. There's a special mass we do down in the tomb of St. Peter's and.

[A]

Oh, that's a, that's probably very prestigious to be doing that.

[B]

Yeah.

[A]

But I'm down with the, with St. Peter Pope 1.0.

[B]

When you come, I'll, I'll take you down there. It's. It's very nice.

[A]

Well, there's a window you can look when you, if you go to the basilica.

[B]

No, no, no, this is the special one. This is the church Chapel of St Clementine. It's right next to the tomb of St Peter.

[A]

It's so is. But isn't the, the altar in St. Peter's right above that straight up?

[B]

Yeah, but, but there is, there's no way to look down there. There, there is ways to look down, but not into this chapel. This is a special chapel.

[A]

Oh, wow. Wow. So I'm telling you, that's a pretty high end mass. Do you, do you draw straws for that? Who has to get. How do you get that?

[B]

You just have to ask the right people. I'm. I'm at the point where I'm, I actually know some of the right people. Now that's.

[A]

And now how. It can't be a very big chapel.

[B]

Oh, a maximum is 12 people sitting or maybe 17 people standing.

[A]

So who, who gets invited to that mass?

[D]

Yeah, yeah. Of

[B]

dignitaries, diplomats.

[D]

It's the Illuminati, isn't it?

[A]

It's the Illuminati is what it is. That's.

[B]

The guy's running the antimatter machine. The guys running all the, the alum stuff.

[A]

So cool.

[C]

We've got to go another three hours to keep him awake now.

[A]

Oh yeah. Well, we might as well entertain half. Holy cow. Wow. And you, do you do that in the vernacular? Do, do you do that in Latin?

[B]

Italian, actually.

[D]

Italian.

[C]

Italian. Do you do.

[A]

I'm sorry, no, you do a homily.

[D]

Homily or.

[C]

Yes, yes you do.

[B]

But a homily is short. We. So it's five minutes, which for some people that's difficult for me, that's normal. I, I always do short homilies.

[A]

What will. Do you know what your homily will be today?

[B]

I make it up as a go. I, I don't write it.

[A]

It.

[B]

I, I always, I like it to

[A]

be, you know, what the subject. Do you have a verse?

[B]

Oh, no, no. I, I've read the Scriptures and I, I know what, like the ideas that I want, but I, I make it as I go just so that it's dynamic. I'm not one of these guys who goes up and says, mar.

[D]

I'm picturing the Stone Cutter song. Who controls the British? We do.

[A]

I. I did make a song. You were here for that? Out of the.

[B]

Oh, yes. On six, seven. Six seven.

[D]

Yeah.

[A]

Well, six seven, by the way.

[D]

Yeah.

[A]

I have been told by a number of our club members is so catchy, they can't stop singing.

[B]

Leo, today is six seven, by the way.

[A]

Okay.

[D]

Yeah. And today.

[A]

So we will end the show with 6, 7. Although I'm told by the kids it's now 6 8. What? I don't think that's true. I think they're messing with me.

[B]

No, no.

[A]

I will play 6 7, 6 7.

[D]

By the way, 668 is the neighbor of the beast next door.

[A]

It's right next door. Coming up in just a little bit. Father Robert Balasar, who is staying up late tonight. So nice to have you. Wow. I'm just blown away. That is amazing. What's the name of the chapel? St. Clementine.

[B]

St. Clementine.

[A]

Clementine, wow. The patron saint of oranges,

[B]

is he.

[A]

Aren't there clementine oranges?

[C]

Clementine oranges.

[B]

Sometimes for Christmas, I. I lose track of the. The patronages.

[A]

Oh, I know.

[B]

It's complicated now.

[D]

Yeah.

[A]

Abby took the name of St. Abigail because she's Abby. Yeah. So that was nice. Yeah. Jeff Jarvis is also here. He's a good Presbyterian. Don't. Don't joke.

[C]

Well, no, actually, a very bad one, but that's what I mean.

