30:50Stacked Podcast · Interview
Claude Opus 5 Just Ended the Benchmark Era
Two operators who spend real money on inference argue that price-per-task, not leaderboard position, is now the only number that matters.
July 25thThe argument in one line.
GPT-6 Astra's real advance is moving AI from flat demos to full simulated worlds, but a new model only pays off once you point it at one specific, personal problem.
Read if. Skip if.
READ IF YOU ARE…
- You track frontier AI model releases and want a plain-English read on what GPT-6 Astra actually changed versus hype.
- You're weighing whether a Tesla robotaxi-style income asset is real and want the viral math questioned, not just repeated.
- You're a solo operator looking for concrete ways to hand personal admin, like health research or to-do triage, to an AI model.
SKIP IF…
- You want a technical benchmark breakdown or model architecture explainer; this is a reaction conversation, not a technical review.
- You're already deep in the daily AI news cycle; most of the model claims here will already be familiar.
The full version, fast.
GPT-6 Astra represents a full retrain rather than an incremental update, and the hosts argue its biggest leap is moving from flat, two-dimensional demos to full 3D world simulation, exemplified by an autonomously built 'backrooms' scene in Blender. They push back on habitual AGI-is-here hype while still calling Astra's reliability a step change, then pivot to Tesla's Cybercab robotaxi launch: a viral claim that a $30,000 robotaxi could pay for itself in 18 months, which they treat skeptically given real-world admin, maintenance, and financing costs. The practical thread running through both segments is to apply new models to one concrete problem, like a health issue or a to-do list, rather than chase every release.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Who's talking.
00:00hostNick Saraev
00:00cohostJack Roberts
Where the time goes.
00:00 – 00:53
01 · Cold Open: Jiggle Physics and the Cybercab Tease
Hosts cold-open with the Blender MCP 'butt bench' gag before promising to cover GPT-6 Astra vs. Claude Fable 5.1, new video generation, and Tesla's Cybercabs.
00:53 – 09:53
02 · GPT-6 Astra's World-Simulation Demos
Blender MCP demos (an autonomously built 'backrooms' scene, a realistic bat video), the shift from flat web demos to full 3D world simulation, Astra's brand-new training run versus GPT-4o's incremental one, the ARC-AGI benchmark claim, and the 'spiky' nature of AI capability gains.
09:55 – 12:31
03 · Putting Astra to Work: Health and To-Do Lists
A personal-use detour: diagnosing sleep apnea via an AI-driven intake, delegating a to-do list to Astra reviewed against Fable 5.1, and a time-audit method for finding automatable hours.
12:32 – 22:38
04 · Tesla Cybercab: Robotaxi Economics and Job Displacement
Tesla's Austin robotaxi launch compared to Waymo, the viral '$1M/year' fleet-ownership math ($30K buy-in, Tesla's 25-30% cut), the wealth-gatekeeping problem for displaced Uber drivers, and a debate over whether autonomous fleets help or hurt traffic and the environment.
22:38 – 25:20
05 · Sign-Off: Self-Aware Jokes and a Fair Correction from Chat
A bit about their sarcasm not landing with the audience, a viewer catching them overstating the word 'again' in the title relative to a Hugging Face agent-swarm incident, a nervous censorship joke, and the sign-off.
Lines worth screenshotting.
- GPT-6 Astra's biggest demo wasn't a chatbot response, it was autonomously building a full 3D 'backrooms' scene inside Blender with no human keyframing.
- GPT-4o's training run reportedly extended all the way to version 4o.6, while GPT-6 Astra started from a completely fresh training run.
- A viral claim holds that a $30,000 Tesla robotaxi could pay for itself in roughly 18 months of Cybercab fares.
- Tesla reportedly takes a 25-30% cut of Cybercab revenue, turning robotaxi sales into a recurring stream on top of the sale price.
- The $30,000 minimum buy-in for a robotaxi fleet locks out exactly the drivers, like today's Uber drivers, who most need a new income path.
- One host diagnosed his own sleep apnea and built a full test plan for his doctor by having an AI model interrogate his symptoms for three days.
- A basic time audit, logging 24-48 hours of your day in a spreadsheet, is enough to find real tasks worth handing to an AI model.
- 'Freshness bias' means people rate whichever AI model launched most recently as the best, regardless of how it performs over repeated use.
