WEBVTT 00:00:00.669 --> 00:00:03.509 Today's tech news moves fast, but understanding 00:00:03.509 --> 00:00:06.730 it shouldn't be hard. That's why Snarfle Solutions 00:00:06.730 --> 00:00:09.310 Group brings you Connecting the Dots, where we 00:00:09.310 --> 00:00:11.810 break down what's happening in tech and why it 00:00:11.810 --> 00:00:14.890 matters to you. Now, let's dive in with your 00:00:14.890 --> 00:00:18.809 hosts. The multi -trillion dollar cloud computing 00:00:18.809 --> 00:00:23.089 industry is, I don't know, about to be completely 00:00:23.089 --> 00:00:25.769 dismantled by this this little silver box sitting 00:00:25.769 --> 00:00:28.339 right there on your kitchen counter. Welcome 00:00:28.339 --> 00:00:30.579 to Connecting the Dots. I'm Alex. And I'm Morgan. 00:00:30.879 --> 00:00:33.000 And we are joining you from Snarfle Solutions 00:00:33.000 --> 00:00:35.299 Group in Sacramento, California. Yes, we are. 00:00:35.560 --> 00:00:40.039 It is Tuesday, August 25th, 2026. And you know, 00:00:40.119 --> 00:00:42.159 before we really look closer at this massive 00:00:42.159 --> 00:00:44.560 shift in the tech world today, let's just get 00:00:44.560 --> 00:00:46.520 right into your morning briefing. Sounds good. 00:00:46.700 --> 00:00:49.320 So if you are stepping out here in Sacramento 00:00:49.320 --> 00:00:52.579 today Expected to be mostly sunny. We're looking 00:00:52.579 --> 00:00:55.920 at a high of about 94 degrees Fahrenheit So yeah, 00:00:56.000 --> 00:00:57.640 it's gonna be a hot one definitely stay hydrated 00:00:57.640 --> 00:01:00.140 out there for sure. It's accompanied by a fairly 00:01:00.140 --> 00:01:02.600 light breeze though Yeah, wind is sitting at 00:01:02.600 --> 00:01:05.560 about I'd say three to eight miles per hour But 00:01:05.560 --> 00:01:08.040 if we kind of pull back and look at the national 00:01:08.040 --> 00:01:10.920 picture things are significantly more turbulent 00:01:10.920 --> 00:01:13.250 right now. Yeah, they really are Currently we 00:01:13.250 --> 00:01:15.469 are seeing a hundred and eighty one active US 00:01:15.469 --> 00:01:17.590 weather alerts and ninety seven of those are 00:01:17.590 --> 00:01:20.150 actually overland. Oh, well ninety six overland 00:01:20.150 --> 00:01:23.590 So wherever you are listening from this morning, 00:01:23.590 --> 00:01:26.269 please make sure to check weather gov for the 00:01:26.269 --> 00:01:28.829 specific details in your local area You don't 00:01:28.829 --> 00:01:30.650 want to get caught in any of that. No, definitely 00:01:30.650 --> 00:01:33.629 not Well moving over to the financial markets. 00:01:33.629 --> 00:01:36.109 It was actually a fairly quiet close yesterday 00:01:36.109 --> 00:01:39.450 The Dow Jones dips slightly down just about a 00:01:39.450 --> 00:01:42.010 quarter of a percent close at fifty three thousand 00:01:42.010 --> 00:01:44.209 four hundred sixty six point zero five okay so 00:01:44.209 --> 00:01:47.390 barely moved right and the s &p 500 largely mirrored 00:01:47.390 --> 00:01:50.209 that movement just inching down zero point zero 00:01:50.209 --> 00:01:52.590 seven percent to settle at seven thousand six 00:01:52.590 --> 00:01:55.049 hundred and seventy one point six five so pretty 00:01:55.049 --> 00:01:57.049 stable across the board for the traditional markets 00:01:57.049 --> 00:01:59.530 but i mean the the real momentum right now is 00:01:59.530 --> 00:02:02.159 in crypto right oh exactly yeah Bitcoin is having 00:02:02.159 --> 00:02:04.579 just a really fascinating week. So it was up 00:02:04.579 --> 00:02:06.719 fractionally yesterday about zero point zero 00:02:06.719 --> 00:02:08.819 six percent, putting it at seventy nine thousand 00:02:08.819 --> 00:02:11.479 thirty point two nine. Which doesn't sound like 00:02:11.479 --> 00:02:13.479 a lot. No. Right. For one day. But if you zoom 00:02:13.479 --> 00:02:15.400 out just a few days, it's actually up over eight 00:02:15.400 --> 00:02:18.159 percent since last Friday alone. Wow. Yeah. It's 00:02:18.159 --> 00:02:20.020 definitely on a tear showing some major resilience 00:02:20.020 --> 00:02:21.780 in the market right now. Eight percent in the 00:02:21.780 --> 00:02:24.099 weekend is wild. Definitely a space to keep an 00:02:24.099 --> 00:02:26.340 eye on. And for our sports fans, just a quick 00:02:26.340 --> 00:02:29.210 baseball update from yesterday's action. The 00:02:29.210 --> 00:02:31.669 SF Giants absolutely dominated the Cincinnati 00:02:31.669 --> 00:02:33.849 Reds. Oh, they did. Yeah, shutting them out five 00:02:33.849 --> 00:02:35.789 down. Jason Foley was the winning pitcher on 00:02:35.789 --> 00:02:37.669 that one, and he was just commanding the mound 00:02:37.669 --> 00:02:39.590 beautifully. Yeah, he looked great out there. 00:02:39.650 --> 00:02:42.569 He really did. And well, in a much higher scoring 00:02:42.569 --> 00:02:44.830 affair, the Athletics lost a pretty tough game 00:02:44.830 --> 00:02:48.370 to the Minnesota Twins six to nine. Ah, tough 00:02:48.370 --> 00:02:50.770 break. Yeah, with Zebby Matthews taking the win 00:02:50.770 --> 00:02:54.360 on the mound for the Twins. Right. Well. Taking 00:02:54.360 --> 00:02:57.120 a pivot from the baseball diamond over to the 00:02:57.120 --> 00:03:00.259 tech world, we have a really fascinating narrative 00:03:00.259 --> 00:03:02.960 today regarding the future of where our data 00:03:02.960 --> 00:03:06.259 actually lives. So let's unpack this top tech 00:03:06.259 --> 00:03:08.800 news, but before we dive into Apple's latest 00:03:08.800 --> 00:03:11.020 silicon announcements, let's hear a quick word 00:03:11.020 --> 00:03:13.180 from our sponsor. Today's episode is brought 00:03:13.180 --> 00:03:15.840 to you by Strong Coffee. the premium instant 00:03:15.840 --> 00:03:18.280 coffee that tastes like a traditional brew, but 00:03:18.280 --> 00:03:21.599 is packed with 15 grams of protein, MCTs, and 00:03:21.599 --> 00:03:24.439 adaptogens like ashwagandha. Get clean, sustained 00:03:24.439 --> 00:03:27.620 energy without the jitters. Use promo code SNARFL 00:03:27.620 --> 00:03:31.560 for 20 % off at strongcoffeecompany .com. So 00:03:31.560 --> 00:03:33.979 when we talk about desktop computers historically, 00:03:35.680 --> 00:03:38.139 there has always been a very specific expectation, 00:03:38.280 --> 00:03:41.039 right? It's a box. It sits on a desk. And it 00:03:41.039 --> 00:03:42.960 connects to the internet so you can access all 00:03:42.960 --> 00:03:45.180 the heavy lifting servers out in the world. Exactly. 00:03:45.360 --> 00:03:48.300 For like the last 10 to 15 years, the trend has 00:03:48.300 --> 00:03:51.139 been to make the local computer as thin and light 00:03:51.139 --> 00:03:53.879 as possible while sending all the actual processing 00:03:53.879 --> 00:03:57.159 power to these remote server farms. I mean, everything 00:03:57.159 --> 00:03:59.580 from your photo storage to running AI chatbots, 00:03:59.819 --> 00:04:02.699 it all happens in the cloud. But based on yesterday's 00:04:02.699 --> 00:04:05.199 announcements, it looks like Apple is just aggressively 00:04:05.199 --> 00:04:07.240 trying to pull that pendulum back in the other 00:04:07.240 --> 00:04:09.500 direction. So let's look closer at their new 00:04:09.500 --> 00:04:11.500 silicon. Announcements and I want to start with 00:04:11.500 --> 00:04:14.120 the most accessible machine in their lineup the 00:04:14.120 --> 00:04:17.199 the brand new Mac Mini Yeah, the Mac Mini has 00:04:17.199 --> 00:04:19.740 always been that you know entry -level stepping 00:04:19.740 --> 00:04:23.480 stone, but this new iteration Completely rewrites 00:04:23.480 --> 00:04:25.319 what an entry -level machine is even capable 00:04:25.319 --> 00:04:28.709 of so they're doing the brand new m6 chip And 00:04:28.709 --> 00:04:32.709 the base model starts at $899. Okay, $899. Which 00:04:32.709 --> 00:04:36.189 gets you that M6 processor, 256 gigabytes of 00:04:36.189 --> 00:04:38.949 storage, and 16 gigabytes of RAM right out of 00:04:38.949 --> 00:04:41.550 the gate. Nice. And if you want to bump it up, 00:04:41.670 --> 00:04:44.269 there is an M5 Pro version that starts at $1 00:04:44.269 --> 00:04:47.209 ,699, with both of those models actually starting 00:04:47.209 --> 00:04:50.149 to ship on September 22nd. But I want to pause 00:04:50.149 --> 00:04:52.870 on the M6 chip itself. Sure. Because the architectural 00:04:52.870 --> 00:04:55.649 leap here is what's truly driving this shift 00:04:55.649 --> 00:04:57.910 away from the cloud. The specs say it features 00:04:57.910 --> 00:05:01.610 a 12 -core CPU and a 12 -core GPU, which is up 00:05:01.610 --> 00:05:04.170 from the 10 -core setup in the older M4. Plus, 00:05:04.269 --> 00:05:06.410 we're seeing Wi -Fi 7 and Bluetooth 6. Pretty 00:05:06.410 --> 00:05:08.750 standard for a spec bump, yeah. Right. But the 00:05:08.750 --> 00:05:10.750 phrase that kept coming up over and over in the 00:05:10.750 --> 00:05:12.889 source material is that it's built on a two