[A]

He's a bad Presbyterian, even worse. But he's a good Scotsman. Thank you for being here, Jeff. It's always great to see you. Of course. He'll be back Wednesday for Intelligent Machines and Joey de Villa. Congratulations. Netfoundry IO. Hey.

[D]

Hey.

[A]

He's putting ZD on everything.

[D]

Oh, yeah.

[A]

ZD on all the places. I wonder if maybe I'll put open ZD on my. On my router. That's a really good idea. I do have tailscale running, but you're right, once they're in the network, they can do anything they want. Gotta limit that. Very, very good. Very, very good. Our show today, brought to you by my mattress. I love my mattress. Gosh, I had such great dreams last night. Benito, you were in my dreams. I think I was yelling at you, though, so I apologize. I apologize for that. I love my Helix Sleep mattress. About a year ago now, before we were doing the ads, Lisa and I realized that we'd had our mattress for almost eight years. I think almost nine years. And I read probably my agent Told me that you're supposed to change your mattress between every six and 10 years because they were out. They sag, they bow, they don't cool as well. It was time for a new mattress. So we did a lot of research and we found Helix Sleep and we are so glad we did. You will be too. It's time to upgrade to a Helix mattress and get a great night's rest. No more night sweats, no back pain, no motion transfer. And you know, we didn't want to settle for a mattress made overseas with kind of low quality and questionable materials and packed into a box stuffed on a container ship across the sea. Six months sitting next to the bunker. Fuel oil. No. But we chose the Helix mattress because we found out it's assembled, packaged and shipped from Arizona. And they make it to order. So they do all that within days of you placing your order. Then they make the mattress so it is fresh. It is amazing. We also did what I recommend you do, which is take the Helix Sleep Sleep quiz. This matches you with the perfect mattress based on your preferences. You like firm, you like soft. What do you like and your sleep needs. I'm a side sleeper and I wanted something, you know, for side sleeping. In a Wesper sleep study, Helix measured participants sleep performance. After doing what we did, switching from their old mattress to a Helix mattress, you know, it's interesting, our experiences exactly matched this stuff. Here's what they found. 82% of participants saw an increase in their deep sleep cycle. In fact, participants on average achieved 25 more minutes of deep sleep per night. That doesn't sound like much, but you really, you only get between half an hour and an hour a night of deep sleep. But that's the most important 1. A 50% increase in your deep sleep. That's huge. Really impacts how you feel the rest of the day. Participants on average also achieve 39 more minutes of overall sleep per night. So you sleep better, you sleep longer, longer, you feel great the next day. Time and time again, Helix Sleep remains the most awarded mattress brand. I think that's what we did. We looked around and we saw all of these awards tested and reviewed and ranked number one by experts like Forbes, like Wired, like Oprah, Helix delivers your mattress right to your door with free shipping in the US and rest easy, they have seamless returns and exchanges. They call it the Happy with Helix guarantee. It's a restaurant risk free customer first experience ensuring you are completely satisfied with your new mattress. We didn't need to return ours. You're never getting it back from Me, I love. I love my Helix. Sleep you will too. Go to helixsleep.com TWIT and find your perfect mattress match. Make sure you enter our show name after checkout. Just say twit so they know we sent you. Go to helixsleep.com twit that. That's H E L I X sleep.com twit. You're going to love it. Helixsleep.com twit it is the seventh day of June in the year 2026. I'm going to play just a little bit of this. I'll jump to the. This is. I'll jump to the chorus.

[D]

It's nonsense half the time it blooms.

[A]

Everybody sing along don't overthink it, that's

[D]

the trick it lands like a shrug then it sticks A number with swagger, a wink, not a clue they say

[C]

it for the vibe not to hand

[D]

it to you 6, 7, 6, 7 that's the whole scene 6767, 67 it's a flex, not a fact.

[A]

Oh, Leo,

[B]

it is catchy. And I. I hate that. I hate catchy.

