Tech Brew Ride Home - New Macs

Starting point is 00:00:04 Welcome to the TechBoo ride home for Tuesday, August 25th, 2026. I'm Brian McCullough today. Apple refreshed the Mac Studio with M5 Max and M5 Ultra Chips and gave the Mac Mini M6 Silicon all at higher prices. Opening eye is jalapeno chip, beat Nvidia on efficiency, perplexity went fully local, WhatsApp, toughened logins, and Uber is letting parents live stream when your kid rides. Here's what you miss today in the world of tech. Every day, shareholders meet to discuss important matters about the companies you invest in. you can make your voice heard too. Vanguard investor choice makes it easy to set your proxy voting preference for your eligible Vanguard index funds. Whether you hold a Vanguard fund directly or through another brokerage firm, all it takes is a few clicks to select your proxy voting preference and be heard on important shareholder topics like executive pay and director elections. Visit vanguard.com slash investor choice to learn more. It's your shares. It's your voice. It's easy. Vanguard investors own shares of our index funds. And those, Those funds own shares of the companies they invest in, Vanguard Marketing Corporation distributor. Apple has updated the Macs Studio with M5 Max and M5 Ultra Chips with up to 4.3 times faster AI performance, faster graphics, and up to 512 gigabytes of unified memory for $2,500.

Starting point is 00:01:30 They also unveiled a Mac Mini with M6 and M5 Pro chips with up to four times faster AI performance and up to two times faster graphics for 900 bucks plus with the M6 and 1700 bucks plus with the M5 Pro. Now, before we get into the details, notable that the M6 and M5 Pro Mac minis are $100 above the previous M4 models in terms of price and the M5 Ultra Mac Studio is up $200. Quoting the Verge. While pre-orders start today, the new Mac minis won't ship until September 22nd. Hopefully enough time for Apple. to stock up on inventory as the last-gen models became scarce due to higher than expected demand for running AI agents. Like last generation, the new entry-level Mac Mini with an M6 chip starts

Starting point is 00:02:20 with 16 gigabytes of unified memory and 256 gigabytes of storage, and the M5 Pro Mac Mini starts with 24 gigabytes and 512-gabyte configurations all in on the edges. But the storage in all of these models is now twice as fast, just like the recent crop of MacBook's Air and Pro. In addition to the double-speed storage, both M6 and M5 Pro Mac minis now support 2.5 gigabit Ethernet ports with the same optional upgrade to 10 gigabit. And they use Apple's N1 chip for Wi-Fi 7 and Bluetooth 6 support. The M6 Mac Mini now has a 12-core CPU and 12-core GPU along with 170 GbPS of memory bandwidth compared to the M4's 10 core, 10 core setup, and memory bandwidth of just 120 GbPS. Its CPU consists of two supercores, a naming convention that began with the M5 Pro and

Starting point is 00:03:16 Max chips, four performance cores and six efficiency cores, and it's Apple's first chip on a two-nanometer process. Apple claims the M6 more faster cores lead up to a 40% uplift in CPU performance over the M4 generation, and the new 12-core GPU with neural accelerators is up to twice as fast at graphics rendering with ray tracing than the M4 in some games. The M5 Pro of the new Mac Mini is configurable, much like it is on the MacBook Pro. It can be had with up to 18 CPU cores and 20 GPU cores, and it has the same 307 GbPS of memory bandwidth. Apple claims the Mac Mini with an M5 Pro has significant gains in AI compute and graphics over the outgoing M4 Pro model, thanks in part to the neural accelerator in

Starting point is 00:04:04 each GPU core. This should permit heavier AI workloads, including photo and video upscaling and running larger diffusion models. Some key things haven't changed on the new Mac minis, most notably all the ports besides the upgraded Ethernet. Both minis maintain two front-facing 10 GbPS USBC ports, and three on the rear of the M6 are faster Thunderbolt 4s, while the rear three of the M5 Pro are even speedier Thunderbolt 5 ports. But both the M6 and M5 Pro Mac Minis will ship with MacOS 27 Golden Gate preloaded out of the box, offering full compatibility with Apple's new Siri AI and Apple Intelligence features. There's no official launch date for the final public release of MacOS 27, but with today's Mac announcements, it should arrive no later than their September 22nd