nanometer 00:05:12.889 --> 00:05:15.529 process. What does that actually mean under the 00:05:15.529 --> 00:05:17.569 hood? I mean, I like to think of it kind of like... 00:05:17.500 --> 00:05:20.879 Packing a massive v8 engine into a tiny compact 00:05:20.879 --> 00:05:24.199 car. It's getting smaller but somehow wildly 00:05:24.199 --> 00:05:27.740 more efficient and powerful That is a great analogy 00:05:27.740 --> 00:05:30.360 because two nanometers sounds like this abstract 00:05:30.360 --> 00:05:33.120 marketing jargon, right? But it's actually entirely 00:05:33.120 --> 00:05:36.139 rooted in physics. So think of a computer chip 00:05:36.139 --> 00:05:39.879 like a massive sprawling city Okay, historically 00:05:39.879 --> 00:05:42.600 the businesses the homes the factories, which 00:05:42.600 --> 00:05:45.439 are the transistors in this analogy. They were 00:05:45.439 --> 00:05:47.420 spread out, right? And that meant the electrons 00:05:47.420 --> 00:05:49.540 had to commute long distances across the chip 00:05:49.540 --> 00:05:52.199 just to get anything done. And in computing, 00:05:52.540 --> 00:05:54.939 a long commute generates friction, and friction 00:05:54.939 --> 00:05:58.500 creates heat. Exactly. Heat is the ultimate enemy 00:05:58.500 --> 00:06:01.100 of performance. If a chip gets too hot, the computer 00:06:01.100 --> 00:06:02.939 just has to artificially slow it down to prevent 00:06:02.939 --> 00:06:05.930 the whole thing from literally melting. So by 00:06:05.930 --> 00:06:09.209 moving to a two nanometer process, Apple is essentially 00:06:09.209 --> 00:06:11.709 shrinking the layout of that entire city. They 00:06:11.709 --> 00:06:13.629 are moving all those microscopic transistors 00:06:13.629 --> 00:06:15.750 incredibly close together. Okay, so the electrons 00:06:15.750 --> 00:06:18.069 have practically zero distance to travel. Right. 00:06:18.269 --> 00:06:19.930 Nobody has to commute. Oh, that makes sense. 00:06:20.050 --> 00:06:22.009 And because that distance is so ridiculously 00:06:22.009 --> 00:06:25.410 short, almost no energy is wasted as heat. So 00:06:25.410 --> 00:06:28.449 the chip doesn't just run faster, it runs incredibly 00:06:28.449 --> 00:06:32.439 cool and uses significantly less power. That 00:06:32.439 --> 00:06:35.379 is how you can pack a massive 12 core processing 00:06:35.379 --> 00:06:38.980 unit into this tiny silent silver box on your 00:06:38.980 --> 00:06:42.279 desk without the fan screaming or the thing overheating. 00:06:42.360 --> 00:06:44.779 Okay, I get the physics of it now, but here is 00:06:44.779 --> 00:06:46.439 where I kind of want to push back a little bit. 00:06:46.500 --> 00:06:49.480 Okay, go for it. Apple is stating this M6 chip 00:06:49.480 --> 00:06:52.779 boasts up to four times faster AI performance 00:06:52.779 --> 00:06:56.019 than the previous generation. Forex is a huge 00:06:56.019 --> 00:06:58.819 job it's massive but for a normal family rate 00:06:58.819 --> 00:07:01.160 of a family that is just running microsoft excel 00:07:01.160 --> 00:07:04.060 to manage their budget or or just browsing safari 00:07:04.060 --> 00:07:07.459 for recipes. Why on earth do they need a forex 00:07:07.459 --> 00:07:10.459 leap in a processing like is this just overkill 00:07:10.459 --> 00:07:13.180 to sell a fancy new machine. Well, if you only 00:07:13.180 --> 00:07:14.959 look at the computer as a faster web browser, 00:07:15.120 --> 00:07:17.000 then yes, it's total overkill. Right. But if 00:07:17.000 --> 00:07:18.959 we connect this to the bigger picture of where 00:07:18.959 --> 00:07:21.319 all software is currently heading, it's about 00:07:21.319 --> 00:07:24.160 a fundamental shift toward what the tech industry 00:07:24.160 --> 00:07:27.079 is calling agentic computing. Agentic computing. 00:07:27.259 --> 00:07:30.300 OK, let's unpack that term. How does agentic 00:07:30.300 --> 00:07:32.779 computing differ from the AI we're already used 00:07:32.779 --> 00:07:36.839 to? So until now, AI has mostly been a chatbot. 00:07:37.129 --> 00:07:39.389 You go to a website, you type a prompt into a 00:07:39.389 --> 00:07:42.029 text box, and you sit there and wait for a server 00:07:42.029 --> 00:07:44.310 hundreds of miles away to generate an answer. 