[D]

Padre.

[A]

We need a 767.

[D]

We need a karaoke version.

[B]

Yeah. We're Filipino, so we're gonna.

[D]

Yeah, exactly. Yeah.

[A]

Well, you did hear the. The. The Gregorian chant. I think I made it while you were there. There to the Pope's encyclical. To section 238, your favorite section.

[C]

Yes. My mind was 99.

[B]

Did you put that text in there or did you just.

[A]

But it had. But the lyric writer, which is probably chat. GPT. I'm thinking clearly can go out because I didn't tell it what 67 was. Was. And it wrote those lyrics. So it went out and figured out what it was, you know, what six seven was all about. And it nailed it, frankly. Nice that it means absolutely nothing. The kids just use it to mess with your head. Oh. I just closed the window of both AT&T and Verizon loss in the Supreme Court. They upheld the FCC's power to fine them over data sales 8 to 1. But it's a mere $104 million in fines for AT&T and Verizon.

[D]

A rounding error.

[A]

It's nothing.

[D]

Who's the one?

[A]

Yeah, really. AT&T and Verizon didn't just sell access to customer location data. They failed to prevent that data from reaching bounty hunters and even a sheriff who used it to track people without their knowledge. And then, of course, they sold it to data brokers,

[C]

who are the real bad Agents.

[A]

Oh, yeah. I mean, bad enough. I mean, law enforcement gets it, but data brokers, who knows, you know, they're going to sell it to anybody.

[D]

So. So which Supreme Court justice is the friend of Dog the bounty hunter one?

[A]

That's a good question. Did they, did they say who the. They must have. Let me look and see if I can find this. I mean.

[B]

Yes, whoever wrote the dissent.

[D]

Yeah, who is. Yeah, who, who is most in the pocket of AT&T and Verizon?

[A]

Take a guess.

[B]

Oh, Clarence. It was Clarence Thomas.

[A]

Clarence Thomas.

[D]

Ah, okay.

[B]

Yeah, I thought, surprise, surprise.

[D]

Remember, it's not, it's not selling out, it's buying.

[B]

In fact, the fact that you don't have a unanimous decision that it's not okay to circumvent constitutional protection, that's scary to me. That really is scary. This should have been a no brainer.

[A]

Yeah.

[B]

The idea that I'm going to allow law enforcement or non law enforcement to have information that should only be obtainable via warrant, I mean, then what are we doing here if we're now pretending that I can ignore civil rights as long as I buy the data from someone, that's horrible precedent.

[A]

This was a week that YouTubers won big at the box office. Huge success.

[B]

You saw this coming? You saw this coming.

[A]

It was coming, wasn't it? Yeah, yeah. Back Rooms, which I went to see on the 29th. Opening night, actually. Kane Pixels, the director was there. He's A local boy, 20 years old, got $10 million from a 24 to make this movie. It has already made more than $100 million in its first weekend. It is a huge success.

[B]

So does that put it over Marty supreme all already or not yet?

[A]

I hope so. Marty supreme not. I finally watched that.

[B]

I did not enjoy it.

[A]

It was not a good movie.

[D]

I have not yet seen it.

[A]

Don't.

[B]

It got a lot of buzz, But

[A]

I got nine Academy or eight or nine Academy Award nominations. 1 0. So that'll tell you something.

[D]

Okay, so YouTubers out there start making movies. I want to see Annoying Orange the movie soon.

[A]

No, we already saw Annoying Orange the TV show. That wasn't good.

[B]

There are. But that's what's interesting creators out there.

[A]

It's interesting that you brought that up because that was the first round of Hollywood saying, oh, you know, this YouTube thing's big. This was 10 years ago. Let's bring some of these people and they put them on Nickelodeon and stuff. And it was not good because it

[C]

was the wrong way.

[D]

Yeah, yeah.

[A]

How so?