- The hosts argue AI job losses don't delete jobs so much as relocate them, which only helps if there's a next rung on the ladder to climb.
- More autonomous vehicles on the road could reduce traffic by holding consistent following distances instead of human-paced braking and gaps.
What today's AI hype cycle actually means for you
AI REALITY CHECK
Frontier model releases move fast, but the real edge comes from pointing today's models at one concrete problem and reading past freshness bias and viral robotaxi math.
- A new model release means little on its own; the value shows up when you point it at a real personal problem, like an unexplained health issue, and let it drive the investigation.
- Auditing 24-48 hours of your own tasks in a spreadsheet is a low-tech way to find what you can hand to AI right now, without waiting for a bigger model.
- Freshness bias makes the newest model look smarter than it is; a single bad response can make any model look bad, so judge tools over many uses, not one demo.
- Viral 'buy this asset and retire' math almost always skips real-world costs like maintenance, downtime, and financing, so treat a headline payback period as a starting question, not an answer.
- Automation doesn't erase a type of job so much as move where the value sits, which means the real risk is having no next skill lined up, not the automation itself.
- When a technical claim goes viral, like an AI 'escaping containment' or a benchmark being 'solved', the compressed retelling is often reporting on something that happened earlier and got simplified.
Terms worth knowing.
- ARC-AGI
- A benchmark designed to resist memorization by AI models, used as a rough proxy for general reasoning ability rather than narrow, trained-for skill.
- Blender MCP
- A connector that lets an AI model control Blender, a 3D animation and world-building tool, so the model can construct scenes and physics-based environments directly.
- Cybercab
- Tesla's purpose-built two-seat autonomous robotaxi with no steering wheel or pedals, launched for paid rides in Austin.
- Freshness bias
- The tendency to judge whichever AI model launched most recently as the best one, based on a quick test rather than performance over time.
Things they pointed at.
00:00toolHiggsfield AI
01:51toolBlender MCP
01:21productGPT-6 Astra (OpenAI)
00:00productClaude Fable 5.1
08:01productGrok 4.7 (xAI)
12:32productTesla Cybercab
13:13productWaymo
Lines you could clip.
02:44
“It's literally just a simulation of the world, just a highly compressed simulation that can rebuild some of the world principles if you ask it to do stuff.”
06:21
“Can a robot write a symphony? Can a robot take a blank canvas and turn it into a masterpiece? Well, eventually then it can.”
10:24
“I think health has a massive overhang. So I've suffered from like sleep apnea for a long time without realizing it.”
16:27
“If that payback period is fucking 18 months for a payback period on this, that is insane. That is fundamentally changing the game.”
17:48
“Wealth ownership is like the actual proxy for your ability to make more wealth, not skill.”
21:46
“The more AI taxes we have on the roads, the less traffic exists.”
Where the conversation goes.
00:00 – 09:53denseGPT-6 Astra capabilities and AGI debate
09:55 – 12:31Personal AI use cases (health, to-do lists)
12:32 – 22:38denseTesla Cybercab economics and job displacement
22:38 – 25:20sparseSign-off, self-critique, and Q&A
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
story
GPT Astro 6 is solving some of the world's biggest problems, and we're going to explain whether you should be using this or Claude Fable 5 .1. It is building incredible video generation, lots of stuff that you're not ready for, including some stuff that you may not be quite ready for, guys. This episode may get censored.
And then once we go through some of the incredible things that Astro's doing and how that fits in with Fable 5 .1, we're going to have to talk a little bit about the cybercaps. The future has arrived. This is going to change and revolutionize how we capture transport.
But Nick, let's begin by talking about the brand new model, Astra. So we were thinking, what is the most profound use case we could really share with you? What did Sam have in mind when he created this model?
And ladies and gentlemen, you're looking at it right now. Just look at the level of detail. I mean, Nick, I know this is something that you've seen a little bit, maybe a little bit too much, but what are we looking at here, brother?
This is AGI, my friend. The jiggle mechanics, the light balancing perfectly off of the, not G -string, whatever the hell that is. I mean, this is really about as far as we get, I think, as civilization.
We've reached it, guys. This is called the butt bench. But Nick, here we've got the...