Starting point is 00:04:54 ship date, end quote. More on the new chips, also from the verge. Apple has a announced a new M6 chip for Macs along with an M5 Ultra that it says is its most powerful chip ever designed for tasks like 3D rendering and running frontier AI models. The M6 is Apple's first two nanometer chip, which Apple says will bring improvements in performance and power efficiency, and it also has a dual 16-core neural engine for on-device AI workloads. The M6 has a 12-core CPU up from 10 in the M5, and it includes two supercores for performance cores and six efficiency cores. Apple's supercore terminology used to refer to performance cores, but it introduced a new type of performance core, starting with the M5 Pro and M5 Max.

Starting point is 00:05:36 The company says the CPU offers the world's fastest single-threaded performance and up to 1.2 times faster multi-threaded performance in comparison to the M5. The chip also has a 12-core GPU up from 10 cores in the M5 and includes a neural accelerator with each core, Apple says. It supports as much as 32 gigabytes of unified memory. Apple says the M5 Ultra uses the company's Ultra Fusion Interconnect technology to connect two dual-dye M5 Max chips together as one quad-dye architecture. The M5 Ultra has an up to 36-core CPU with 12 supercores and 24 performance cores, and an up to 80-core GPU and a 32-core neural engine. It supports up to 512 gigabytes of high bandwidth unified memory, which, as usual with the M-Series,

Starting point is 00:06:22 is an upgradeable after purchase, end quote. Speaking of chips, OpenAI says it's Halapeno chip delivered 1.5x to 1.9x more AI work per watt and 1.7X to 3.6x lower latency than Nvidia chips across GPTOSS. DeepSeek R1 Kimi K2.51T, quoting The Verge once more. First introduced in June, Halapeno is an application-specific integrated circuit, ASIC, made in partnership with Broadcom. It's designed for AI inference, the process of running a trained AI model to complete a task or deploy an agent. To measure Halapeno's performance, OpenAI used inference X, a benchmarking platform that shows how well AI systems handle inference. The test compared Halapeno's performance against the best results recorded at the time, which were with NVIDIA's GB 200 or GB300 superchips.

Starting point is 00:07:20 OpenAI says Halepinio delivered 1.5 to 1.9 times more AI work per watt across GPDOSS 120B. deep-seek R1 and Kimi K2.51T, then the comparison systems while offering 1.7 to 3.6 times lower end-to-end latency across the three models. That means the chip can provide users with faster responses, more responsive agents, and more reliable access as the demand grows, according to the company. Open AI plans to deploy jalapeno in small volumes by the end of this year, but we'll begin to ramp the volume up into 2027, the company added. The company doesn't say, how many chips it plans to deploy next year, however. Even with these performance improvements, OpenAI says they don't expect to replace their

Starting point is 00:08:06 entire chip lineup with Halapeno saying its overall compute strategy includes very good partners like Nvidia. OpenAI will continue developing the second and third generations of the new chip in the meantime, end quote. WhatsApp has upgraded its two-step verification process letting users replace the six-digit pin with a longer alpha-numeric password and added support for multiseption. multiple pass keys, quoting TechCrunch. The meta-owned app is also adding context to calls from unknown numbers. The updates come as WhatsApp and other messaging apps such as Signal and

Starting point is 00:08:45 Telegram look to make account security easier and more accessible for everyday users, especially as fishing and hacking attempts had become more sophisticated over the years. Until now, WhatsApp's two-step verification has used a six-digit pin as an extra layer of protection to prevent someone from taking over an account, even if they're able to obtain the user's one-time passcode. Now, users can choose a longer alpha-numeric password with a special character to make the pin harder to guess. As for pass keys, users can now add more than one pass key to their account, which

Starting point is 00:09:18 WhatsApp says will be useful if a person uses both iOS and Android. Pass keys, which WhatsApp launched support for in 2024, are far stronger than passwords and allow users to log in to their accounts using face ID or their fingerprint. Pass keys remove the need to rely on username and password combinations, which are ordinarily susceptible to phishing attacks. Pass key logins make it harder for bad actors to remotely access your accounts, as an attacker would need physical access to a person's device that stores the user's part of the pass key.

