Apple and Meta Are Destroying the Reason You Pay for AI
Meta gave away a frontier-class open model, Apple supplied the hardware to run it, and the case that follows is that AI ends up a device you buy once, not a subscription you rent.
Posted
today
Duration
Format
Essay
educational
Views
15.4K
412 likes
57 · 43
Big Idea
The argument in one line.
Meta commoditizes AI models by giving them away, Apple and Nvidia sell the hardware to run them, and together they weaken the reason anyone pays monthly for closed frontier models.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You pay $20 to $200 a month for Claude or ChatGPT and want to know whether a local model on your own machine will replace that bill.
You run a small business that handles client financials, notes or tax returns and doesn't want that data pasted into someone else's server.
You follow AI strategy and want a business-model read on Meta, Apple, Nvidia and the closed labs rather than a benchmark rundown.
You are deciding whether to buy a Mac mini or Mac Studio for local AI and want the memory argument behind it.
SKIP IF…
You want a hands-on tutorial for installing or running a local model; this is a strategy argument with no setup steps.
You need neutral analysis; the video argues one future with confidence and its predictions about Apple are opinion.
TL;DR
The full version, fast.
The video argues that AI is heading toward a one-time device purchase instead of a subscription. Meta, whose business is ads, treats models as a cheap bun for its hot dog, so on October 15 it released the open-weight Muse Spark Open, tuned to run on Macs with Apple's MLX. Apple lets others spend first and then owns the device, and Nvidia is shipping 128 GB machines and funding open models. If free models trail paid ones by only months, closed labs priced near a trillion dollars become vulnerable. The author predicts Apple either buys one cheaply or just sells the hardware.
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.
To a company that sells ads, an AI model is the bun on the hot dog, so Meta wants it as cheap as possible and gives it away.
Meta's best model, Muse Wonder, stays closed; the open release sits a tier below it, which is exactly how a commoditize-your-complement play is supposed to work.
Running the full Muse Spark Open model takes about 493 GB of memory, so the model is free but the machine to run it is still a $10,000-plus problem.
Apple chips share one pool of memory across CPU and GPU, so a Mac can hold models that a 32 GB gaming card cannot.
A year-old paid flagship scores 16 to 28 on the Artificial Analysis index while a free 27B Qwen model that fits a 24 GB Mac scores 34.
Open models trail the best paid models by about four months on average, according to Epoch AI as the video cites it.
Apple has rarely been first (Apple Pay, big iPhones, Apple Watch) and wins by arriving later with the better device.
Meta wins AI distribution by reaching the most people, while Apple wins by reaching the most valuable ones, since the US is 54% iPhone.
A local model needs no subscription and no internet connection to think, so after purchase the main cost is electricity.
A coach, accountant or consultant holding client data gets a privacy case for local AI that no cloud plan can match.
When the public, the government and retirement accounts all own pieces of the closed labs, everyone is rooting for the labs to win.
All three of Sam Altman's reported AI endgames keep the most powerful AI inside a handful of labs; none has the model running on your desk.
DeepSeek's free model helped knock about $589 billion off Nvidia in one day, which shows what an open model can do to a valuation.
At a $2,500 to $6,000 yearly subscription bill, the same monthly payment could finance a Mac mini with 128 GB of unified memory.
Takeaway
Why free models and your own hardware change what AI costs
WHAT TO LEARN
Free open models plus memory-rich hardware could turn AI from a monthly rental into a one-time purchase, and the remaining question is whether local models get good enough.
Companies that earn money elsewhere, like Meta with ads, give models away to commoditize them, so expect the free tier to keep improving whatever the closed labs charge.
Open models trail paid ones by only a few months, and a 27B free model on a 24 GB Mac already beats flagship paid models from a year ago on one index.
Memory, not raw compute, is the bottleneck for local AI, which is why unified-memory Macs and 128 GB Nvidia machines matter more than gaming cards.