[C]

And, and well, I, I I, I, I think that you've got to recognize that the new things are built in the new medium and you leave them there. You don't try to bring them into the old medium.

[A]

Although making a movie out of it did. Okay.

[C]

Yeah, well, I think we're going to see more and more of that. I mean we had Robert Tic on Intelligent Machines last week who's very Hollywood. He was, talked about, he was great. And he talked about the AI in the lot conference where we're going to see movies made there. There's one movie that's, it's, it's premiering at Tribeca Film Festival made entirely in AI. So between the distribution side and the making side, it, it's, it's, it's exploding.

[A]

And then what was the other one? Obsession. Is it Obsessions?

[D]

Obsession, obsession.

[A]

Another YouTube movie. YouTuber making a movie and a huge success.

[B]

The YouTubers I really want to see make it. Have you ever heard of Viva la Dirt League? Yes, they're a New Zealand.

[D]

I watch them regularly.

[B]

They started they were just making funny games about gaming stereotypes or funny shorts about gaming stereotypes. But they actually have chops. They have acting and directing chops. They've done a few longer formats pieces and they're fantastic. That's the kind of YouTuber I want to see transition into long format storytelling.

[D]

The other one I'd be interested in seeing would be the Archive in Between. So they do these, they do these shorts that feel like 50 style info info films.

[B]

Okay.

[D]

About the about. They're basically tour tourism for the multiverse where they talk about different, different cities or different parts of the patchwork city, which is where every universe in the multiverse intersects and what you can expect to see. And it's fascinating.

[B]

What's it called again? What's the name?

[D]

The Archive in Between.

[A]

Okay.

[D]

And they also have news alerts talking about, oh, beware this creature from Universe X157 is now terrorizing the neighborhood. Retreat to your home. Follow the instructions in your in Intergalactic Interloper kit. Inter Universal Interloper kit. And they're done so well. And it feels like a 1950s, 1960s civil defense film. And it's all A.I. it's written, I believe it's written by humans, but definitely AI generated images. But yeah, really great stuff that, that

[B]

goes in hand in hand with that story that YouTube took the streaming hours crown back from Netflix.

[A]

Absolutely. Worldwide. YouTube has longer daily average viewing around the world over Netflix. This is huge.

[C]

I'm so frustrated with YouTube these days because the new kind of slop that's being made there is. I'll get a fake script for. You know, I'll see something I like, like. Like Campbell's Soup, the original Campbell soup. Since I used to live next to the Campbell Soup Test Kitchen farm in New Jersey.

[A]

Yes.

[C]

New Jersey. Yes. South Jersey.

[A]

Where all the best tomatoes are made.

[D]

Yes.

[C]

So you'll see something like how Campbell's soup rose and fell. Okay, that's interesting. I'll watch that for. No, it's. It's awful. It's.

[D]

It's.

[C]

It's a slop script. Fake voice, a bunch of images picked up from nowhere. And it's wrong. It's made up stuff.

[A]

It's cheap, though.

[C]

My YouTube is just.

[B]

But it's confidently wrong. Yes, that's.

[A]

AI so you. So what happened is Netflix went down in 2024. It was a 100 minutes a day average down to 93, while YouTube went up in 2025. 99.1 minutes a day on average.

[B]

I'm up way above that.

[A]

More than 99, more than an hour.

[B]

I have YouTube running. It's like background music for me.

[D]

Same here.

[B]

Just keeps going.

[A]

So. And what do you have on there? Is educational and informational stuff or is it music?

[B]

Well, unfortunately, because I do fuzz the data. If I just let it go, I get weird.

[C]

That'll teach you. That'll teach you. You liar, you.

[A]

So you turned on Autoplay. See, I turned that off.

[B]

If I turn on Autoplay, I will get the weirdest content.

[A]

This might be kind of fun.

[B]

I can't do that. No, it's just.

[A]

Get a doobie, sit back, Maybe a beer bong.

[D]

Yeah, there we go.

[A]

Oh, that's right. You're not allowed. Probably.