Starting point is 00:09:49 Additionally, Android users will now see more information when they're getting a call from someone who isn't in their contacts, like whether the number is from a different country and if they have any groups in common with the caller. The latest features join the growing list of updates. WhatsApp has rolled out in recent months. The company launched usernames in late June to allow people to share their profiles without disclosing their phone number. In late May, Meta launched a subscription plan for the messaging app that unlocks extra features like profile customization, super reactions, story insights, and more. The WhatsApp Plus plan launched alongside similar Plus offerings for Instagram and Facebook, end quote.

Starting point is 00:10:26 With this summer's record-breaking heat, it's hard to cool down enough to get a good night's sleep. That's where the pod comes in. The pod by eight-sleep is a smart mattress cover that goes over your existing mattress and actively heats or cools each side of your bed independently. It connects to your wearable, analyzing your daily activity. Then it predicts how you'll sleep and adjusts the pod's temperature automatically. What I love about the eight-sleep is the simplest of things, the ability to be cool when it's time to go to sleep and warm.

Starting point is 00:11:01 when it's time to wake up, and my side is entirely under my control. My wife can be whatever temperature she wants on her side of the bed, too. Use code ride home at 8Sleep.com slash ride home for up to $350 off. That's code ride home at 8Sleep.com slash ride home. Is your multi-entity management creating more confusion than clarity you need the Intuit ERP, Intuit Enterprise Suite? It's the AI native ERP solution that's powerful, painless, and proven. Learn more at Intuit.com slash ERP. Perplexity has launched portable computer, a local AI agent platform running fully on device with zero token costs, starting with Nvidia DGX Spark and RTX Linux PCs, quoting Venture Beat. The launch developed in close partnership with Nvidia is one of the most aggressive attempts

Starting point is 00:12:02 yet to move serious AI agent workloads off the cloud and onto local devices. The model, the user's files, and the work itself can all stay on the machine. Work completed locally consumes no billing credits, and the company says every task starts on the device by default with the system asking permission before sending any individual step to a more powerful frontier model in the cloud. We've basically brought the exact same UI to a fully local app, said Nate, perplexity's vice president of engineering for infrastructure and enterprise during a press briefing Monday. This incorporates the entirety of the agent harness and inference and everything needed to do work locally.

Starting point is 00:12:41 Perplexity Computer, the company's agentic platform for knowledge work, orchestrates AI models, files, tools, and web access to complete multi-step tasks, reviewing folders of documents, analyzing data, producing reports, and pushing results into business systems. Portable computer replicates that experience locally. the local models agent harness inference engine tools app connectors and a security sandbox come packaged together in a single system that bundling is the point with most local AI stacks today users must assemble and operate those pieces separately downloading model weights standing up in inference server wiring together tools and tuning performance historically it's just been really painful to bring up

Starting point is 00:13:24 the local AI stack Nate said with portable computer we really focused on on just making this a really straightforward experience where you can get up and running very quickly. In one demo Monday, the system played the role of a retail investor reviewing a folder of 1099s and investment documents that kind of sensitive financial material many users would hesitate to upload to a cloud service, running a 27 billion parameter Quinn model at full GPU utilization on a DGX Spark. The agent reviewed each document and flagged cases where the hypothetical investor was paying unnecessary fees.