The full Muse Spark Open model needs about 493 GB of memory, so the free model is real but the hardware to run all of it is still expensive.
Apple's pattern is to wait, let others fund the experiments, then own the device, which explains why its weak AI start may not decide who wins.
A model on your own machine keeps client notes, tax returns and financials in the room, a practical reason to run local that does not depend on cost.
Closed-lab valuations rest on people renting intelligence, so a free model that is good enough for most work puts pressure on them the way DeepSeek did on Nvidia.
Compare your yearly AI subscription total, $2,500 to $6,000 at the high end, with the cost of financing a Mac mini or Studio, and watch for the moment a local model matches your needs.
Glossary
Terms worth knowing.
Open weights
A model whose trained parameters are published so anyone can download and run it on their own hardware, even if the training data and code are not shared.
Commoditizing your complement
A strategy where a company makes the product that goes alongside its own cheap or free so demand for its own product rises, like a hot dog seller wanting cheap buns.
Unified memory
A design where the CPU and GPU draw from one shared pool of memory, letting a Mac load large AI models that would not fit on a graphics card with its own small memory.
MLX
Apple's machine learning software framework for running models efficiently on Apple silicon.
Parameters
The learned numerical weights inside a model. A bigger count generally means a more capable model that needs more memory to run.
Tokens per second
How fast a model generates text. About 25 to 36 tokens per second is faster than most people read.
Frontier model
One of the most capable AI models available at a given time, usually closed and sold by a major lab by subscription or API.
The Script
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.
17px
I believe that one day you're going to pay for AI the same way you pay for your smartphone once. you buy the device and the two companies making this happen have spent years trying to destroy each other on october 15th meta gave away an ai model that you can run on your desk for free and it's a good model and the company that helped make it run that way is apple in the previous video of which this one is an important follow -up claude was swallowing up the applications ai is becoming the software but in this video i'm going to show you the strategic plan that these companies have to swallow up the companies that make those models.
And I'm also going to share with you the move that I believe that Apple is going to make when it's all said and done. Because right now, there are people who think that capitalism, it's on its way out. And I can see why they would think that.
Anthropic just put 5 % of their company into a public trust fund. And according to CNBC, that could be worth $50 billion or more, which is bigger than the entire economy of some countries. Which could be worth at least $50 billion.
Committing 5 % of its company shares for a new fund. The company announcing this type of public benefits. I remember when Sam Altman first pitched the idea that the AI companies would become service providers or utility providers, kind of like the water company or the power companies.
And now we're beginning to hear more people think that that's the way the AI is going to monetize in the future, because it's still very unclear how these companies are actually going to become profitable. and what that economy is going to look like. And to be perfectly fair, things could go that way.
But I think Meta and Apple and even some others have pointed in the opposite direction. And by the end of this video, I'm going to show you why I think they might be destroying the reason you pay for AI, why that might be the best thing to happen to your job and your business, and why I think that Apple is going to end up being the quiet winner of this entire AI race.
And so to understand why that partnership is such a big deal you need about two minutes of history meta and apple spent years as enemies in 2020 apple announced a pop -up for the iphone that would let you stop apps from tracking you so facebook took out full page newspaper ads attacking apple and the wall street journal reported that zuckerberg told his team they needed to inflict pain on Apple.
And when the pop -up shipped, most people tapped no. And in February 2022, Meta lost about $232 billion in market value in a single day, which at the time was the biggest one -day drop in the history of the US stock market. And the tracking pop -up was a big part of the reason.
He plunged over 26%. Meta platforms plunged. A meta plunged by around 20%.
And the second thing you need to know is how meta competes because it runs the same play every time. When Snapchat got popular, Instagram copied it with stories. And within eight months, Instagram stories had more people using it every day than all of Snapchat.
A blatant copy of Snapchat stories. Instagram is now going head to head with Snapchat. Instagram launches stories to compete directly with Snapchat.