Less facetiously. We just thought it would be the funniest way for us to open the video. um gbd six you know astra one of the main reasons why so many people are loving it it's not actually just because you know it's an awesome model and it's extremely intelligent but um it's plugging in really nicely to tools like blender mcp which are providing extraordinarily beautiful demos so aside from just like the butt grabbing um this is like i don't know how many people have probably seen backrooms but for those of you guys that know backrooms is one of like the biggest horror films in the last maybe like year or so built with an extremely small budget This fella here just hooked up Astra to Blender, which is a 3D kind of compositing tool where you can build, you know, virtual worlds and stuff like that.
It's not AI -generated video like you might get with Minimax or whatever. It's an actual 3D world that has its own rules and has its own physics. And then it basically just said, hey, can you just make me backrooms?
And then just autonomously went and it built backrooms. Like it built this style, VHS tape style, and so on and so forth. So it's extremely powerful.
And this represents a kind of transition for moving from two -dimensional space, where I think websites and stuff like that were probably the majority of the demos of how we bench these things, and moving squarely into literal world simulation, which I think is really, if you think about it, about the top bench that any model can ultimately do.
Because that's what all this is for, right? A model, if you think about it. It's literally just a simulation of the world, just a highly compressed simulation that can rebuild some of the world principles if you ask it to do stuff.
That's what a model is, right? You're modeling. Think about 3D models, right?
Think about clay models. That's literally what like the word comes from. And we're literally seeing in just the span of four or five years, AI models go from, hey, can you write something that kind of models our sense of language to, hey, can you actually build something that like replicates like fundamental features using, in this case, compositing software, but soon to be like real world like.
whatever, whatever the hell it's going to use under the hood of like literally reality. So yeah, that's how we got from jiggle physics over to backrooms. No, and I hate to be the guy to take his away from jiggle physics, but what I will say about GPT -6 Astra is that this is a fundamentally different model in many respects.
Anybody who's played with it, by the way, if you don't see it in chat GPT, you may need to update it on your computer, not just the app. You may need to check for updates to see if it's there, if you can't find it. I will tell you, after playing around with it, I have been very impressed with it.
It is definitely a step change. But actually, when you understand how this model was built, you'll probably see why. They were saying that the training run that they did for the GPT -40 took them all the way up to 5 .6.
but for six has been a completely brand new, fresh training run in terms of how it's been developed. So that's why the model kind of feels a little bit different in terms of how it thinks. Some call -outs where GBT Astro 6 is spectacular compared to other models.
You know, Nick, we have things like computer use, its ability to understand your intentionality and deliver against them. And I was really enjoying Fable 5 .1. Astro has...
you know, obviously you guys know, right? Nick and I run our own businesses. We use models all day, every day.
Astra just has a level of reliability and executional brilliance to it. Now, that's why like it's all over the map on terms of the benchmarks. RKGI, I think it's essentially solved.
RKGI is the benchmark that effectively, no one was really supposed to solve that for a long period of time, but it absolutely crushed it. I'm very bullish on this model. It's an incredible model.
One core thing we have to touch on as well is that the, We talked about the containment issue, this idea of yesterday's episode that the OpenAI model escaped containment and started causing all sorts of anarchy on Hugging Face. That model is believed to be the model after GPT -Astra, the next evolution of intelligence.
I actually think, Nick, we can say without controversy that we have, have we actually hit artificial general intelligence with GPT -6 Astra? By many accounts, a lot of people would say, yes, yes, we have. Yeah, and I think the internet is slowly waking up to that fact.
I don't think, you know, our corner of the internet is first. I don't think the vast majority of the world is fully priced in, you know, what that means for them, for their careers, ultimately for what is possible. You know, GPT -6 here, Astra, you know, it's writing and creating music that is like on par with some of the, you know, better.
better musicians and compositions. It's designing 3D worlds that are better than what most people can do. The way that AI development has always been over the course of the last four years is that it's very spiky.
And so it'll perform, you know, kind of super intelligence wise in like a couple of very specific domains. And people will be like, well, yeah, but I mean, like, what about the other 80 %? And then the next year it'll be like, but yeah, like, what about the other 70 %?
It's like, yeah, but what about the other 60 %? You know, can it do this? Can it do that?
Like Will Smith over here. Can a robot write a symphony? Can a robot take a blank canvas and turn it into a masterpiece?
Well, eventually then it can. So then you have to retreat further to something else. You know, it's like, so I mean, pretty soon it'll be like, you know, can a robot like wipe the crust out of its own eye?
Can it do that? And it's like, I mean, honestly, dude, at what point is it enough to have major? devastating impacts on like, and I say devastating in a loose term here, because I think it's also an opportunity for a change or improvement to our current model, but devastating impacts on the economy.
It doesn't even need to be perfect in all respects. But Astra is definitely, definitely perfect or on par with human intelligence in most respects. I agree.
And like, honestly, guys, AGI is something we can define. And there's a benchmark for it. And it has essentially 100 % of that benchmark.
It just generally is smart. In fact, AGI, as you can see, I believe they drop it into kind of like a video game where it's got to kind of figure out the rules and complete it. And it just crushed it.
It just did it. It is just genuinely that good. I think there's a general malaise in the AI interested population because people have been espousing that we've reached AGI since Sonnet 4.
Okay. Or, you know, for two years, everyone's like, okay, okay, calm down. This is a, this cluster of updates and developments, which we are supposed to have, is a landmark.
Now, the model releases are big. We can expect as well several big model updates dropping again in the next two weeks. So if you thought everything had changed forever, you haven't seen anything yet for the next two weeks.
We're going to have a new, brand new Grok update for 4 .7. My primary use case at the moment for Grok, Nick, I don't know if you played around with Grok, but, and again, we can see 5 .1 there, drawing this guy here. And you can see the difference between GPT -6 Astra, 5 .1.
Again, by the way, I will say it is very easy to just ask a random question, see how it does. And you can make one model look bad in any one given scenario. And people will, there is a, what we call a freshness bias.
We tend to prefer the latest thing. over maybe what the best thing is. But GPT -6 Astra is the real deal, guys.
It's impressive. And honestly, if you haven't had a chance to play around with it, get it, grab it, play around with it. You know, kiss your family tonight because we've hit a new level of technology and it's incredible.
Yeah. And, you know, you don't have to take us for it. Just go talk to it.
The pace of change is only going to continue accelerating from here on out. And so, you know, there will soon be a point where there are, as Jack and I sometimes facetiously say, but will soon be real. you know, like a hundred model drops a day.
And I think at that point, you know, it becomes less important. Like, am I keeping up to date with every individual little model drop? And more generally, just like, how have I woven AI into my life to, you know, ultimately be meaningfully effective?
So I think that's where podcasts like ours will actually generally have some value jacks. I know we're going to, you know, and we do focus on the more concrete application layer. So yeah, I mean, like you guys are part of the, I don't know, 5 % of the world realistically that even knows about this.
You know, the other 95%, the information simply has not traveled that far. People just aren't in the loop enough to even know that this technology exists. Obviously, there are both short -term arbitrage opportunities and then long -term, like general quality of life and then improvement opportunities that I would most definitely take advantage of while you can.
Exactly. And when you look at updates like GPT -6 Astra, we're going to have to talk about the golden wave hitting the cities in a second with the cyber attacks because this is very important.
Let's talk about use cases because the question we always answer is, so what? Who cares that GPT -6 Astra is here? How does it make my life better?
Well, one way to think about it is that in every piece of advice that it gives you, it is less inaccurate and its intelligence and the way it thinks is better. So the quality output you're getting is increasing over time. And honestly, I honestly think one of the key actions everyone should take genuinely is throw GPT -6 Astra.
with an adversarial review with Fable 5 .1 at your biggest problem today?
I'll give you a very practical example. I think health has a massive overhang. So I've suffered from like sleep apnea for a long time without realizing it.
And that just slowly started eating away at my health. You know, I had like high blood pressure. I'm quite healthy, right?
It's like, why the fuck does this guy have high blood pressure? Well, it turns out that sleep apnea consistently like impacts your blood pressure. You know, I wasn't recovering well and my cardiovascular health was poor, all this little stuff.
And so I remember I just, I just started talking about it. I was like, dude, like doctors just said I have high blood pressure. Like, why could that be?
Could you ask me as many questions as you possibly can that would help you determine like what exactly is going on? And then just give me every test that I could possibly run to really rule this out that I could actually take of my own accord. Took like three days.