00:07:44.550 --> 00:07:46.670 Right. It's very transactional. Exactly. It's 00:07:46.670 --> 00:07:49.269 completely reactive. Agentic computing means 00:07:49.269 --> 00:07:52.810 the computer has AI models or agents running 00:07:52.810 --> 00:07:56.250 silently in the background all the time, proactively 00:07:56.250 --> 00:07:59.709 doing complex multi -step tasks for you without 00:07:59.709 --> 00:08:01.470 you having to prompt it every step of the way. 00:08:01.560 --> 00:08:04.220 So instead of me specifically asking it a question, 00:08:04.379 --> 00:08:06.720 it's actively working on my behalf, like an assistant 00:08:06.720 --> 00:08:09.420 that's actually anticipating what I need. Precisely. 00:08:09.759 --> 00:08:11.639 And because of this incredibly efficient new 00:08:11.639 --> 00:08:14.439 M6 chip, the Mac Mini is becoming really popular 00:08:14.439 --> 00:08:16.420 among tech enthusiasts for running these local 00:08:16.420 --> 00:08:19.120 AI agents natively. There's an open source model 00:08:19.120 --> 00:08:21.139 called OpenClaw, for example, that you can run 00:08:21.139 --> 00:08:23.980 entirely on the device. Oh, I see. So instead 00:08:23.980 --> 00:08:26.660 of taking your family's personal highly sensitive 00:08:26.660 --> 00:08:29.420 data like your banking emails or private photos 00:08:29.420 --> 00:08:31.720 of your kids and sending all of that up to some 00:08:31.720 --> 00:08:34.820 massive cloud server to be analyzed, this Mac 00:08:34.820 --> 00:08:36.799 Mini is strong enough to process all of that 00:08:36.799 --> 00:08:39.419 locally on the desk. That is the crucial takeaway 00:08:39.419 --> 00:08:42.570 right there. It suddenly brings massive totally 00:08:42.570 --> 00:08:45.370 private AI capabilities right into the home. 00:08:45.710 --> 00:08:48.129 It can index your family photos, it can organize 00:08:48.129 --> 00:08:50.889 your finances, or summarize long chains of emails 00:08:50.889 --> 00:08:53.610 natively without a single bite of your data ever 00:08:53.610 --> 00:08:56.830 leaving your home Wi -Fi network. So that 4x 00:08:56.830 --> 00:08:59.610 AI leap isn't about loading web pages faster, 00:09:00.049 --> 00:09:01.789 it's about giving you a private intelligence 00:09:01.789 --> 00:09:04.009 that lives securely in that little silver box. 00:09:04.200 --> 00:09:06.840 That is a massive paradigm shift, but I guess 00:09:06.840 --> 00:09:09.000 it begs the question, if the Mac Mini is this 00:09:09.000 --> 00:09:12.019 compact daily driver for the home, what happens 00:09:12.019 --> 00:09:14.600 when you need industry -level, just uncompromised 00:09:14.600 --> 00:09:16.360 horsepower? Because they certainly didn't stop 00:09:16.360 --> 00:09:18.360 at the Mini today. No, they did not. They also 00:09:18.360 --> 00:09:21.070 announced the newly updated Mac Studio. Yes, 00:09:21.070 --> 00:09:23.370 and this is where the hardware capabilities get 00:09:23.370 --> 00:09:26.509 truly staggering. The Mac Studio is aimed squarely 00:09:26.509 --> 00:09:28.490 at the highest end of professional workflows, 00:09:28.549 --> 00:09:32.830 so it starts at $2 ,499 for a machine equipped 00:09:32.830 --> 00:09:36.570 with the M5 Max chip, and that comes with 36GB 00:09:36.570 --> 00:09:40.090 of unified memory and 512GB of storage. Which 00:09:40.090 --> 00:09:42.490 is already a beast, but then there's the M5 Ultra 00:09:42.490 --> 00:09:44.830 model. Right, the M5 Ultra. That one starts at 00:09:44.830 --> 00:09:48.549 $5 ,499, which is actually a $200 bump from the 00:09:48.549 --> 00:09:51.200 previous generation. But for that premium, you 00:09:51.200 --> 00:09:52.899 were getting a machine that could be configured 00:09:52.899 --> 00:09:56.019 with up to 512 gigabytes of unified memory. Wait, 00:09:56.120 --> 00:09:58.879 wait, wait. Let's pause right there. Yeah. 512 00:09:58.879 --> 00:10:01.899 gigabytes of unified RAM in a desktop computer. 00:10:01.960 --> 00:10:04.580 It's insane. My current laptop has 8 gigabytes 00:10:04.580 --> 00:10:06.639 of RAM. And honestly, it handles everything I 00:10:06.639 --> 00:10:08.879 throw at it perfectly fine. Why on earth would 00:10:08.879 --> 00:10:12.179 any desktop ever need 512 gigabytes of memory? 00:10:12.519 --> 00:10:14.080 Well, for traditional computing, it absolutely 00:10:14.080 --> 00:10:16.139 doesn't. But we have to look at how artificial 00:10:16.139 --> 00:10:18.500 intelligence actually operates at a professional 00:10:18.500 --> 00:10:21.470 scale. When developers or data scientists want 00:10:21.470 --> 00:10:24.850 to run a large language model in LLM, they have 00:10:24.850 --> 00:10:28.549 to load the entire brain of that AI into computer 00:10:28.549 --> 00:10:31.789 short -term memory, or RAM, all at once. Oh, 00:10:31.889 --> 00:10:34.889 so the intelligence of the AI is directly bottlenecked 00:10:34.889 --> 00:10:37.629 by how much memory the computer has? If the brain 00:10:37.629 --> 00:10:40.250 is bigger than the RAM, it won't work. Exactly. 