712: Mildly Technologically Traumatized
Episode
137 min
Read time
3 min
Topics
Productivity, Health & Wellness, Relationships
AI-Generated Summary
Key Takeaways
- ✓Waymo Adoption Curve: First-time Waymo riders typically experience disorientation for only a few minutes before treating the service as a standard taxi. Pricing sits within the same range as Uber and Lyft. The service is geographically limited but performs reliably within its operational zones, including executing moderate-traffic left turns without excessive hesitation. Riders who trust statistical safety data over emotional bias will find the transition straightforward after just one or two rides.
- ✓AI-Assisted Reverse Engineering: When attempting to automate BMW app interactions, Claude identified that the app uses Approov, an attestation service that issues short-lived tokens to verify app authenticity, making man-in-the-middle attacks impractical. Claude then independently suggested bypassing browser-based VNC entirely and communicating directly with the GLINet CometQ's underlying PyKVM API, a more efficient solution than the human operator had conceived, demonstrating AI's capacity to reframe the problem rather than just execute instructions.
- ✓GLINet CometQ as Remote KVM: The CometQ, currently $119 on Amazon, runs a native Tailscale client, making it a self-contained network node accessible from anywhere without requiring screen sharing to be enabled on the host machine. It supports DisplayPort-capable iPhones and can be used to lock and unlock devices programmatically. For anyone managing Mac minis in remote locations, closets, or data centers, one unit per location eliminates the need for screen sharing as a recovery mechanism when software access fails.
- ✓Apple Silicon Cache Scaling: Die-shot analysis of the M6 and A20 Pro confirms the E-core L2 cache has doubled from 4MB on the A18 Pro to 8MB on the A20 Pro, progressing through 6MB on the A19 Pro. The shared S/P-core L2 cache has remained at 16MB across all three generations. Cache expansion on chip costs real estate and power budget, so Apple's aggressive doubling signals that E-core cache was a confirmed bottleneck in real-world workloads, not a speculative optimization.
- ✓M6 Memory Bandwidth Tier Split: The 16GB M6 configuration uses the same memory bandwidth as the M5 at 154.6 GB/s. Only the 24GB and 32GB configurations use faster 10,667 MHz DRAM, reaching 170.7 GB/s. For users running local AI models or other bandwidth-sensitive workloads, this makes the RAM configuration choice a performance decision, not just a capacity decision. Buyers who prioritize throughput over cost should select at least the 24GB tier regardless of actual storage needs.
What It Covers
ATP episode 712 covers first-hand Waymo autonomous vehicle experience, Casey's BMW home automation project using Claude AI and a GLINet CometQ KVM device, Verizon dropping cable card support forcing TiVo abandonment, M6 chip architecture analysis from die shots, rumored OLED MacBook Pro weight reduction concerns, and Apple's new smart home push via an LG hardware partnership.
Key Questions Answered
- •Waymo Adoption Curve: First-time Waymo riders typically experience disorientation for only a few minutes before treating the service as a standard taxi. Pricing sits within the same range as Uber and Lyft. The service is geographically limited but performs reliably within its operational zones, including executing moderate-traffic left turns without excessive hesitation. Riders who trust statistical safety data over emotional bias will find the transition straightforward after just one or two rides.
- •AI-Assisted Reverse Engineering: When attempting to automate BMW app interactions, Claude identified that the app uses Approov, an attestation service that issues short-lived tokens to verify app authenticity, making man-in-the-middle attacks impractical. Claude then independently suggested bypassing browser-based VNC entirely and communicating directly with the GLINet CometQ's underlying PyKVM API, a more efficient solution than the human operator had conceived, demonstrating AI's capacity to reframe the problem rather than just execute instructions.
- •GLINet CometQ as Remote KVM: The CometQ, currently $119 on Amazon, runs a native Tailscale client, making it a self-contained network node accessible from anywhere without requiring screen sharing to be enabled on the host machine. It supports DisplayPort-capable iPhones and can be used to lock and unlock devices programmatically. For anyone managing Mac minis in remote locations, closets, or data centers, one unit per location eliminates the need for screen sharing as a recovery mechanism when software access fails.
- •Apple Silicon Cache Scaling: Die-shot analysis of the M6 and A20 Pro confirms the E-core L2 cache has doubled from 4MB on the A18 Pro to 8MB on the A20 Pro, progressing through 6MB on the A19 Pro. The shared S/P-core L2 cache has remained at 16MB across all three generations. Cache expansion on chip costs real estate and power budget, so Apple's aggressive doubling signals that E-core cache was a confirmed bottleneck in real-world workloads, not a speculative optimization.
- •M6 Memory Bandwidth Tier Split: The 16GB M6 configuration uses the same memory bandwidth as the M5 at 154.6 GB/s. Only the 24GB and 32GB configurations use faster 10,667 MHz DRAM, reaching 170.7 GB/s. For users running local AI models or other bandwidth-sensitive workloads, this makes the RAM configuration choice a performance decision, not just a capacity decision. Buyers who prioritize throughput over cost should select at least the 24GB tier regardless of actual storage needs.
- •Verizon Cable Card Discontinuation Path: Verizon ended cable card support after the legal mandate requiring it lapsed. Customers who call to cancel TV service are told to keep the cable cards rather than return them. Dropping TV service from a legacy triple-play bundle can reduce monthly bills substantially — one reported case dropped from $250 to $100 per month. Verizon is also offering 5-gigabit Fios for $10 more per month, though adoption requires a technician visit and full home switch infrastructure upgrades.