[D]

No, no, no. But you can. In the incense burner. You hide it, right?

[A]

I'm sorry, have you.

[D]

Have you never been an altar boy? Do you know frankincense?

[B]

The incense burner? You mean the Thor fur?

[D]

Yeah, that thing. Yeah, yeah.

[A]

Gen Z remained YouTube's most engaged age group last year, averaging 111 minutes a day. But growth was strongest among priests. Age. Oh, no, wait a minute.

[C]

Men.

[A]

Sorry. There we go. Age 55 to 64, where viewing has increased 15% since 2024. Daily average YouTube users also increased for women. Of all age groups. South Koreans watch YouTube the most 161 1/2 minutes a day. France recorded the biggest growth, up by a third.

[B]

How much of the South Korean numbers are from mukbangs?

[A]

Yeah, I don't know what that is. Should I ask.

[B]

That's just watching people eat an enormous amount of food. That's what like.

[D]

Yeah, like I, I'm going to try every burger in this burger place.

[B]

That's a mukbang.

[A]

Yeah, that is, that, is that specific to South Korea?

[D]

No, no, there's.

[B]

But it's popular.

[D]

It's popular there.

[C]

But they've, it's on American cable. You can see stuff like that.

[D]

But they've now adopted the word. Yeah, they've just adopted the word for everything now. So somebody's done a McDonald's mukbang.

[A]

Wow.

[D]

Yeah.

[C]

Right now.

[D]

Yeah.

[B]

YouTube for me is mostly like legal. I get a lot of legal content.

[A]

Interesting.

[B]

Auditing courts and, and, and lawyers giving out their, their personal opinions on personal cases. It's fun stuff for me. I like the law. I would have been a lawyer if I wasn't a priest.

[A]

I feel like I should be more consistent my YouTube viewing, because it's good angel, bad angel. Well, yeah, there you go.

[B]

I know they compete for the same eyeballs and stuff, but YouTube and Netflix aren't the same. No, that's true. No, no, they're not at all.

[C]

But YouTube, that's the point, I think Benito, is that YouTube recognizes the culture, is recognizing a different genre of, of entertainment. Netflix tries to recreate the old genres, movies and TV shows shows. YouTube is something different. And the fact that it's bigger says a lot.

[D]

And I have to say, half, maybe half my YouTube consumption is listening rather than directly watching.

[B]

Yeah, that's true.

[A]

Oh, interesting. Yeah, that's what they say. Podcast. Yeah. Are very big on YouTube. And of course, you know, our show, even though we do video is really, there's nothing to see here.

[B]

Just.

[C]

Hey, hey,

[D]

hold on. Like, at least an hour of my day on YouTube is actually me on my bike and I'm just listening to it. I've got, I've got my phone clipped on my shoulder and I'm just listening to YouTube videos lately. A lot of Nate B. Jones talking about AI.

[A]

I love Nate B. Jones is great.

[D]

Yeah. So I, I, I just listen. I just listen to him. It's either that or what else? On the other hand, either Adam Conover or Behind the Bastards.

[A]

Do you think? I bet you everybody's AI. I mean, YouTube viewing is completely unique. Right. You could fingerprint from what they watch. Pretty, Pretty much. Because you're mentioning channels that are huge, that I've never heard of, and yet I watch AI Pro. I mean, YouTube, I keep calling it AI.

[C]

I don't know why.

[B]

Actually, I think this is actually the primary difference between YouTube and Netflix is that everything under the sun is on YouTube.

[D]

Like everything.

[A]

That's right.

[D]

Yeah.

[A]

Netflix is probably a lot of commonality. What people watch on Netflix. We all watch the same. Roughly the same stuff.

[D]

Well, yeah, yeah.

[B]

And so on.

[D]

So on YouTube, you create.