00:10:40.409 --> 00:10:43.029 If an AI model is too big, it simply won't fit 00:10:43.029 --> 00:10:44.950 into your RAM and the computer just crashes. 00:10:45.149 --> 00:10:48.309 It won't even start. So by offering 512 gigabytes 00:10:48.309 --> 00:10:50.870 of unified memory, professionals can now run 00:10:50.870 --> 00:10:54.330 massive frontier class LLMs entirely on their 00:10:54.330 --> 00:10:57.629 devices. We're talking about AI models that previously 00:10:57.629 --> 00:10:59.870 required literally millions of dollars in cloud 00:10:59.870 --> 00:11:02.230 server infrastructure just to boot up. And that 00:11:02.230 --> 00:11:04.929 solves a colossal problem for enterprise businesses. 00:11:05.419 --> 00:11:08.179 Because the privacy implications here are enormous. 00:11:08.620 --> 00:11:10.840 Huge. If you are a hospital dealing with patient 00:11:10.840 --> 00:11:13.980 records or a high -end law firm analyzing confidential 00:11:13.980 --> 00:11:17.179 client contracts, you simply cannot legally or 00:11:17.179 --> 00:11:19.399 ethically upload that sensitive data to a public 00:11:19.399 --> 00:11:22.000 AI cloud. No, absolutely not. With this new Mac 00:11:22.000 --> 00:11:24.299 Studio, the data never has to leave the desk. 00:11:24.559 --> 00:11:26.659 It completely eliminates those corporate privacy 00:11:26.659 --> 00:11:29.519 hurdles. Yeah. And it also eliminates the incredibly 00:11:29.519 --> 00:11:32.860 costly cloud computing fees. And you buy the 00:11:32.860 --> 00:11:35.899 machine once, and you run your massive AI models 00:11:35.899 --> 00:11:39.200 247 without paying per token or per minute of 00:11:39.200 --> 00:11:42.080 server time. Plus, Apple has introduced Thunderbolt 00:11:42.080 --> 00:11:45.440 5 to these machines, which includes this brilliant 00:11:45.440 --> 00:11:47.940 clustering feature. OK, I found this part of 00:11:47.940 --> 00:11:50.919 the report so fascinating. To me, it's essentially 00:11:50.919 --> 00:11:53.919 like assembling Voltron. Oh, I love that. Right. 00:11:54.139 --> 00:11:56.720 You take individual Mac studios, you link them 00:11:56.720 --> 00:11:59.500 together using these insanely fast Thunderbolt 00:11:59.500 --> 00:12:02.759 5 cables, and they share their memory pools to 00:12:02.759 --> 00:12:06.779 create one massive supercomputer. You're literally 00:12:06.779 --> 00:12:10.580 combining the robotic lines into one giant unstoppable 00:12:10.580 --> 00:12:13.139 robot. That is a really fun way to look at it. 00:12:13.129 --> 00:12:15.269 But technically, this is powered by something 00:12:15.269 --> 00:12:18.450 called RDMA, or Remote Direct Memory Access. 00:12:18.590 --> 00:12:20.610 Okay, let's unpack RDMA. Keep it simple for me, 00:12:20.710 --> 00:12:22.389 though. Right. In simple terms, it is a hardware 00:12:22.389 --> 00:12:24.750 protocol that allows one computer to reach directly 00:12:24.750 --> 00:12:27.070 into the brain or the memory of another connected 00:12:27.070 --> 00:12:29.830 computer instantly. It completely bypasses the 00:12:29.830 --> 00:12:31.990 operating system, so there's no software asking 00:12:31.990 --> 00:12:34.590 for permission and slowing things down. It just 00:12:34.590 --> 00:12:37.450 grabs what it needs. So if one Mac Studio with 00:12:37.450 --> 00:12:40.429 512 gigs of RAM isn't enough for the massive 00:12:40.429 --> 00:12:43.289 AI model you want to run... four of them together. 00:12:43.750 --> 00:12:46.409 Exactly. Suddenly, you have over two terabytes 00:12:46.409 --> 00:12:49.710 of shared, ultra -fast memory processing your 00:12:49.710 --> 00:12:53.429 data as a single unified entity. Which is mind 00:12:53.429 --> 00:12:55.529 -blowing. And it circles right back to the theme 00:12:55.529 --> 00:12:58.690 we started with. We have spent well over a decade 00:12:58.690 --> 00:13:01.350 moving absolutely everything to the cloud. Our 00:13:01.350 --> 00:13:03.610 documents, our high -end processing, all our 00:13:03.610 --> 00:13:05.409 enterprise software, it's all running in a remote 00:13:05.409 --> 00:13:08.669 server farms. Are we really witnessing a massive 00:13:08.669 --> 00:13:10.710 U -turn back to local computing? We absolutely 00:13:10.710 --> 00:13:13.250 are, specifically for these high -end compute 00:13:13.250 --> 00:13:16.129 -heavy tasks. Yeah. Because the cloud is amazing 00:13:16.129 --> 00:13:17.950 for convenience, right, sensing your data across 00:13:17.950 --> 00:13:20.570 your phone and your laptop. But for heavy compute, 00:13:20.750 --> 00:13:23.250 it introduces a major bottleneck, which is latency. 