Latest / Connecting the Dots / Apple's Price Pressures, AI Export Restrictions, and Unfixable Guardrails
Transcript
- 0:00Today's tech news moves fast, but understanding
- 0:03it shouldn't be hard. That's why Snarfl Solutions
- 0:06Group brings you Connecting the Dots, where we
- 0:09break down what's happening in tech and why it
- 0:11matters to you. Now, let's dive in with your
- 0:14hosts. I told you that being absolutely terrible
- 0:18at coding is actually your biggest advantage
- 0:20in the new economy. I mean, it sounds crazy,
- 0:23but it's true. Right. and that the very same
- 0:27software that is about to make your next smartphone
- 0:29hundreds of dollars more expensive is, you know,
- 0:33actively causing a massive diplomatic crisis
- 0:35in Europe. Yeah, it's all connected. We're looking
- 0:37closer at the hidden forces reshaping your wallet,
- 0:40your workplace, and, well, the global balance
- 0:43of power. Welcome to the show. I'm Alex, and
- 0:46sitting across from me is Morgan. We are your
- 0:49professional, hopefully somewhat charismatic,
- 0:52co -hosts of Connecting the Dots, broadcasting
- 0:54live from Snarfful Solutions Group right here
- 0:57in Sacramento, California. It is fantastic to
- 1:00be here with you. Today is Thursday, June 18,
- 1:032026. Before we get into the heavy hitters of
- 1:06the tech world, let's do a quick vibe check on
- 1:08the day's data. Always a good idea. Right. So
- 1:11if you are stepping out here in Sacramento, you're
- 1:13in luck. We've got a beautiful sunny day hitting
- 1:1586 degrees Fahrenheit with a really nice light
- 1:18breeze somewhere between like five and 14 miles
- 1:20per hour. Oh, that is just pristine patio weather.
- 1:23Exactly. But if you are traveling or listening
- 1:27from. Elsewhere in the country you might want
- 1:29to double check your itinerary nationally. We're
- 1:31looking at 400 active weather alerts Wow 400
- 1:34yeah and 283 of those are overland So definitely
- 1:38head over to weather gov to see if your area
- 1:40is affected before you you know hit the road
- 1:42Yeah, the weather systems have been incredibly
- 1:44volatile this week always good to be prepared
- 1:46for sure moving over to the markets It was well
- 1:49bit of a red day on Wall Street yesterday The
- 1:51Dow closed down 445 points, which is about a
- 1:550 .86 % drop. The S &P 500 also dipped, losing
- 2:01104 points, or 1 .39%. And if you are watching
- 2:05the crypto boards, Bitcoin took a slight dip
- 2:07as well, closing at $64 ,318. It really looks
- 2:11like a broader cooling off period across, you
- 2:14know, both traditional equities and digital assets.
- 2:16It definitely does. And finally, just a mixed
- 2:19bag for local sports from the 17th. The SF Giants
- 2:22gave us something to cheer about with a solid
- 2:247 -2 win against the Braves. Robbie Ray took
- 2:26the win on the mound for that one. But on the
- 2:28other side of the bay, the Oakland A's dropped
- 2:30their game to the Pirates 4 -12. And for the
- 2:32basketball fans out there, no Kings games report
- 2:35right now. Well, we will absolutely take the
- 2:38Giants win where we can get it. Oh, no complaints
- 2:40here. All right, let's get into today's Morningstar
- 2:43Tech Report for this deep dive. We have a massive
- 2:46lineup. We're looking at Apple's impending and
- 2:49apparently unavoidable price hikes. Right. Then
- 2:52we are shifting into this intense geopolitical
- 2:54standoff over the world's most advanced artificial
- 2:57intelligence. Let's unpack this. Today's episode
- 3:01is brought to you by Strong Coffee, the premium
- 3:03instant coffee that tastes like a traditional
- 3:05brew but is packed with 15 grams of protein,
- 3:08MCTs, and adaptogens like ashwagandha. Get clean,
- 3:11sustained energy without the jitters. Use promo
- 3:14code SNARFL for 20 % off at strongcoffeecompany
- 3:17.com. So the AI boom is officially coming for
- 3:20your wallet. Yeah, it really is. Tim Cook recently
- 3:22sat down for an interview with the Wall Street
- 3:24Journal and he delivers some news that... frankly
- 3:27no consumer ever wants to hear. He stated that
- 3:29massive price hikes on upcoming Apple products
- 3:32are simply unavoidable and that the current supply
- 3:35chain reality they're facing is just unsustainable.
- 3:39And you know, the culprit here is directly tied
- 3:42to the physical infrastructure required to run
- 3:44artificial intelligence. Which is wild to think
- 3:46about. Right, because when we visualize AI, we
- 3:49usually picture like lines of code, algorithms,
- 3:51chat interfaces. Is that for stuff? Exactly.
- 3:55But AI is incredibly reliant on physical hardware.
- 3:58Specifically, it is devouring the global supply
- 4:01of memory and storage chips. Wait, hold on. When
- 4:04I think of AI, I think of massive computing power,
- 4:07right? Companies buying up huge NVIDIA GPUs that
- 4:10do... billions of calculations a second. Sure.
- 4:13So why is Apple getting outbid on memory and
- 4:15storage? Isn't memory just the digital filing
- 4:18cabinet where things are saved? Why is that the
- 4:20bottleneck? That is a very common misconception.
- 4:23AI isn't just compute heavy. It is fundamentally
- 4:25memory bound. Okay, how so? Well, think about
- 4:28how a large language model works. It has billions,
- 4:32sometimes trillions of parameters or weights.
- 4:35Every single time the AI generates a single word,
- 4:38it has to load those billions of parameters from
- 4:40the memory into the processor. Oh, wow! Do the
- 4:43math and then move on to the next word. It's
- 4:45a massive amount of data moving constantly. Ah,
- 4:49I see. So the processor is like the factory,
- 4:51but the memory is the conveyor belt feeding materials
- 4:54into the factory. Exactly. It doesn't matter
- 4:56how fast your factory can build things if the
- 4:59conveyor belt is too slow or too small. The whole
- 5:02operation just stars. That is the perfect way
- 5:04to look at it. And to prevent that starvation,
- 5:07AI data centers require something called High
- 5:09Bandwidth Memory, or HBM. Okay. It is a highly
- 5:12specialized, incredibly dense type of memory
- 5:14chip. Because the AI companies are willing to
- 5:17pay massive, I mean, almost unbelievable premiums
- 5:20for HBM, the chip manufacturers, companies like
- 5:23Samsung and SK Hynex, are actively shifting their
- 5:26production lines. Makes sense, follow the money.
- 5:28Right. They are literally tearing down the manufacturing
- 5:30lines that used to make consumer device memory.
- 5:33You're kidding. And re -dinging them to make
- 5:35server -grade AI memory. Which obviously creates
- 5:38a massive supply. vacuum for consumer electronics.
- 5:41Yeah. It really reminds me of a normal everyday
- 5:43family trying to buy a starter house in a crazy
- 5:46real estate market. Oh, absolutely. They save
- 5:48up, they put in a competitive offer and then
- 5:50suddenly a massive corporate buyer swoops in
- 5:54with an all cash over asking offer and just takes
- 5:58the property. Yeah. The AI corporations are the
- 6:01all cash buyers here. And even Apple, one of
- 6:04the wealthiest, most powerful supply chain operators
- 6:07on earth, is getting entirely outbid. And Tim
- 6:10Cook actually likened this dynamic to a 100 -year
- 6:12flood in the supply chain. He stated he hasn't
- 6:15seen anything disrupt the market this violently
- 6:18in over 40 years. So what does that mean for
- 6:20you, the consumer? Well, the real -world impact
- 6:22is pretty steep. Market analysts are projecting
- 6:25that the upcoming iPhone 18 Pro could cost around
- 6:28$270 more. Just for the phone. Yeah, just so
- 6:30Apple can maintain its current profit margins.