- •OLED MacBook Pro Weight Reduction Risk: Gurman's October 4 report describes the upcoming OLED touchscreen MacBook Pro as significantly lighter through internal component miniaturization. The primary concern is that meaningful weight reduction almost always requires reducing battery capacity, since the battery and metal chassis account for most laptop mass. The M5 Pro and Max chips in higher-end configurations are not on the 2nm process, so they receive no efficiency dividend to offset potential battery cuts, making the base M6 model the only variant with a credible path to lighter weight without battery life regression.
Notable Moment
Casey built a working BMW home automation system by combining a GLINet CometQ KVM attached to an iPhone, with Claude writing code that directly queries the device's PyKVM API. The system automatically wakes the phone, reads car status, pushes data to Home Assistant via MQTT, then relocks the phone to prevent OLED burn-in — all without any human-written code.
Episode Transcript
Have either of you ridden in a driverless car yet? No. Nope. I had the opportunity to ride my first Waymo. Oh, interesting. How was that? It was really... So it's exactly what everyone says. It was really weird for like the first few minutes, and then I got bored and looked at my phone, and it was just a taxi. And then on the way home, I'm like, let's book another one. And I booked another one in homes. I'm like, well, as long as Waymo's available, I know, like, mathematically, it's probably safer for me to take a Waymo than a a human driven car. And so I literally immediately was converted. So one thing to know, like, was staying at a hotel, and the Waymo would not come into the hotel's little like loop driveway thing. Oh, interesting. It would only pick me up like outside, like across the street. So I... And I'm sure there's probably some reason maybe maybe Waymo doesn't go into private property. I I don't know. So I was thinking, I'm like, even even, like, where it should pull over to drop you off or pick you up, that is itself a very difficult problem. And I was thinking like, wow. Everything about this is a very difficult problem. Like, you you know, you get past the entire self driving thing, which is not a small thing, and then you're like, oh, but where do you park? Like, even that is a difficult problem. But it was really nice. Like, it pulled up and, you know, it it... The the app is fine. Like, it's it's a modern decently seemingly well made app. Price wise, I I compared, like, you know, comparing the Waymo to an Uber or a Lyft at the same time. You know, the prices jump around a lot, but they were all within the same ballpark of each other. So it wasn't it wasn't significantly cheaper or significantly more expensive, and the Waymos are everywhere, like, there there are tons of them, and you just don't re- like, stop noticing after a while. Even like like, you know, like we had we had rented our own car, and so we were also like driving that around, but like, when you're a driver on the road surrounded by Waymo's, it was fine, and being a passenger in a Waymo was fine. It was unbelievable, like, because I know, I have said on this show years ago that I think it's gonna be a very long time before we have fully self driving cars on all roads, way, you know, the way that we think. Waymo is not that. It is not on all roads, all... Like, everywhere. You can't you can't have Waymo pick you up anywhere, and it can't drive you anywhere. But in the in the places it operates, it seems remarkably good. You know, and That's super cool. And there was even like, was worried like, …
Get the full transcript (29,536 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 134-minute episode.
Get Accidental Tech Podcast summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
More from Accidental Tech Podcast
711: Hot Dog on an Actuator
Oct 1 · 119 min
The Vergecast
The real story of the iPhone 18 Pro's camera
Sep 18
More from Accidental Tech Podcast
710: That New Mouse Pad Feeling
Sep 24 · 133 min
We Study Billionaires
TIP844: Uber (UBER): The Autonomy Referendum — Is Mr. Market Completely Wrong? w/ Daniel Mahncke & Shawn O’Malley
Sep 6
Books, tools, and gear mentioned in this episode
SignalCast may earn commission on purchases via these links. As an Amazon Associate, SignalCast earns from qualifying purchases.
Tools
- ClaudeRecommended
by Anthropic
“Casey's BMW home automation project using Claude AI and a GLINet CometQ KVM device. When attempting to automate BMW app interactions, Claude identified that the app uses Approov, an attestation service that issues short-lived tokens to verify app authenticity, making man-in-the-middle attacks impractical.”
- TailscaleRecommended
by Tailscale
“The CometQ, currently $119 on Amazon, runs a native Tailscale client, making it a self-contained network node accessible from anywhere without requiring screen sharing to be enabled on the host machine.”
- Home AssistantRecommended
by Home Assistant
“The system automatically wakes the phone, reads car status, pushes data to Home Assistant via MQTT, then relocks the phone to prevent OLED burn-in — all without any human-written code.”
Gear
- GLINet CometQ KVMRecommended
by GL.iNet
“Casey's BMW home automation project using Claude AI and a GLINet CometQ KVM device. The CometQ, currently $119 on Amazon, runs a native Tailscale client, making it a self-contained network node accessible from anywhere without requiring screen sharing to be enabled on the host machine.”
More from Accidental Tech Podcast
We summarize every new episode. Want them in your inbox?
Similar Episodes
Related episodes from other podcasts
The Vergecast
Sep 18
The real story of the iPhone 18 Pro's camera
We Study Billionaires
Sep 6
TIP844: Uber (UBER): The Autonomy Referendum — Is Mr. Market Completely Wrong? w/ Daniel Mahncke & Shawn O’Malley
The Vergecast
Aug 25
Does smart home tech work in a college dorm?
No Priors: Artificial Intelligence | Technology | Startups
Jul 31
Building an Autonomous Enterprise for Real-World Services with Netic Founder Melisa Tokmak
Cognitive Revolution
Jul 27
Nathan Goes to China – Part 1: Tech & Agent Setup, Chinese AI UX, WAIC, and Attitudes on AI
Explore Related Topics
This podcast is featured in Best Tech Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's Health & Longevity Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into Accidental Tech Podcast.
Every Monday, we deliver AI summaries of the latest episodes from Accidental Tech Podcast and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime