Best of LinkedIn: Next-Gen Vehicle Intelligence CW 03/ 04

00:00:00: Brought to you by Thomas Allgaier and Frennus, this edition highlights key LinkedIn posts on NextGen vehicle intelligence in weeks three and four.

00:00:08: Frennus supports automotive enterprises and consultancies with market and competitive intelligence decoding disruptive technologies, customer needs, regulatory change, and competitive moves.

00:00:19: So product teams and strategy leaders don't just react, but shape the future of mobility.

00:00:24: Welcome back to the deep dive.

00:00:26: We're looking at weeks three and four of twenty twenty six and you can really feel the mood has shifted.

00:00:32: It really has.

00:00:33: I was going through the sources you sent over and it feels like the whole honeymoon phase with the software defined vehicle is well it's over.

00:00:39: It's definitely over.

00:00:40: We're not looking at Chinese concepts anymore.

00:00:42: No.

00:00:43: We've sort of entered what I'd call the hangover phase.

00:00:45: Yeah.

00:00:45: The industry is waking up looking at the bill and asking okay how are we actually going to execute on this.

00:00:50: And

00:00:50: pay for

00:00:50: it.

00:00:51: And pay for it.

00:00:51: Yeah.

00:00:52: The big theme for these two weeks is really this clash between the dream of a new software architecture and, you know, the cold hard reality of organizational drag.

00:01:02: That's a great way to put it.

00:01:03: The tech is ready to sprint, but the companies are like trying to run a marathon in work boots.

00:01:10: We've got a lot to cover today.

00:01:11: This whole organizational mess, the technical nitty-gritty of diagnostics, which is way more interesting than it sounds, I profit.

00:01:20: It really

00:01:20: is.

00:01:21: And finally, this pressure to provide evidence that any of this AI stuff actually works.

00:01:26: Let's start right there with that first point.

00:01:28: the organizational drag.

00:01:30: Yeah, this came up over and over.

00:01:32: We all talk about the software-defined vehicle, the SDV.

00:01:35: Right.

00:01:36: But no one's talking about the hardware-defined organization that's trying to build it.

00:01:40: That's

00:01:40: the core of the problem.

00:01:41: Andrea Leitner really nailed this in her analysis.

00:01:45: She brought up Conway's law.

00:01:46: Are you familiar with it?

00:01:47: Vaguely,

00:01:48: something about communication structures, right?

00:01:50: Exactly.

00:01:51: Conway's law basically says that any system you design will end up mirroring the communication structure of your organization.

00:01:58: Okay, so if your org chart is a mess, your product is going to be a mess.

00:02:02: Specifically, if your company is siloed, your product will be fragmented.

00:02:07: Lightner's point is that most OEMs think the SDV challenge is a technical one.

00:02:11: Just hire more coders.

00:02:13: Right.

00:02:13: Let's hire five hundred Picon developers and we'll win.

00:02:16: Right.

00:02:16: But if you have a powertrain department that doesn't talk to the chassis department, who doesn't talk to infotainment, You can't build a centralized vehicle brain.

00:02:26: You just end up stitching things together.

00:02:28: You end up with three separate computers duct taped together.

00:02:30: Yeah.

00:02:31: And there was this perfect, almost painful story shared by Lucas Tim.

00:02:35: that just illustrates this.

00:02:36: Oh,

00:02:36: the BMW bug story.

00:02:38: I saw that.

00:02:39: It went completely viral.

00:02:40: For a reason.

00:02:41: It's painfully real.

00:02:42: So for anyone who missed it, an SDV expert finds a software bug in his own new BMW.

00:02:48: Right.

00:02:48: He posts about it.

00:02:49: The post blows up.

00:02:51: BMW social media reaches out, sales reaches out, the developers even reach out.

00:02:55: Everyone is super helpful.

00:02:56: They find the bug, they write the fix.

00:02:58: Yeah.

00:02:59: Success.

00:02:59: You would think so.

00:03:00: But here's the punchline.

00:03:01: They couldn't actually push the fix to his

00:03:03: car.

00:03:03: Even though the car is fully capable of over-the-air updates.

00:03:07: Yes.

00:03:08: The car had to physically go to a dealership for a software patch that was already finished and sitting on a server somewhere.

00:03:15: Luke has called it an over-the-auto house update.

00:03:17: Over-the-auto

00:03:18: house.

00:03:19: It's funny, but it's also.

00:03:21: It's devastating.

00:03:22: It's embarrassing.

00:03:22: This isn't a talent problem.

00:03:24: The engineers fixed

00:03:25: it.

00:03:25: It's a pipeline problem, a bureaucracy problem.

00:03:28: He mentioned something like, what was it?

00:03:30: Fourteen sign-offs.

00:03:31: Fourteen.

00:03:32: Fourteen different approvals needed between code is ready and code is deployed.

00:03:38: That's insane.

00:03:39: And this connects perfectly to what Anton Semyon was saying.

00:03:42: The hardest part isn't writing the code.

00:03:45: It's stopping the organization from just rebuilding its old, slow, hardware-based delivery systems around the new software.

00:03:52: It's like trying to run Netflix, but you need a committee meeting every time you want to upload a new show.

00:03:57: That's a perfect analogy.

00:03:58: And it's why this concept of rightsizing from Kristoff Hursig is getting so much attention.

00:04:03: His argument is we're overbuilding these systems.

00:04:05: Making

00:04:06: them too complex.

00:04:07: Yeah, we think complexity means sophistication, but in SDV complexity is just liability.

00:04:13: Rate sizing means simplifying the architecture to give control back to the OEM, but also allowing suppliers to integrate faster because the boundaries are clear.

00:04:21: Okay, so let's say just for a minute we can fix the org chart.

00:04:24: A huge assumption.

00:04:26: A

00:04:26: huge one.

00:04:27: We still have to fix the actual plumbing inside the car.

00:04:30: You mentioned diagnostics?

00:04:31: Yes, diagnostics.

00:04:32: It sounds dry, but this is really where the rubber meets the road for maintenance, for updates, for everything.

00:04:39: So when I hear diagnostics, I'm thinking of a mechanic plugging that greasy laptop into the port under the steering wheel.

00:04:45: That is exactly the world we're trying to leave, that's UDS Unified Diagnostic Services.

00:04:50: It's a query response system.

00:04:52: The tool has to know exactly what to ask the car.

00:04:55: Like playing twenty questions with your engine.

00:04:57: Exactly.

00:04:58: But Marcus Rettstat points out this huge problem.

00:05:00: We're putting these super powerful high performance computers or HPCs in cars.

00:05:06: But they're still speaking that old twenty questions language.

00:05:09: So we have a super computer that communicates like a pager from nineteen ninety

00:05:13: nine.

00:05:13: Pretty much.

00:05:15: It's diagnostics over IP.

00:05:17: But it's still the old way of thinking.

00:05:19: What he's advocating for is SOVD service-oriented vehicle diagnostics.

00:05:24: And how is that different?

00:05:26: Think of it like a modern health app versus a doctor's visit.

00:05:29: With UDS, the tool has to ask you questions.

00:05:33: With SOVD, the car is like a patient wearing a smartwatch that's constantly broadcasting its vital signs via an API.

00:05:41: It's self-describing.

00:05:42: So you don't need a special tool that knows all the questions?

00:05:45: No.

00:05:45: Any authorized service can just subscribe to the feed and see the car's health.

00:05:49: This means you can onboard new components way faster.

00:05:51: They just show up on the network and say, hi, I'm a battery.

00:05:54: Here's my status.

00:05:55: That sounds infinitely better.

00:05:57: So why isn't everyone doing

00:05:58: it?

00:05:58: It's the dual world challenge.

00:06:00: You can't just flip a switch on a whole car.

00:06:02: For the next decade, you'll have an HPC speaking this new fast language.

00:06:06: But the window motor is still speaking the old one.

00:06:08: Ah, so you need a translator in the middle.

00:06:10: And that's where the complexity and the bugs can really explode.

00:06:13: Speaking of bugs, Dr.

00:06:15: Ralph Mensenberger brought up something that sounds like a hidden project killer,

00:06:19: timing.

00:06:20: This is so critical and so often overlooked.

00:06:23: He points out that these new platforms are a mix of everything.

00:06:27: You've got Autosar Classic, Adaptive Autosar, Linux, all running on the same hardware.

00:06:31: Which sounds like a traffic jam waiting to happen inside the chip.

00:06:34: It is, but here's his hard truth.

00:06:38: These projects rarely fail because the technology is broken.

00:06:43: They fail because the timing insights come too late.

00:06:45: What does

00:06:46: that mean, timing insights?

00:06:47: It means you realize that a critical signal, like the one that fires the airbag, is arriving ten milliseconds too late.

00:06:54: But you only find this out three weeks before the start of production.

00:06:57: Ouch.

00:06:58: That is a stop the line kind of problem.

00:07:00: It's a disaster.

00:07:01: And the data to spot that lag probably existed in a simulation six months earlier, but no one was looking at it in the right way.

00:07:08: It's a management failure.

00:07:09: And Johannes Hoppe added another layer here.

00:07:11: He was talking about zonal architecture.

00:07:13: Right, zonal is the big buzzword.

00:07:15: Organizing the cars wiring by location instead of function.

00:07:19: But he says

00:07:19: that's not enough.

00:07:20: Not even close.

00:07:21: He says zonal alone doesn't make a vehicle software defined.

00:07:26: It actually just moves all the complexity from the wiring harness into the network itself.

00:07:31: Okay.

00:07:31: And if you don't have deterministic communication on that network, you're in huge trouble.

00:07:36: Deterministic.

00:07:37: Break that down.

00:07:38: It just means predictable.

00:07:40: If you hit the brake pedal, that signal has to reach the brake caliper in, say, exactly two milliseconds every single time.

00:07:48: Right.

00:07:49: It can't be two milliseconds today and then five milliseconds tomorrow because the kids are streaming a movie in the back and clogging up the network.

00:07:55: That is a terrifying thought.

00:07:56: Sorry, I crashed.

00:07:57: My car was buffering.

00:07:58: That's exactly it.

00:08:00: You need something like time-sensitive networking, TSN, to guarantee that delivery time.

00:08:05: Without it, the car behaves differently every time you drive it.

00:08:08: Let's

00:08:08: zoom out a bit from the tech to the business side.

00:08:12: The suppliers, the tier ones, they seem to be in a really tough spot right now.

00:08:17: Tough is putting it mildly.

00:08:19: Yeah, Justin Friedl had a great term for it.

00:08:21: The uncomfortable sandwich.

00:08:22: Which does not sound delicious.

00:08:24: No, definitely not.

00:08:26: The tier one giants are being squeezed from both sides.

00:08:29: From the top, you have the OEM saying, we're going to insource the high value software.

00:08:33: Thank you very much.

00:08:34: And

00:08:34: from the bottom?

00:08:35: From the bottom, you have the platform, players, Nvidia, Qualcomm, Google.

00:08:41: providing the base layers directly to the OEMs.

00:08:43: So the tier ones are stuck doing the messy integration in the middle, but with shrinking value pools.

00:08:49: They're carrying all the integration risk, but losing access to the profit.

00:08:53: Arbrikesha actually broke the whole supplier landscape down into three camps.

00:08:57: Okay,

00:08:57: who's in the safe zone?

00:08:58: First, what he calls the resilient integrators.

00:09:01: Think Bosch or Magna.

00:09:03: They have so much scale and deep system knowledge, they can just manage the whole mess for an OEM.