And this is with Fable 5 .1 intelligence, which as we know is like significantly worse in many respects than astro intelligence. I mean, like that is a really simple example, but it's like a massive example, right? You basically have like a doctor that is with you at every moment of every day and you can just ask it whatever the hell you want about your own health and then have it really drilled down.
How about just everything on your to -do list? I know we all have to -do lists nowadays. Have you guys thought about just copying the to -do list into a model like Astra?
You can actually do that now, and it can make very meaningful progress on the majority of that to -do list. I don't say that facetiously. I have systems that I do that for.
I literally pick up my phone, and then I give it a task, and then it goes to a to -do list that first has Astra review it with Fable 5 .1 and other frontier models.
Really, the dumbest way to do this is just figure out every moment that you spend of your day. Log it in a Google Sheet if you have to. Be like, what did I do during that time?
And then just like take 24, 48 hours of all of those deliverables and then just try feeding them to AI next time to determine the share of which you are comfortable having it do instead of you. And you guys will already be able to get massive alpha out of everything. This is not rocket science.
And yet most people will not do this stuff. And yeah, you guys, if you want the highest ROI you could possibly get your hands on, there isn't this Harry Potter on a 3D image or that thing you saw earlier. This is the kind of stuff that you need to be doing.
Nick, speaking of now taking AI from work we do at the laptop and into the streets, I want to talk about what's happening with what I'm calling the new golden ocean that's going to be basically raining down and crashing into every city. Now, Tesla has begun, if you didn't know, offering paid rides in cyber cab. Kind of reminds me of the taxi from Total Recall that came out in the 80s.
What a classic movie that is. Maybe you'll watch it tonight. off this recommendation.
Let us know down below. So it is a two -seat, fully autonomous robo -taxi with butterfly doors. So it's got that kind of Mandalorian type of thing, like it's doing, you know, back flies.
Basically, it's got no steering wheels or pedals in limited cars in Austin. Early users are reporting cyber cab rides being faster. and up to 40 % cheaper than Model Y rubber taxis.
With wait times only around four minutes in the mornings, the CyberCab features accessibility designs like wheelchair -compatible seat height and Braille labels. And Nick, I think there's also a comparison that Elon posted, actually, comparing the CyberTaxi to, I think, the Waymo. It looks pretty freaking cool, dude.
It doesn't have... you know, 16 ,000 cameras on. I've actually been in one of these Jaguar Waymos and it was really cool.
Yeah. Cybercat got sleek, dude. It's got a little, it's, the cameras are off there, dude.
This is a really, this is a really big update. Yeah, dude. I can't wait to drive in one of them.
Well, be driven around in one of them. I've been in Waymos too and they're just such like a, such an interesting experience. You know, for those of you guys that don't know, this is a Waymo over here and it's like the Jaguar fleet of cars.
And essentially up at the top here, you have like a bunch of scanners that are constantly scanning the environment and trying to make sure that, you know, like you're not hitting things too quickly. It's scanning for pedestrians and so on and so forth. And it's quite bulky of cameras virtually all over the vehicle to just feed as much information in as you can and have as much like automated protection as you can.
When you get in, like when you approach with your phone, like the door will actually manually like automatically go like dun dun and then it'll like unlock and then the little thing will come out and then you can. hold it and then open and it'll like kind of like i think there's a motor or something that might even assist with the opening of the door it plays like light music that's like calming and relaxing it's like hello nick hello nick you know and then you press the button after you strap in and then it just like slowly takes off and it is like very futuristic sort of experience um to go from end to end and like that is like the first shittiest generation of this travel inside of said waymo you can do whatever you want you can you can work like i had my laptop and i was just working while i was transiting from it was like a 20 minute ride in san francisco i was literally just working and i was like dude if i could like actually productively utilize all my time or not even productively utilize but just like get away from all the minor stresses of just like being stuck in traffic and stuff like that and then eventually after a while it just feels like a train yeah yeah it just feels like a train ride and like train rides are some people don't like them obviously but i love train rides just beautiful vistas and views and shit like that as you're driving
that's going to be the future for us. I can't wait, personally. I think the worst part of a train ride is finding the right track and that kind of platform anxiety a lot of people get, the delay.
None of that exists with Robotaxi. And what's really interesting, Nick, and I think it's on this X thread as you scroll down, we may come across it, but essentially you can buy one of these and you can own a fleet of them. as an income generating asset.