[B]

You curate it yourself. You. You curate YouTube. Netflix kind of does it for you, right? Yeah, I. I went through months where I was binging on air traffic control conversations, tc. I don't know why, but I just started like liking those and then I fell out of love with it and I moved on to van life and then I moved on to legal stuff.

[A]

So.

[D]

Okay, there we go.

[C]

Yeah.

[A]

But are you doing that under your name or do you have some special account that is.

[B]

I create a special account for curating because otherwise too chaotic.

[C]

Yeah, yeah, I'm watching a lot of that stuff on Tick Tock.

[A]

Well, we thank everybody who's watching us on YouTube, 657 people right now. Thank you. Thank you. We do this show on YouTube, Twitch X, Facebook, LinkedIn and Kik. But it's most consistently YouTube is where most people watch. Of course, if you're in the club, you can also watch in the club. Twit Discord. We do the show every Sunday, 2 to 5 Pacific, 5 to 8 Eastern, midnight to 4:30am in the. In Italy. Rob, Robert, thank you so much for being here. We appreciate it. Father Robert Balaser Digital Jesuit Jesuit pilgrimage app is the app that he designed. But there's a whole lot more to Robert. He is on bluescom.

[C]

Some we know about, some we don't know about.

[A]

Yeah. Always a pleasure.

[B]

Always a pleasure.

[A]

We love having you on. It's great to see you.

[C]

Good to see you again. I'm so privileged to see you twice in a week.

[B]

This has been a fantastic week. I get to work with you. It makes me feel like I did back when I was still twit. So.

[A]

Oh, we miss having you here. Miss you living in our basement in the no hole.

[B]

The no hole.

[D]

Yeah.

[A]

K, N O W. Thank you, Robert.

[C]

Now we get through the basement, the Clementine Chapel.

[A]

It's a higher quality basement, but there are by far. There are no beanbags in chapel.

[B]

We have hard chairs.

[A]

Very uncomfortable, I'm sure. I don't know. Do you say good mass? Have a good mass. Have a good Mass.

[B]

Yeah, that works. Break a leg.

[D]

We get it.

[C]

Break a leg.

[A]

Don't break a leg. Just have a great time down there in the chapel.

[B]

I'll bring you down there when you come Over Leah.

[A]

I would be honored. I would love to see that. Seriously, I want to bring Abby with me because it would be so meaningful.

[C]

Oh, yeah, yeah, yeah, yeah.

[A]

Jeff Jarvis, of course, will be back on Wednesday. We'll interview Jeffrey Cannell from Noose Research about Hermes. We talk AI on intelligent machines. His book Hot Type, available from his website.

[C]

Jeffjarvis.com and audiobook finish. I'll do the last pickups on Tuesday. The things I muffed up, which are many.

[A]

Do they stop as you read in it? Do they stop you as you're reading or do they let you kind of roll in the.

[C]

No, the first time I did an audiobook, you just went on from wherever, you screwed up and did it again. Now that means they gotta edit now. Now you screw up and they say, oh, no, you gotta take that again. And they back up and you get, let's say, a three word cue and then you have to pick up. Right.

[A]

Then that's. So it matches.

[C]

Yeah, so it matches.

[A]

Yeah, I've. Because that's. Sometimes you'll hear that in audiobook, you can tell there's an edit. It just changes so dramatically. Yeah. So then you, they want you to hear how you were talking and then just kind of.

[C]

Yeah, and then I did the, the pickups from when I screwed up. Things like I said could instead of wood. That kind of stuff.

[A]

Right.

[C]

And don't you get to say, hey,

[A]

I'm the author and could is fine

[C]

to some extent, but they get pretty. When they were kids. Cases. There were cases where I said, this doesn't make sense in audio. I'm changing this. They said, okay, it's your book, you can do that. But no, generally they want it to be. They want to be able to say,

[A]

you'll hear that on audiobooks. They'll say if you're listening instead of if you're reading, which is what the text said. Yeah, that makes sense.