Starting point is 00:14:00 interface element that normally displays a running tally of cloud credits is just parked at zero, Nate noted, because all of this is happening on the device. A second demo showed the hybrid side of the product playing a startup founder, Nate asked the agent to analyze a CSV of user funnel data locally, then push the finished analysis to a Slack channel using Perplexity's connector ecosystem, proof that Local First does not mean disconnected. The system also connects to Google Drive, Gmail and GitHub, and can escalate to a frontier cloud model when the local model hits its limits. At launch, users can set up Quen 3.827B or PPLX 27B, a version perplexity has post-trained on its own harness with NVIDIA's Nemotron 3.5 Lightning coming soon.

Starting point is 00:14:47 Portable computer arrives today for Pro Max Enterprise Pro and Enterprise Max subscribers on Linux with Windows support following in September. any RTX GPU with at least 24 gigabytes of V RAM, roughly a GIFRTX 3090 or newer, clears the bar, a threshold Nate called sort of the floor where we really want to make sure that we can deliver a great experience, but balance that with making it broadly available. The strategic logic behind the launch becomes clear when you consider how AI workloads have changed. Chat was bursty, a question, an answer, and then done. Agents are different. With agents, you want these agents always on if you can't.

Starting point is 00:15:25 You want the agent to really consume as many tokens as they can, Nader said. What we're seeing is an insatiable demand for tokens, and that's something that makes local AI so great. As you saw through all these demos, you were not metered by the token. You were not paying for the token, so it's really killer for agents. This reframes the value proposition of local hardware. An agent that runs for hours reviewing documents, verifying its own work, and iterating on analyses, would rack up substantial API bills in the cloud. On a device the user already owns, the marginal cost of those tokens approaches zero.

Starting point is 00:16:00 Perplexity's paper makes the enterprise version of this argument explicitly. As agents scale across individual workflows and entire organizations, token expenditure and data movement become increasingly difficult to govern. Local first execution addresses both at once. Spend because inference is free and privacy because sensitive tokens never leave the device boundary. Perhaps the most commercially interesting result concerns the hybrid, middle ground. On terminal bench 2.1A challenging coding benchmark, the fully local Kwan model, scored 59.6% at essentially zero marginal cost. Letting it escalate to a Claude Opus 5

Starting point is 00:16:38 advisor in the cloud raised the score to 73% at an estimated 41.5 cents per task. Running the frontier model alone scored 82.4% at 65 cents per task. Escalation, in other words, recovered roughly three-fifths of the gap to frontier performance at about two-thirds of the cost, and the user decides when that trade is worth making. Before any advisor call, the harness runs a classifier over the outgoing context and shows the user exactly what would leave the device. The remote model returns text guidance only it never touches local files or tools, end quote. Finally today, Uber has launched an optional safety feature allowing parents or guardians to watch a live stream of their teen's rides via the driver's front-facing phone camera, quoting Bloomberg.

Starting point is 00:17:32 The feature uses the selfie camera on the driver's phone and notifies an authorized guardian if a live stream is available for viewing within the Uber app. The rideshare company said in a statement on Tuesday, both the teen and the driver will be alerted when the parent is tuning in. The safety measure will roll out nationwide over the coming weeks and is currently being tested in nine markets across the U.S. Uber is announcing the live stream feature during back-to-school season when parents may require car services to pick up or drop off their teens at school or extracurricular activities. The teen ride product, which allows children ages 13 to 17 to be matched with highly rated experienced drivers, has proven popular, with tens of millions of trips completed since its launch in 2023.

Starting point is 00:18:12 It is available in more than 50 countries worldwide. Live streaming will only be available if drivers choose to have it enabled four rides. And Uber spokesperson said, live video streaming runs. in the background to adapt to the driver's device. The representative added, the app monitor's battery, device temperature, and performance, and will pause the stream automatically if needed, including if the driver's connection is unstable, end quote. Nothing more for you today. Talk to you tomorrow.

Tech Brew Ride Home - New Macs

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.