00:13:23.509 --> 00:13:26.110 Just the time it takes for data to travel. Exactly. 00:13:26.590 --> 00:13:28.710 Latency is the time it takes for your data to 00:13:28.710 --> 00:13:31.149 travel from your computer to a server in Virginia, 00:13:31.370 --> 00:13:34.279 be processed, and travel all the way back. By 00:13:34.279 --> 00:13:36.460 giving professionals the power of a server farm 00:13:36.460 --> 00:13:39.740 locally on their desk, you eliminate the latency 00:13:39.740 --> 00:13:42.580 entirely and you permanently cap the financial 00:13:42.580 --> 00:13:45.500 cost. It's an incredible paradigm shift. We've 00:13:45.500 --> 00:13:48.139 seen the physical machines and we've explored 00:13:48.139 --> 00:13:50.919 the raw hardware specs. But let's look closer 00:13:50.919 --> 00:13:53.480 at how Apple is actually bringing all this silicon 00:13:53.480 --> 00:13:56.059 to life under the hood. Specifically, I want 00:13:56.059 --> 00:13:58.960 to talk about this M5 Ultra chip and the software 00:13:58.960 --> 00:14:02.120 because raw metal is totally useless without 00:14:02.120 --> 00:14:04.750 the right architecture. Oh, absolutely. The hardware 00:14:04.750 --> 00:14:07.190 is only half the story. If we look at the M5 00:14:07.190 --> 00:14:09.870 Ultra chip itself, the way they achieve this 00:14:09.870 --> 00:14:11.950 level of performance is through a specialized 00:14:11.950 --> 00:14:15.049 technology they call Ultra Fusion. Right, and 00:14:15.049 --> 00:14:16.690 how does it actually work? Because from what 00:14:16.690 --> 00:14:19.429 I read, they are quite literally taking two separate 00:14:19.429 --> 00:14:22.590 M5 Max chips and stitching them together to create 00:14:22.590 --> 00:14:25.269 a massive quad -y processor. Yep, that's exactly 00:14:25.269 --> 00:14:27.509 what it is. But physically bridging two computer 00:14:27.509 --> 00:14:30.590 chips without losing speed has historically been 00:14:30.590 --> 00:14:33.330 like one of the hardest physics problems in computing? 00:14:33.710 --> 00:14:36.610 It is incredibly difficult. Normally, if you 00:14:36.610 --> 00:14:38.289 connect two separate chips on a motherboard, 00:14:38.870 --> 00:14:41.049 the data traffic jam between them slows everything 00:14:41.049 --> 00:14:45.519 down to a crawl. But Ultra Fusion uses a microscopic 00:14:45.519 --> 00:14:48.379 silicon interposer, which is essentially a high 00:14:48.379 --> 00:14:50.759 -density bridge that connects the chip so perfectly 00:14:50.759 --> 00:14:53.960 with so many tiny pathways that the software 00:14:53.960 --> 00:14:56.320 literally cannot tell it looking at two separate 00:14:56.320 --> 00:14:58.740 pieces of silicon. Oh, wow. So the computer just 00:14:58.740 --> 00:15:01.500 treats it as one giant brain. Exactly. And the 00:15:01.500 --> 00:15:04.679 result of that flawless connection is a mind 00:15:04.679 --> 00:15:08.360 -bending 1 .2 terabytes per second of memory 00:15:08.360 --> 00:15:10.399 bandwidth. To put that in perspective, you could 00:15:10.639 --> 00:15:13.279 transfer entire hard drives worth of data instantly 00:15:13.279 --> 00:15:15.059 across the memory. In a fraction of a second. 00:15:15.100 --> 00:15:17.600 Yeah. And to control all of this unprecedented 00:15:17.600 --> 00:15:20.440 local power, both the new Mac Mini and the Mac 00:15:20.440 --> 00:15:22.559 Studio are launching with a brand new operating 00:15:22.559 --> 00:15:25.759 system. Right. Mac OS 27, codenamed Golden Gate. 00:15:26.100 --> 00:15:28.019 Golden Gate. And it seems like Golden Gate is 00:15:28.019 --> 00:15:30.580 completely built around the deep native integration 00:15:30.580 --> 00:15:33.080 of Siri AI. And to be clear, this isn't the Siri 00:15:33.080 --> 00:15:34.980 we've been using for the last 10 years just to 00:15:34.980 --> 00:15:36.700 set kitchen timers or ask about the weather. 00:15:37.360 --> 00:15:41.509 Not at all. Siri AI is an entirely new architecture 00:15:41.509 --> 00:15:45.049 that possesses deep personal context. Because 00:15:45.049 --> 00:15:47.190 all that processing power is local, like we talked 00:15:47.190 --> 00:15:50.009 about. It can constantly index and search across 00:15:50.009 --> 00:15:52.250 your private emails, your text messages, your 00:15:52.250 --> 00:15:54.629 calendar, and your photos to connect the dots 00:15:54.629 --> 00:15:56.409 of your life securely. Without sending it to 00:15:56.409 --> 00:15:59.029 the cloud. Right. They also highlighted a feature 00:15:59.029 --> 00:16:02.190 called visual intelligence. How does that practically 00:16:02.190 --> 00:16:05.389 work for the end user? Well, visual intelligence 00:16:05.389 --> 00:16:09.539 is essentially giving the AI Eyes. Okay. Using 00:16:09.539 --> 00:16:12.360 a simple keyboard shortcut, you can ask Siri 00:16:12.360 --> 00:16:14.500 AI about whatever is currently visible on your 00:16:14.500 --> 00:16:17.480 screen. Say you have a massive complicated spreadsheet 00:16:17.480 --> 00:16:20.519 open or a complex architectural diagram. Okay, 00:16:20.580 --> 00:16:22.440 or maybe even just a recipe you found online. 00:16:22.559 --> 00:16:25.059 Sure, exactly. You hit the shortcut and the local 00:16:25.059 --> 00:16:27.519 AI analyzes the screen's contents, summarizes 00:16:27.519 --> 00:16:29.860 the data, or acts on it like drafting an email 00:16:29.860 --> 00:16:31.600 to your team based on the spreadsheet numbers 00:16:31.600 --> 00:16:33.360 you're looking at right that second. That is 00:16:33.360 --> 00:16:35.919 so wild. And they've updated the overall look 00:16:35.919 --> 00:16:38.100 and feel of the OS too with something they are 00:16:38.100 --> 00:16:41.500 calling liquid glass, which adds uniform toolbars 00:16:41.500 --> 00:16:44.279 and edge to edge sidebars, just making the whole 00:16:44.279 --> 00:16:47.789 OS feel highly responsive. Gorgeous. It really 00:16:47.789 --> 00:16:50.370 does. I also think it's worth briefly highlighting 00:16:50.370 --> 00:16:52.769 the environmental angle Apple included in the 00:16:52.769 --> 00:16:55.070 report today. Oh, definitely. Yeah, the new Mac 00:16:55.070 --> 00:16:58.370 Mini is made of 50 % recycled material overall 00:16:58.370 --> 00:17:01.629 and the Mac Studio sits at 35 % but is built 00:17:01.629 --> 00:17:04.869 with 100 % recycled aluminum enclosures. Which 00:17:04.869 --> 00:17:07.369 is a significant step toward their stated corporate 00:17:07.369 --> 00:17:10.049 goal of being completely carbon neutral across 00:17:10.049 --> 00:17:12.650 their entire manufacturing footprint by 2030. 00:17:13.240 --> 00:17:15.519 They are even matching the electricity customers 00:17:15.519 --> 00:17:18.519 use to power the Mac Mini with direct investments 00:17:18.519 --> 00:17:20.599 in renewable energy. That's great. But I want 00:17:20.599 --> 00:17:23.920 to pull us back to Mac OS 27 and Siri AI for 00:17:23.920 --> 00:17:25.559 a second, because this is where I start to wonder 00:17:25.559 --> 00:17:28.839 about the philosophical implications of all this. 00:17:28.920 --> 00:17:32.519 OK, let's hear it. With Siri AI now able to actively 00:17:32.519 --> 00:17:36.559 read our screens, write emails for us, edit our 00:17:36.559 --> 00:17:40.299 documents, and draw on our deeply personal context. 00:17:40.750 --> 00:17:43.450 At what point does the operating system cross 00:17:43.450 --> 00:17:46.210 the line from being a tool we use to an active 00:17:46.210 --> 00:17:48.609 agent that practically does our jobs for us? 00:17:49.009 --> 00:17:51.809 Wow, yeah. I mean, that is arguably the defining 00:17:51.809 --> 00:17:54.250 philosophical question of this new era of computing. 00:17:54.430 --> 00:17:57.329 Right, because a tool, by definition, requires 00:17:57.329 --> 00:18:00.680 a human to swing the hammer. But an agentic AI, 00:18:00.940 --> 00:18:03.880 which is what Mac OS 27 and these new M -series 00:18:03.880 --> 00:18:06.680 chips are leaning heavily into, swings the hammer 00:18:06.680 --> 00:18:08.700 for you while you just point at the nail. Or 00:18:08.700 --> 00:18:11.259 honestly, maybe the AI just decides which nail 00:18:11.259 --> 00:18:13.359 needs hitting based on analyzing your calendar 00:18:13.359 --> 00:18:15.660 and your pending emails. Exactly. And on one 00:18:15.660 --> 00:18:17.619 hand, that can be incredibly liberating. It frees 00:18:17.619 --> 00:18:19.680 us up from the mundane, repetitive tasks, you 00:18:19.680 --> 00:18:22.059 know, the endless digital paperwork of our modern 00:18:22.059 --> 00:18:24.759 lives. But it fundamentally challenges our relationship 00:18:24.759 --> 00:18:27.200 with our computers. It really does. We have to 00:18:27.200 --> 