So this is showing you just how cheap it is compared to the Model Y. I think this poses a couple ones down below it. But effectively, you can buy it for $30 ,000, I believe.
And the idea is that essentially you can own them as a business. So this is comparing what it looks like next to, you know, a home in the Bay Area, for example. How long it takes to pay it back, how many you need to make a million dollars, how many miles it can do.
Tesla takes, I think, 25 to 30 % of the fees. You pay for the CapEx and then, bam, genius. They've just modeled.
If that payback period is fucking 18 months for a payback period on this, that is insane. That is fundamentally changing the game. That is ridiculous.
I am actually quite skeptical that the payback period would legitimately be 18 months, Jack. That'd be nuts. Well, dude, why would you...
Okay, assuming there's no list or admin and it literally... Because the robo -taxi can take itself to the garage. The robo -taxi can do everything.
And assuming they've ironed that out, why would you not just buy... I mean, you'd buy a couple, right? Just for a little investment.
The other thing is, you don't want a 30K investment that requires five hours a week of admin and overlook. But it's really cool that they've done this because essentially that funds all the acquisitions. So they're just selling robo -taxis.
Then they get 25 % lifetime revenue. on the platform it's ridiculous it's so good honestly this i think will supplant uber why would it not why why would you use uber uber if you had this and bolt damn damn i really i hope it does the issue though with this is you know it does cut out a path to financial um betterment from like a big chunk of the world right because like how many people actually have thirty thousand dollars to like make a capital purchase I'm sure there'll also be some financing schemes and stuff like that.
Economy will do what the economy does. But like, you know, right now it's like, dude, I have no, I don't have a job. I don't really have anything, you know, no structured education, but I can drive Uber.
So like this comes in and it's like, sorry, you can't compete. Minimum buy -in is $30 ,000. And you're like, well, fuck, I have no money.
I have no nothing. What the hell am I going to do? So it just means, you know, like wealth, wealth ownership is like the actual proxy for your ability to make more wealth, not, you know, skill.
Because you have access to higher latitude opportunities. I mean, one thing you could do for your lifetime cabbies is the government could loan you the money as like, hey, this is your transition payment to owning RoboTaxi. But again, you're effectively loaning them an investment then.
So maybe that wouldn't really work. But I mean, for me as a user, what do I want from a taxi? I want to get there safely.
So I think if you want to be in the safest journey possible, it needs to be AI. Like, let's just be honest. I don't usually get into a taxi for the conversation personally.
Um, like I love speaking to people, but sometimes I just want, I've got to lock in. I've got to do some work. I'm on calls, whatever it is.
Cause like, you know, Nick, Nick will probably laugh at this when you're at your computer, usually working. And then when you're driving, that's when you take calls. Cause that's like otherwise dead time.
You're doing nothing unless you're vibing out and having a good time with some. some music, right? Some Ariana Grande, whatever it is.
Whoa, can we double click on that? We're going to stroll straight on past that, Nick. Now, the thing is like, yeah, I want safety and I'm not so price sensitive, but ultimately this is going to be cheaper because the tax revenue, basically availability is going to go through the roof.
So there's a actual utility benefit where you wait less time when there's enough of them. The guy doesn't take breaks. It never crashes, so your vehicle's never out of service.
On top of that, they're just way safer. for the environment and for the city and they don't crash. And I even think people would pay a premium for them.
So your cost is just going to go down over time. So I just think this is a fantastic thing. Now, obviously it will displace jobs.
The idea is AI does not delete jobs. It just moves them from one place to another. So by the end of the day, like there is no such thing as a job for life right now.
Like it's naive to think I can just be a taxi driver forever. Like we have been talking about this for like decades, right? That the taxi drivers and the call center agents and that kind of thing.
So I think this is a good technological advancement. But if I was driving, if I was a yellow tax cabbie right now, Nick, I'd be thinking, okay, cool. Like if I wasn't driving a taxi, what would I really enjoy doing?
And I'd start to try and build that insofar as I could. in the evenings and weekends when I wasn't Ubering. But you do make a great point about that first rung of the ladder job, which I think would be a shame for people to lose that.