When TikTok blew up. Meta built Reels. And when India banned TikTok in 2020, Instagram started rolling out Reels out there within days.
And I've talked to people who saw Reels before they ever saw TikTok. And to them, TikTok looked like the copy. When Be Real took off, Instagram added a dual camera and Be Real's daily users filled by more than half.
And when Twitter stumbled, Meta launched threads and got 100 million signups in five days. Yeah, 100 million. It's massive.
We're surpassing 100 million users in just five days. 100 million signups in just five days. And so the play is to take whatever is working somewhere else, make it free, put it in front of billions of people before the original can get there.
And there is a name for this in business strategy. It's called commoditizing your compliments. And the easiest picture is a hot dog stand.
Because if you sell hot dogs, you actually want buns to be as cheap as possible. Meta sells ads and attention. So to Meta, an AI model is the bun.
And Zuckerberg pretty much said this in 2024 when he wrote that selling access to AI models isn't Meta's business model. And the third thing is that Meta's AI didn't go smoothly at first. Its Lama 4 models disappointed people in 2025.
And so Meta spent $14 .3 billion for about half of a company called Scale AI. and brought in its founder, Alexander Wang, to run a new super intelligence lab. And that lab's first model, MuseSpark, came out this April as a closed model.
And so a lot of people assume Meta was done giving its AI away. And then October 15th happened. And this is where I want to slow down because I don't think people understand how big this is.
Meta released a model called MuseSpark Open, an 816 billion parameter model with open weights, which means anyone can download it and run it themselves. And Meta calls it the most capable American open weight model. model to date and meta work with apple the company zuckerberg once wanted to inflict pain on to make it run on mac studios connected together using apple's own machine learning software called mlx and apple published the results on its own research blog now partnering with meta on ai with john turnus now at the helm is this a new era that actually brings these two stalwarts closer together Zuckerberg publicly thanked John Ternes, Apple's new CEO, and Meta's chief AI officer, Alexander Wang, told Axios he was excited for mommy and daddy to be playing nice again.
Now, if you want to understand what Meta actually gave away, you want to make sure you're precise because there is some confusion about this. On the main independent intelligence index from a group called Artificial Analysis, Meta's best model right now is Muse Wonderful. And it's the number three model in the world.
Right behind Anthropix, two best models and ahead of GPT -6 and Gemini. And Meta kept that one closed. What it gave away sits a tier below that.
And that's the playbook working exactly the way it's supposed to. Because all you have to do is give away something good enough. that everyone else's price starts to look ridiculous and the free models are already good enough for a lot more than people think on today's version of the same index the paid flagships from about a year ago models like gpt5 and claw sun at 4 .5 score somewhere between 16 and 28.
a free model from alibaba called quinn 3 .8 27B scores 34, and it runs on a Mac with 24 gigabytes of memory. The best free model in the world right now, Mimo from Xiaomi, ties XAI's newest paid flagship, Grok 4 .7.
And a research group called Epic AI found that open models trail the best paid models by only about four months on average, so the gap between free and paid is down to a few months. But here's the bottleneck, and it's the part that makes Apple so important. Running the full Muse Spark open model takes about 493 gigabytes of memory, which means two of the biggest Mac studios Apple makes linked together.
And those 512 gigabyte machines aren't even shipping until late October and are expected to cost well over $10 ,000 each. On two of them, the model writes about 25 tokens a second and on four of them about 36, which is faster than you can read it. So the model is free.
but the machine to run the full thing isn't cheap yet. And that's going to be the bottleneck. And whoever solves that problem.
And then think about what Wang said next, which is that once developers are running the model on their own machines, Meta has no control or visibility to any of that. That's coming from the company whose entire fight with Apple was about wanting visibility into what you do on your phone. So the two companies that spent years at war just teamed up to put AI somewhere neither of them can watch it, which is your desk.