[C]

And I hate saying I still. It's still a book. I won't say it's audiobook for listeners versus readers. I won't say that. But, but they'll, they'll play what I screwed up and then right after that I have to say the same thing corrected so that I think I've heard how my tone was right then.

[A]

And you refuse to read the part at the beginning about what at the

[C]

end where it says, no AI company may take this in this universe or any future universe because you're evil bastards. I won't read that part. So they have to find other. They, they, they have Plenty of other voices who've read that because.

[A]

Because like, like me, you and embrace the AI.

[D]

Yeah, I do.

[C]

Let him. Yes. I want to be discovered there.

[A]

I think we made a deal with somebody that if people wanted AI, wanted Twits content for AI, they would go to this company and license it. But who's going to do that? It just seemed odd to me. But I guess we give you a legal.

[C]

Is that Pro Rata?

[A]

I don't know how it works. I should ask Lisa.

[C]

That was the name of the company.

[A]

It's. Oh, I don't know who it is. No, it's Pro. I mean, I don't even know if it's on our page. Is it on our page? Does it say like, if you'd like to license this content, please contact Pro rata. I don't know, it should say that somewhere. But I mean we could put all

[B]

the Twit content in the archives over here we have storage vault now.

[A]

So like what not to do? Like how, how not to be. Right. Thank you very much, Jeff. Thank you. Joey, congratulations on the new gig@netfoundry IO developer advocate. Working on some what AI thing there. What are you working on?

[D]

What I'm doing is I'm promoting a lot of the new AI tooling. So it's built on top of OpenZD and it is for agents to talk to LLMs, agents to talk safely, MCP servers safely, safely. And agents to talk to other agents. All zero trust. Basically, you're either using zero trust or you're going bust.

[A]

Zero trust or bust.

[D]

Zero trust or bust.

[A]

Joey Devilla. You should write a accordion song for that. Joey's a great accordionist as well.

[D]

There we go.

[B]

I mean, you've got an LLM to generate that song, Leo.

[A]

Well, no, I, I believe in humans. I bring in. We brought in a chorus or do that six, seven song Organic.

[D]

As a, as a musician, I am a protein chauvinist. Let the meat make the music.

[A]

Carbon based life forms for me. Thank you very much. Thank you all for watching. A special thanks to our Club Twit members who make this show possible. Yes, we have Advertising covers about 70%, maybe 60% of our costs. In order to do these shows, we need your help and boy, you've really stepped up and we appreciate it. Twitt TV Club Twit. If you're not amazing, if you're not a member, you will not be able to see our coverage of WWDC tomorrow. All the keynote stuff is I don't like paywalls. But we have to put it behind the paywall so we don't get taken down by Apple. So if you want to see it, Twit TV Club Twit, you also get ad free versions of all the shows, special programming we do only for our club members, and a lot more. But mostly you get the good warm and fuzzy feeling of knowing you're supporting independent journalism beholden to none. Except you are the users because we're all users here. Thanks, Joey, Robert, Jeff, thanks to all of you for joining us. We will see you next time. And as I've been saying for 21 freaking years, another twit is in the can. We'll see you later. It's amazing. Twit, baby, doing the twit. All right, Ryan Reynolds here from Mint Mobile. I don't know if you knew this,

[C]

but anyone can get the same Premium Wireless for $15 a month plan that I've been enjoying.

[A]

It's not just for celebrities. So do like I did and have

[C]

one of your assistant's assistants switch you

[A]

to Mint Mobile today. I'm told it's super easy to [e-mail]

[B]

Switch upfront payment of $45 for 3

[A]

month plan equivalent to $15 per month

[D]

Required intro rate first 3 months only,

[C]

then full price plan options available.

[D]

Taxes, extra fee, full [e-mail] Some follow the noise. Bloomberg follows the money.

[A]

Whether it's the funds fueling AI or crypto's trillion dollar swings, there's a money side to every story.

[B]

Get the money side of the story.

[A]

Subscribe [e-mail].