00:18:29.500 trust that the AI understands are human intent 00:18:29.500 --> 00:18:31.859 perfectly. And we have to be comfortable with 00:18:31.859 --> 00:18:34.519 a machine having that level of intimate unregulated 00:18:34.519 --> 00:18:37.160 access to our digital lives. Which brings us 00:18:37.160 --> 00:18:39.900 full circle right back to why running these models 00:18:39.900 --> 00:18:42.819 locally on your own hardware rather than in the 00:18:42.819 --> 00:18:46.470 cloud is so appealing. The trust factor is infinitely 00:18:46.470 --> 00:18:49.390 higher when the brain doing the thinking is a 00:18:49.390 --> 00:18:51.369 silver box sitting right there on your physical 00:18:51.369 --> 00:18:53.769 desk, entirely disconnected from the broader 00:18:53.769 --> 00:18:55.890 internet if you choose. Precisely. It creates 00:18:55.890 --> 00:18:58.450 a closed loop of privacy. You get the intelligence 00:18:58.450 --> 00:19:00.309 of the cloud, but with the security of a vault. 00:19:00.609 --> 00:19:03.049 Before we wrap up today's analysis, a quick thanks 00:19:03.049 --> 00:19:05.910 to Old Glory, an iconic music and sports fan 00:19:05.910 --> 00:19:08.410 merch store, perfect for the techie in your life. 00:19:08.829 --> 00:19:12.069 Use promo code SNARFL for 15 % off at oldglory 00:19:12.069 --> 00:19:15.839 .com. So just to quickly recap the rather expansive 00:19:15.839 --> 00:19:17.740 journey we've been on today with this massive 00:19:17.740 --> 00:19:20.380 tech report, we started by looking at the incredible 00:19:20.380 --> 00:19:23.079 everyday accessibility of the new Mac Mini. By 00:19:23.079 --> 00:19:26.099 utilizing that hyper -efficient 2nm M6 chip, 00:19:26.660 --> 00:19:29.440 Apple is proving that true, private AI power 00:19:29.440 --> 00:19:32.160 is finally making its way directly into the family 00:19:32.160 --> 00:19:35.619 home for under $1000. Then we scaled all the 00:19:35.619 --> 00:19:37.940 way up to the professional, boundary -pushing 00:19:37.940 --> 00:19:41.410 Mac Studio with the M5 Ultra. We explored how 00:19:41.410 --> 00:19:43.789 creatives, hospitals, and enterprise businesses 00:19:43.789 --> 00:19:46.670 can finally break free from the expensive, latency 00:19:46.670 --> 00:19:49.450 -heavy cloud by using Thunderbolt 5 clustering 00:19:49.450 --> 00:19:52.730 and massive local memory pools to run LLMs right 00:19:52.730 --> 00:19:55.569 on their desks. And finally, we saw how Mac OS 00:19:55.569 --> 00:19:58.509 27 Golden Gate is weaving all this raw hardware 00:19:58.509 --> 00:20:00.809 together, turning our computers from passive 00:20:00.809 --> 00:20:03.990 tools into active, intelligent agents with Siri 00:20:03.990 --> 00:20:06.890 AI and visual intelligence. It really is a massive 00:20:06.890 --> 00:20:09.130 leap forward across the entire spectrum of how 00:20:09.130 --> 00:20:10.930 we interact. with machines. It truly is. And 00:20:10.930 --> 00:20:12.789 it leaves me with this final thought for everyone 00:20:12.789 --> 00:20:14.650 listening to Chuan as you go about your day. 00:20:14.690 --> 00:20:17.289 Yeah. If desktop computers, machines we can just 00:20:17.289 --> 00:20:19.849 buy and put in our homes and offices can now 00:20:19.849 --> 00:20:22.609 run these incredibly powerful frontier class 00:20:22.609 --> 00:20:25.190 open weight AI models completely locally and 00:20:25.190 --> 00:20:28.190 entirely privately How will this disrupt the 00:20:28.190 --> 00:20:30.809 multi -trillion dollar cloud computing industry? 00:20:30.930 --> 00:20:33.089 Yeah, that's the big question. Are we witnessing 00:20:33.089 --> 00:20:36.069 the great decentralization of artificial intelligence, 00:20:36.369 --> 00:20:38.990 moving it straight out of massive corporate server 00:20:38.990 --> 00:20:41.250 farms and directly into the hands of the consumer? 00:20:41.910 --> 00:20:44.470 It's a fascinating shift in the balance of technological 00:20:44.470 --> 00:20:46.950 power. We're here to help at Snorful, so reach 00:20:46.950 --> 00:20:49.250 out on our website if you have any more questions. 00:20:52.769 --> 00:20:55.890 That's a wrap on today's top stories. Remember, 00:20:56.150 --> 00:20:58.210 tech doesn't have to be complicated, and we're 00:20:58.210 --> 00:21:00.829 here to help connect the dots. If you've got 00:21:00.829 --> 00:21:03.250 questions or want to learn more, visit us at 00:21:03.250 --> 00:21:06.329 snarful .com. Thanks for listening, and we'll 00:21:06.329 --> 00:21:07.190 catch you next time.