And that's the opposite of a future where a few companies run everything. And so you have to ask why Apple would do this. And to answer that, you have to look at how badly Apple's own AI has gone.
In 2024, Apple ran ads for a smarter Siri. In 2025, it said it was going to take longer than they thought. People who bought iPhones because of those ads sued.
And in May of this year, Apple agreed to a proposed $250 million settlement. iPhone 16 Pro is built from the ground up. Apple has agreed to a $250 million settlement.
Which claims its Siri feature didn't live up to... And in January, Apple announced that its next generation of models would be based on Google's Gemini. So on paper, it just looks like Apple is losing the AI race.
But Apple has almost never been first. And it's almost never needed to be. Google launched a phone wallet in 2011.
and Apple Pay came out in 2014. Samsung was selling giant phones in 2011 and Apple didn't ship a big iPhone until 2014. Pebble was selling smartwatches before the Apple Watch came out in 2015.
That's the pattern. Apple lets everyone else spend the money and work out the kinks and then it comes in later, does it better and takes the market. And Apple is too big and too profitable to burn tens of billions of dollars.
trying to win a model race that gets cheaper every few months. So I think it's going to do the same thing with AI, waiting until the intelligence is cheap and then owning the device you run it on. And if that sounds familiar, it's because Meta plays the same game.
Meta actually wasn't first to disappearing stories, short videos, or a Twitter alternative either. Neither company cares about being first. They both wait.
Watch what works and then do it better with the advantage they already have. And the advantage is where they're different. Meta's advantage is reach because it can put something in front of billions of people overnight.
Google has the same kind of reach with phones because about 69 % of the phones around the world run Android. So when Google ships a feature, it's going to go everywhere. Apple has about 31 % of the world, but in the United States, it's about 54%.
And the US is the biggest economy. on earth. So when Apple ships a feature, it lands in front of the most valuable audience on earth, and Apple's going to get more out of every launch than anybody else.
So Meta wins by reaching the most people, and Apple wins by reaching the most valuable ones. And the clearest signal is who Apple just put in charge. In April, Apple announced that John Ternus would replace Tim Cook as CEO, and he took over on September 1st.
15 years. Apple has a new CEO for the first time in 15. Massive leadership shakeup at Apple.
Ternus is a mechanical engineer who joined Apple's product design team in 2001. And Apple credits him with helping introduce the iPad and AirPods and overseeing generations of the iPhone, the Mac, and the Apple Watch. And Tim Cook said he has the mind of an engineer.
And so Apple just handed the company to a person whose whole career has been making the hardware. better. And look at what was already happening before he took over.
In April, Cook told investors that the Mac mini and the Mac studio had sold out because customers were adopting Macs to run AI models locally faster than Apple expected. And he said Apple had under called the demand. Then in August, Apple launched a new Mac mini starting at $899 and pitched it for always on desk side agentic computing.
And that's Apple's own language. about its own product. And if you want to know why it's the Mac in particular, it comes down to memory.
On most computers, the graphics chip that runs AI has its own small pool of memory. So it's like trying to run a restaurant out of a food truck kitchen. And even a top gaming card with 32 gigabytes can't hold a mid -sized model that needs 35 to 40.
But Apple's chips share one big pool of memory across everything. So the whole kitchen is available to the model.
So here's where I think this goes. Turnus' Apple is going to ship a Mac Mini that's affordable and has enough memory to run serious open models locally and the handwriting's already on the wall. And I'll come back to what that means for Apple at the end.
And Apple isn't the only one who sees this coming. On October 16th, Nvidia, the company whose chips train most of these models, started selling rts spark laptops from dale hp lenovo asus msi and microsoft starting at 2 599 dollars with up to 128 gigabytes of memory shared between the processor and the graphics chip which is the same trick that makes the mac so good at this nvidia says they can run 120 billion parameter models right on the laptop and its dgs station desktop is built for models up to a trillion parameters.