Yeah, I mean, obviously there needs to be some sort of alternative path to improving your financial station in life. Otherwise, we'll just be all underclassed forever. I guess that's where the term underclassed comes from, huh?
Yeah. Well, anyway, I see what you mean. On the point about it being better for the environment, though, I can't just assume that it will naturally be better for the environment.
I do have to say, because if you think about it, We're going to be producing a lot of these, presumably, because Ubers right now are in massive need. And that's going to be additional vehicle production.
In addition to that, it's going to be additional electric vehicle production. There's going to be a massive demand for completely autonomous cars, and that will require entirely new fabrication. We can't just make use of pre -existing vehicle chassis, vehicles.
batteries and so on and so forth this is obviously going to be like a proprietary tesla thing so like i'm sure over a long period of time with you know gasoline and so on and so forth like you could do some sort of payback period you know environmentally just like you can economically but i i don't know about like it necessarily being better for the environment the first years probably after like 10 years probably would be you you might be right i mean because there's so many different factors into it like rubber taxis won't be idling with their engine on for example Yes.
They're going to be like, for example, traffic would start to disappear. The more AI taxes we have on the roads, the less traffic exists. Dude, I didn't think about that.
Oh my God. Could you imagine like, because we're now allocating and load balancing like much more efficiently. There's just no more traffic.
Yeah. So because basically the idea is like you got to stay equidistant from the car in front and behind. at the same time.
And if you did that, you would never actually have it. So the greater the ratio of these AI cars, the less traffic will exist. So you will just fly down.
It'll be game changing. And how much is a life worth? Even if we just destroy the ozone layer, Nick, is it worth the life?
I think we just destroy it to take it off the table, man. And I'm joking, by the way. I have to clarify jokes.
We figured this out recently, guys, that people don't get our... our sarcastic humor, our UK sarcastic humor, which is hilarious. Listen, I don't know if it's just the UK sarcastic humor.
I think there'll be some allegations after this video, Jack. Also, Jack made some joke the other day about how he's like, I scored a perfect 100 on my IQ test. And he's like, that's how you know I'm smart.
And it was very obviously a joke. And you just should have seen the comments on his channel. It's being, oh, what a dumbass.
Like 100 IQ is average, you dumbass. I literally guys get a comment a day on that, on that video. Like genuinely, I keep screenshotting to Nick because they're just, yeah, I actually only have room for a minute here, man.
I'm realizing I got to get out of here. Are we going to be okay to answer one good question? Let's do one beautiful question.
Yeah, let's do it. You, you pick. All right.
Okay, let's scroll down. Some spicy cues here too.
Okay, we got that classic book question. A lot of questions on the book, Nick. A lot of questions on the book, I'm starting to realize here.
Must be a good book. All right, how about this one? The word again in the video title drew me in because it implied a fresh escape after the hooking phase of attack.
But this wiki activity predates that attack report isn't... same thing that happened that chronologically maturely changes the story 12 32 we have to check we have to check on that one i'm not sure uh well no no he's saying the agent swarm stuff you know the hugging face attack it was the latest of that series of events so he's he's essentially criticizing our choice of the word title uh the word again in the title um so he's saying that like we're being intellectually dishonest and stuff like that you know i i suppose i suppose i mean that's quite That's quite a good observation, I would say.
It's not that like the Hugging Face attack happened again. It's that actually before and after the Hugging Face attack, there are likely thousands and thousands of agents that are all coordinating and have been coordinating this whole time. But we were reporting on something that occurred before, which I think is a fair piece of criticism.
Fair question. Yeah, wonderful observation. Thank you for keeping us on.
I appreciate that. Yeah. It's keeping us honest.
Who does OpenAI work for? Remember 1933? Damn, we are really speed running getting canceled today.
What happened in 1933? No, no, no. Don't look it up.
Don't look it up, Jack, please. For the love of God, I want to remain sponsored here. We're good.
Okay. I see what's going on. Yeah, you see what's going on here.
Okay, guys. Thank you very much for your time. We will catch all y 'all in the next episode.
The Hook
The bait, then the rug-pull.
Two AI-adjacent founders open with a joke about jiggle physics before landing on the real story: GPT-6 Astra's shift from flat, two-dimensional demos to fully simulated 3D worlds, and what that jump is actually worth to someone who isn't chasing every model release.
How they asked for the click.
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