- 6:33And they're already doing this quietly across
- 6:35their other products, right? Yeah. The base price
- 6:37of the new Mac Mini just jumped from $599 for
- 6:41$799. They didn't even raise the price of the
- 6:44entry level components. They just eliminated
- 6:47the lowest tier option entirely because they
- 6:50can't source those chips cheaply enough anymore.
- 6:52And you know, this hardware crisis has a really
- 6:54fascinating geopolitical twist to it, too. Oh,
- 6:57yeah. President Trump recently announced on Truth
- 6:59Social that Intel has struck a massive deal to
- 7:03manufacture Apple's chips right here in the United
- 7:06States. Which is a huge pivot because Apple has
- 7:08historically relied almost entirely on Taiwan.
- 7:11for its processor manufacturing. Exactly. Bringing
- 7:14that production to U .S. soil is a frantic effort
- 7:16to diversify their manufacturing base and secure
- 7:20a domestic supply chain that isn't vulnerable
- 7:23to international shipping bottlenecks or foreign
- 7:25conflicts. It really is. And the administration
- 7:28had actually laid the groundwork for this earlier.
- 7:30Back in August, the U .S. government made an
- 7:32$8 .9 billion investment in Intel, taking roughly
- 7:35a 10 percent stake in the company. In his recent
- 7:37post, the president highlighted how massively
- 7:40that investment has grown. Intel's overall market
- 7:43value has surged from around $100 billion at
- 7:45the time of the investment to over $600 billion
- 7:48today. It's crazy. It really reframes how we
- 7:52look at physical hardware. It isn't just consumer
- 7:55electronics anymore. Semiconductor manufacturing
- 7:58is straight up being treated as a critical core
- 8:00component of national security. And that reality
- 8:03creates the perfect bridge to our next story.
- 8:05Yeah. Because if Washington is treating the physical
- 8:08hardware like a national security asset, They
- 8:10are treating the actual AI software like a weapon
- 8:13of mass destruction that needs to be locked and
- 8:16evolved. Exactly. So let's look at the gap between
- 8:18that geopolitical panic and the international
- 8:21business landscape. Because if hardware is getting
- 8:23expensive, the software is getting heavily guarded.
- 8:26Right. We are moving from the physical chip shortage
- 8:28to an intense geopolitical tug of war over the
- 8:32AI software itself, specifically. A massive standoff
- 8:35between the leading AI company, Anthropic, and
- 8:38the White House. This is an incredibly complex
- 8:41situation. So the Trump administration recently
- 8:44utilized an export control directive that effectively
- 8:47forced Anthropic to pull its most advanced AI
- 8:50models, known as Mythos and Fable 5, completely
- 8:53offline for international users. And just to
- 8:56clarify for everyone listening, since Fable 5
- 8:58is essentially the public, heavily safeguarded
- 9:01face of their absolute cutting edge frontier
- 9:03model, Mythos, taking it offline is a massive
- 9:07blow to their user base. Oh, definitely. So what
- 9:09actually triggered the administration to drop
- 9:11the hammer like this? Well, the primary catalyst
- 9:13appears to be an anthropic initiative called
- 9:16Project Glasswing. Anthropic was using this program
- 9:19to grant early experimental access to their mythos
- 9:22model to a select group of trusted global organizations.
- 9:25And one of those organizations was SK Telecom,
- 9:28which is South Korea's largest wireless carrier.
- 9:31Wait, South Korea is a major US ally. Why would
- 9:34sharing a model with their largest telecom carrier
- 9:37trigger an export ban? So the White House raised
- 9:39major concerns regarding the corporate structure
- 9:41behind that telecom. The administration alleges
- 9:44that SK Telecom's parent group has extensive,
- 9:47deeply embedded business ties to China, spanning
- 9:50across their energy, semiconductor, and industrial
- 9:53sectors. And compounding that anxiety, Independent
- 9:57researchers from Amazon recently published a
- 9:59paper flagging certain vulnerabilities they found
- 10:02in the public Fable 5 model, demonstrating that
- 10:05it was technically possible to bypass some of
- 10:07the model's safety guardrails. So the administration
- 10:09looks at the board. They see an advanced frontier
- 10:12model going to a foreign telecom with alleged
- 10:15ties to Chinese industry. And simultaneously,
- 10:18Amazon researchers are saying, hey, the public
- 10:21version of this software has security vulnerabilities.