And Nvidia knows the hot dog rule better than anybody because Nvidia sells the chips. So to Nvidia, the model is the bun. It's planning to spend about $26 billion over five years building open models.
And the one it just shipped for these laptops is open all the way down to the training data. And so if anybody's going to beat Apple to the AI bots on your desk, it's Nvidia. Now this is where it connects back to my last video because a lot of people heard the models are becoming the software and assumed i meant everyone was going to sit down and build their own apps and most people are never going to do that so let me be clearer the models are becoming the software i think the ai builds the app for you right when you need it google already showed a version of this in his gemini app where the model creates a custom interactive screen for whatever you asked.
And Anthropic ran a preview where Claude wrote the software in real time while you clicked around in it. And OpenAI recently launched Intelligent UI. And so all you're doing is clicking buttons on an app that didn't exist a minute ago.
Jensen Huang described the same shift when he launched the RTS part. When he said, for 40 years, you launched apps. You clicked.
you typed and now you ask and the pc does the work and now put that on a mac mini running a free model and that machine doesn't need a subscription it doesn't need an internet connection to think and once you've bought it the main thing you pay for is the electricity the utility you already pay for, not intelligence as a utility.
So if you want to know who actually wants that, think about the coach with the client notes, the accountant with tax returns, or the consultant holding the client's financials, because none of them are supposed to be pasting that into somebody else's server. And with the model on your desk, nothing ever leaves the room. And that's the part I'd add to my last video.
The model becomes the software, and then the software moves on to hardware you own and i'm not even gonna go into the privacy that you have with your model and the conversations that you're able to have because that's an obvious benefit that people are going to love for personal reasons and also commercial reasons and anthropic isn't the only ai company trying to hand you a piece of itself elon musk just penned a post saying that next year he's going to give one of his spacex shares to every kid in america and xai is part of spacex now so that's one more ai company trying to put a piece of itself in your hands his spacex company to every single child in america give a spacex chair to quote every kid in one share of spacex And back in July, the Financial Times reported that OpenAI offered the Trump administration a 5 % stake in the company, which would be worth about $42 billion.
And the pitch was that it gives the public a financial interest in OpenAI.
And so three of the biggest closed AI companies all want the American people invested in them. and you can call that a gift or you can just call it a bribe but either way it changes whose side you're on because if your kid owns a piece of the lab and your government owns a piece of the lab and your retirement account owns a piece of the lab then you're rooting for the lab to win and the lab winning is the future where a few companies are the electricity providers and everyone else waits for a check you need the labs to win in that scenario and here's the question almost nobody's asking, which is what happens when the free models catch up?
Because if Anthropix models just keep getting better and swallow everything, that's one future. And it's possible, but Meta's models are catching up right now. they're number two.
And every few months, the gap gets smaller. So what happens when the free model on your desk is going to be good enough for most of what you do? I think that future looks a lot more like the capitalism we already have, where you earn what you get and keep what you build and competition is what protects it.
And you can't have real competition if everyone has to rent their intelligence from the same few companies. So when Meta gives a model away and apple sells you a machine that can run it the intelligence gets spread out across millions of small businesses and the money keeps moving around the economy the way it always has instead of getting sucked up by a couple of new giants and that's why i don't think universal basic income is where this goes if meta and apple have anything to say about it because when the tools are cheap and you own them people keep their jobs and just use the AI to do more with them.
So if you're against AI because you're worried about losing your living, this should actually be good news to you. So yes, I think Meta and Apple have a sinister plan. And it's only sinister if you're one of the companies that wanted to own the whole thing.
And if you want to see how seriously the closed labs take that question, look at how Anthropic talks about open models. In July, Jensen Huang made his first post ever on X. And it was a letter NVIDIA signed called Open Weights and American AI Leadership.