- 10:23Right. The government's response was basically
- 10:25a blanket order. revoke access for all foreign
- 10:28nationals immediately, including foreign nationals
- 10:30working legally inside the United States. Which
- 10:33is a complete logistical nightmare for a tech
- 10:35company. Anthropic couldn't execute that kind
- 10:38of granular nationality -based access control
- 10:41without violating international privacy laws.
- 10:44Right, of course. And completely dismantling
- 10:46their own operational infrastructure. So, facing
- 10:50an impossible compliance demand, they simply
- 10:52pulled the plug on the models entirely. I have
- 10:54to ask, though. Does a blanket ban like this
- 10:57really make us safer? Or is it like a parent
- 11:01grounding the entire neighborhood from using
- 11:03a trampoline just because one kid might eventually
- 11:06invite someone over that the parent doesn't trust?
- 11:08That's a good analogy. I mean, it feels like
- 11:10a really isolationist approach to a globally
- 11:12connected technology. And that is the exact debate
- 11:15happening in diplomatic circles right now because
- 11:17this domestic US dispute is actively freezing
- 11:20Europe out of the AI race entirely. Wait, I'm
- 11:24struggling to see how Europe factors into a bilateral
- 11:26issue between a U .S. tech company and a South
- 11:29Korean telecom. How does the EU get caught in
- 11:32the crossfire? It comes down to regulatory access.
- 11:35The European Union's primary cybersecurity agency,
- 11:38known as ANISA, was actually scheduled to meet
- 11:41with Anthropic this week. Yeah, they had spent
- 11:44months negotiating to get early access to the
- 11:46Mythos model through that exact same Project
- 11:49Glasswing initiative. The EU desperately wants
- 11:52to understand how these frontier systems work
- 11:55so they can build up European cybersecurity defenses
- 11:57against them. But now, because of the US export
- 12:01block triggered by the South Korean issue, that
- 12:04European meeting is completely dead in the water.
- 12:06Precisely. It is a harsh wake -up call for Brussels.
- 12:09It proves to European regulators that their technological
- 12:12sovereignty is entirely dependent on Washington's
- 12:14moods. The EU does not possess a domestic equivalent
- 12:18to Anthropic's mythos model to test their defenses
- 12:21against. So an American export policy designed
- 12:24to counter Chinese industrial ties in South Korea
- 12:26is now directly dictating a European government
- 12:29agency's ability to protect its own digital infrastructure.
- 12:33That is a massive realization. And Anthropic
- 12:35is just caught right in the middle trying to
- 12:37be a trusted global partner. to European regulators,
- 12:40while simultaneously complying with strict, suddenly
- 12:43shifting directives from their own government.
- 12:45Which brings us to a critical breaking point.
- 12:48We really need to dig deeper into why the relationship
- 12:51between Anthropic and the administration is so
- 12:53heavily fractured right now, because the current
- 12:56deadlock revolves around a specific technical
- 12:58demand from the administration that reveals a
- 13:00fundamental misunderstanding of how artificial
- 13:03intelligence actually works. Yeah, so let's unpack
- 13:05this. The Trump administration officials are
- 13:08telling Anthropic that before they are allowed
- 13:10to re -release the Fable 5 model, they must unequivocally
- 13:14prove that the model's safety guard rails are
- 13:17completely impossible to jailbreak. Yes. Meaning,
- 13:21no matter what adversarial prompt a user inputs,
- 13:24the model will never bypass its own safety training
- 13:27to output malicious code or dangerous instructions.