And Elon Musk posted, Jensen is right. Meta and Microsoft signed it. Open AI and Google added their names that week.
in and Anthropic was the only major AI lab that didn't. The investor David Sachs posted that Anthropic won't stop until they kneecap open source. Anthropic CEO Dario Amadei answered with a post saying Anthropic has never advocated for a ban on open weight models and that open models without dangerous capabilities are a public good.
But just last week, Anthropic's head of international told Bloomberg that by definition, there's more risk with open models. And two days before that, Anthropic published research on small open models spreading themselves across cheap Android TV boxes.
And Jensen Wong, whose company is betting billions on open models, told Fortune last week that anyone who believes AI will end civilization really ought to stop working on it. And Fortune read that as a shot at Anthropic. Now, to be fair, Some of that safety research is real and it's worth taking seriously.
But think about what a model on your desk is actually going to do to a company that rents you intelligence by the month. I'll let you make up your own mind. But when a company keeps warning you about the free version of what it sells, it's worth asking.
Who that warning protects. And then this week, the information reported that Sam Altman walked OpenAI employees through three end games for AI. The first is an international watchdog modeled on the agency that inspects nuclear programs.
And that's reportedly the one he leaned toward. The second is the big labs merging into one. The way Europe built CERN.
An idea that goes back to Google DeepMind's Demis Hassabis. And the third. is one company getting so powerful it can melt everyone else's GPUs and he called that one terrible.
But look at what all three have in common and every one of them the most powerful AI still lives inside a handful of labs. And the only question is who watches them, who merges or who wins.
The future where the model runs on the computer on your desk isn't one of his three. And that's the one Meta and Apple are building. And I think the open models are going to drag down the value of the closed AI labs and the numbers they're carrying right now make them vulnerable.
Anthropic raised money in May at a valuation of almost a trillion. And there is talk of it going public at close to two. OpenAI was valued at $852 billion in March.
It lost about $21 billion running the business last year, and its own forecast for this year's revenue was cut by about $20 billion. And XAI lost about twice as much money last year as it brought in, and its newest flagship just tied a model you can download for free. Now, to be fair, Anthropic's revenue went up 45 % in a single quarter.
and big companies still mostly choose paid models. But if you wanna see what an open model can do to evaluation, look at January, 2025, when a free model from DeepSeek helped knock about $589 billion off NVIDIA in a single day, which is more than twice what Meta lost on its worst day. Or shed nearly $600 billion from its market cap.
The largest single day loss for stock ever. And I think the professionals who need them very best are going to keep paying for frontier models but nowhere near the number of people paying today because for most of what most people do the free model on their desk is going to be enough google can survive that because search and ads pay the bills and i think that's why it's been able to play it slow but companies whose whole business is renting you the model are going to fill it first so what's apple going to do at the end of all this i think it's one of two things and i put money money on one of them.
The first is that Apple lets Meta run the value of these labs down and then buys one for pennies on the dollar compared to what it's worth today. And Apple has already tried to buy its way into the frontier once because in January, Bloomberg's Mark Gurman said Apple had planned to rebuild Siri around Claude and the deal fell apart because Anthropic wanted several billion dollars a year with the price doubling every year for three years.
And Gurman said Anthropic was holding them over a barrel. Apple has about $146 billion in cash and investments.
It spent $62 billion buying back its own stock in just nine months. And Tim Cook said last year that Apple is very open to acquisitions that speed up its roadmap. Right now, Anthropic is worth more than six times Apple's cash.
But if their free models keep pulling paying customers away, that number comes down and apple has a habit of partnering with companies right up until it doesn't need them the way it replaced intel chips in the mac in 2020 and started replacing qualcomm modems in the iphone in 2025. we're announcing that the mac is transitioning ending an almost 15 -year partnership with transitioning from Intel CPUs and its Mac.
And Bank of America just told investors it expects potentially bigger acquisitions under the new CEO. The second path is even simpler because Apple keeps its cash and lets Meta keep doing the heavy lifting and giving away models that are close to the state of the art. And Apple just keeps building hardware that runs them locally for free.
And then power users start running Meta's models with no limit. and no subscription just you and your mac mini and either way the company that spent the last two years looking like it lost the ai race ends up owning the thing that matters most which is the device the intelligence runs on and you pay for ai the same way you pay for the camera in your iphone once when you buy the device so let's put it all together now some people think that capitalism is dead And if the closed labs keep getting better and swallow up the economy, things might go that way.
But in the same week, Meta and Apple, two companies who have traditionally gone at each other's neck, have teamed up to release a free model that can run on your desk. Because Meta's oldest play is to give away what everyone else is selling. And Apple's oldest play is to let everyone else go first and then own the device.
And NVIDIA is building the same kind of bots for everybody else. In my last video, I explained how Claude and other AIs were swallowing up software and applications.
But now Apple, Meta, and Nvidia are all lining up to swallow the reason you pay for Claude, ChatGPT, or any other frontier closed model. So the reason you pay for AI is going to keep getting weaker. And I think the future where you own your own AI and you're able to run it on your desk is a lot closer than a future where the frontier models swallow up the entire economy and everyone gets checked for staying at home.
If you have a pro subscription for either of these labs, then you're paying somewhere around $2 ,500 a year. And if you have a pro plan with both of them or multiple accounts, you're spending five grand a year. And if you have the $500 plan with ChatGPT, you're spending $6 ,000 a year.
The only thing that's left is to have an open. open source model that's as good as Opus, Fable or Astra to run locally. And you can get a Mac mini right now with 128 gigabytes of unified memory, all the storage you need and just decked out with everything.
by simply taking the price you're paying to the closed AI labs and giving it to Apple as a monthly payment on that equipment. So the hardware is already present. The only thing keeping people from switching right now is confidence that their local model is going to be as good as the closed models.
And once we have local models that are, I think it's going to be a no brainer for people to just purchase a Mac mini or Mac studio, get all the bells and whistles and make payments on it if they have to, instead of paying. five or $6 ,000 every year in perpetuity for a model you never own. If you got value out of the video, make sure you hit the like button, subscribe to the channel and as always take care, have a great day.
And if you haven't watched that other video, please go and watch the other video so you can see the two different futures that we're heading towards with AI. I'll see you in the next video.
The Hook
The bait, then the rug-pull.
Meta gave away an open model on October 15, and the company that helped make it run on a desk is Apple, the same one Meta spent years fighting. The video argues the pair, with Nvidia, are quietly weakening the reason you pay for closed AI.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Corey McClain downloads a free open-source clone of After Effects, then hands it to a Claude Code agent with zero training on the app and watches it build a real edit.
OpenAI rebuilt voice mode around GPT-5.6 and GPT-6 Astra, gave paying users up to unlimited daily hours, and made the case that talking may replace typing as the default way to use AI.
A creator runs OpenAI's new writing-style feature and his own custom voice skill against the same AI-written passage, then diffs the results line by line.
00:00
00:22
00:44
01:04
01:28
01:49
02:09
02:30
02:49
03:11
03:33
03:53
04:14
04:37
04:58
05:19
05:38
06:01
06:25
06:42
07:04
07:23
07:45
08:20
08:28
08:56
09:10
09:35
09:52
10:13
10:38
10:58
11:19
11:38
12:00
12:25
12:41
13:04
13:23
13:50
14:07
14:28
14:50
15:11
15:30
15:54
16:14
16:36
16:55
17:20
17:38
18:00
18:23
18:43
19:07
19:24
19:45
20:06
20:27
20:49
21:12
21:33
21:53
22:15
22:36
22:55
23:19
23:38
23:59
24:21
24:44
25:00
25:22
25:50
26:09
26:25
26:49
27:09
27:41
28:03
10:45
17:44
17:58
08:03
10:08
29:31