- 13:30And independent cybersecurity experts across
- 13:32the industry are looking at this demand and stating
- 13:34unequivocally that it is literally impossible.
- 13:38Guardrails on large language models are fundamentally
- 13:40just stop gaps. Let's talk about the mechanism
- 13:42here. Why is it impossible to permanently secure
- 13:46an AI? Why can't you just write a rule that says,
- 13:48never explain how to build a bomb and hard code
- 13:51it into the system? Well, LLNs don't understand
- 13:54words the way a traditional software program
- 13:56understands a line of code. They map concepts
- 13:59in high dimensional space. Okay, what does that
- 14:01mean? Think of it like a massive galaxy of meaning
- 14:04where every concept has a mathematical coordinate.
- 14:07So if the government says, block all instructions
- 14:09for making a bomb, the AI puts a wall in front
- 14:13of that direct coordinate path. Right. but human
- 14:16language is endlessly creative. Exactly. A user
- 14:19can just approach that exact same concept through
- 14:21a semantic backdoor. They don't ask for a bomb
- 14:24recipe. They ask the AI to write a fictional
- 14:27story about a rapid exothermic chemical reaction
- 14:30occurring inside a confined metal space for a
- 14:34sci -fi novel. Oh wow. The AI traverses the high
- 14:36dimensional space, bypasses the blocked bomb
- 14:39coordinate, and arrives at the exact same dangerous
- 14:42information from a completely different angle.
- 14:44cannot patch human language. It is literally
- 14:47like demanding that a shipbuilder create a completely
- 14:50unsinkable ship before they are allowed to put
- 14:52it in the water. You can build watertight compartments,
- 14:54you can have lifeboats, you can train the crew
- 14:57perfectly, but you cannot guarantee it will never
- 15:00sink when faced with the infinite variables of
- 15:02the ocean. Demanding absolute Unhackable security
- 15:06is basically demanding that the technology never
- 15:09be released. And this impossible standard has
- 15:12led to deep frustration within Anthropic. According
- 15:15to reports, some Anthropic employees believe
- 15:17the administration is intentionally placing these
- 15:20impossible hurdles in front of them as a form
- 15:23of targeted punishment. Which stems from a backstory
- 15:25that unfolded earlier this year. Anthropic was
- 15:29in negotiations for a Department of Defense contract
- 15:31managed by Pete Hegseth. During those talks,
- 15:34Anthropic pushed back. They requested that specific
- 15:37use case regulations be put in place, ensuring
- 15:40their AI wouldn't be utilized for mass domestic
- 15:42surveillance or autonomous lethal targeting.
- 15:45And following those requests for ethical guardrails,
- 15:47Hegseth publicly rejected the stipulations and
- 15:50labeled Anthropic a supply chain risk. So we
- 15:52have a documented history of friction there.
- 15:55The company wanted specific restrictions on military
- 15:57implementation. The government pushed back on
- 16:00those demands and slapped a risk label on the
- 16:02company. And now we have this geopolitical standoff
- 16:05over a technical demand that experts say can't
- 16:08ever be perfectly solved. It is incredibly messy,
- 16:11but you know. Amidst all this high -level geopolitical
- 16:14panic and impossible regulation, there is actually
- 16:17an incredible silver lining for working professionals
- 16:20right now. Yes. While Washington is treating
- 16:23AI like a locked vault, everyday professionals
- 16:26are currently treating it like a junior intern.
- 16:28This is the part of the deep dive that I find
- 16:30absolutely fascinating because Anthropic just
- 16:33released a massive study analyzing 400 ,000 user
- 16:37sessions with their Claude code agent. Yeah.
- 16:39and the data completely shatters how we think
- 16:41about the value of AI in the workplace. They
- 16:44looked at how complex programming work is split
- 16:46between the human user and the AI agent, and
- 16:49the findings were totally counterintuitive. What
- 16:51did they find? They discovered that lawyers,
- 16:53middle managers, and research scientists, people
- 16:55with absolutely zero software coding background,
- 16:59were performing almost as well as professional,
- 17:02highly trained software engineers when they used
- 17:04these AI coding agents. The non -technical users
- 17:07were finishing complex tasks within just seven
- 17:10points of the professional engineers. Yeah. What
- 17:13the data proves is that a user's domain expertise,
- 17:16their deep, nuanced knowledge of corporate law
- 17:19or molecular biology or supply chain management,
- 17:23matters far more than their actual coding expertise
- 17:26when leveraging AI. I really want to emphasize
- 17:29this to you listening right now. If you have
- 17:31been worried about being left behind because
- 17:33you don't know Python or C++ EF, your specific
- 17:36industry knowledge is actually your greatest
- 17:39superpower in this new economy. It really is.
- 17:41It's like having a universal translator, but
- 17:43for pure logic. A great lawyer already knows
- 17:45the exact, rigorous logical steps required to
- 17:48win a case or structure a contract. They just
- 17:50don't know the programming syntax to build a
- 17:52software tool to automate it. The AI agent bridges
- 17:55that gap. It takes your pure domain logic and
- 17:59translates it into executable code. It perfectly
- 18:01echoes a recent joint study from Harvard on perplexity.
- 18:04AI agents aren't just making people type faster
- 18:07or draft emails quicker. They are allowing professionals
- 18:10to tackle massive cross -disciplinary problems
- 18:13that previously required an entire department
- 18:16of specialists. That's incredible. The value
- 18:18of the AI isn't capped by the model's processing
- 18:20parameters. It is capped by how well the human
- 18:23user understands the fundamental problem they
- 18:25are trying to solve. So the human is the architect,
- 18:29bringing the vision and the structural logic.
- 18:31The AI is just the construction crew swinging
- 18:34the hammers. It is an incredibly empowering shift
- 18:37for non -technical professionals. Before we wrap
- 18:39up, a quick thanks to Old Glory, an iconic music
- 18:42and sports fan merch store. Perfect for the techie
- 18:45in your life. Use promo code SNARFL for 15 %
- 18:48off at oldglory .com. So bringing all these threads
- 18:51together, the AI boom is clearly not just some
- 18:54abstract concept happening in a server farm out
- 18:56in the desert somewhere. It is incredibly tangible
- 18:59and reshaping the physical world. It is directly
- 19:02raising the price of the next smartphone you're
- 19:04going to buy because of the sheer physical limitations
- 19:06of hardware memory. And it is simultaneously
- 19:09creating massive diplomatic friction, putting
- 19:12the United States at odds with European regulators
- 19:15and international telecom giants over who gets
- 19:18to control the most powerful digital frameworks
- 19:21in the world. And at the exact same time is sitting
- 19:24quietly on your laptop right now, empowering
- 19:27you to execute the work of a seasoned software
- 19:30engineer. even if you've never written a single
- 19:33line of code in your entire life. Yeah. It is
- 19:36a technology defined by its massive contradictions.
- 19:39It is deemed too expensive to manufacture cheaply,
- 19:42yet incredibly accessible to the end user. It
- 19:45is considered too dangerous and unpredictable
- 19:47to export, yet it is currently revolutionizing
- 19:50mundane daily office work across the globe. Which
- 19:54leaves us with a really fascinating thought to
- 19:55ponder as we close out today. If the government
- 19:58is demanding absolute perfect security from AI
- 20:01before we are allowed to use it globally, but
- 20:04regular people are already using it to completely
- 20:06revolutionize their daily jobs right now, who
- 20:08really gets to decide when the future is safe
- 20:10enough to release? We're here to help at SNARFL,
- 20:13so reach out on our website if you have any more
- 20:14questions. That's a wrap on today's top stories.
- 20:21Remember, tech doesn't have to be complicated,
- 20:23and we're here to help connect the dots. If you've
- 20:26got questions or want to learn more, visit us
- 20:29at snarful .com. Thanks for listening, and we'll
- 20:32catch you next time.