00:09:08: Okay.

00:09:09: Second are the specialists.

00:09:10: CATL for batteries, Hyundai Mobis for chassis tech.

00:09:13: If you own a critical piece of hardware that's really hard to make, you're safe.

00:09:17: And the third group, the sandwich meat.

00:09:20: The cotton the middle.

00:09:21: He mentions names like ZF, Continental, Forvea, massive companies, but they're being squeezed.

00:09:27: They have to decide, and quickly, if they're an integrator or a specialist, because that middle ground is just evaporating.

00:09:33: And while they're trying to figure that out, you've got the tech tiers moving in.

00:09:37: Right.

00:09:38: Matt Damascino highlighted this.

00:09:40: Huawei, Xiaomi, Nvidia, they're not just tech companies anymore.

00:09:44: They're core automotive players.

00:09:46: And they don't have that organizational drag that Conway's law problem.

00:09:50: They were born digital.

00:09:52: Which brings us to the general vibe from CES, twenty twenty six.

00:09:55: Reading the post, it really felt like the industry had sobered up.

00:09:59: It felt like a very serious CES.

00:10:01: Sharif Hussein coined a great phrase for it.

00:10:03: Time to evidence.

00:10:04: Time to evidence.

00:10:05: I like that.

00:10:06: What's the implication?

00:10:07: It means we were done with presentations that say, trust me, this will work.

00:10:11: The new competitive gap is how fast can you prove continuously and economically.

00:10:16: that it works safely.

00:10:17: So it's not about building the feature anymore.

00:10:19: No, it's about validating it.

00:10:21: You can't drive enough physical miles to validate a complex AI.

00:10:25: It's impossible.

00:10:26: You need virtual validation, a digital twin to do it at scale.

00:10:30: And that proof is how you build trust.

00:10:33: Liz Santoni from Cisco shared a great personal story about this.

00:10:37: She was in a Mercedes-Benz CLA with Drive Pilot

00:10:40: Pro.

00:10:40: Right, on the Autobahn, at real speeds.

00:10:42: And she said something really profound.

00:10:45: Trust comes before technology.

00:10:48: The driver was calm not because he was brave, but because he understood the system's reliability.

00:10:53: Trust is the prerequisite, not the outcome.

00:10:56: And that's getting harder as we move toward what H. Manguino calls the ADV, the AI-defined vehicle.

00:11:01: Great.

00:11:02: Just when I got used to SDV.

00:11:03: I

00:11:03: know, it never stops.

00:11:05: But his point is critical.

00:11:06: Software isn't the premium feature anymore.

00:11:08: Intelligence is.

00:11:10: We're moving to cars that continuously self-optimize.

00:11:12: Like what?

00:11:13: Like Michelin's smart wear concept.

00:11:14: The car knows exactly how worn your specific tires are, and it adjusts the braking and suspension in real time to compensate for that exact level of grip.

00:11:22: Wow.

00:11:23: Okay, that's not just running code.

00:11:24: That's thinking.

00:11:25: Correct.

00:11:26: And that creates a mixed criticality problem.

00:11:29: AI is probabilistic.

00:11:30: It makes a best guess.

00:11:32: Braking has to be deterministic.

00:11:33: It must work.

00:11:35: So you need a federated compute architecture to keep those two worlds safely separated, but still talking to each other.

00:11:41: It's a huge architectural challenge.

00:11:43: It is.

00:11:44: which is why we're hearing more about egenic AI.

00:11:46: Amity Chokashou and Maitayals Bizarra, we're both discussing this.

00:11:50: This is moving way beyond just voice commands, isn't it?

00:11:52: Far beyond.

00:11:53: It's not, hey car, turn on the seed heater.

00:11:56: It's the car understanding context.

00:11:58: The example from Qualcomm's Snapdragon chassis is that the agent sees you're stressed.

00:12:02: It knows about the traffic jam ahead, checks your calendar, and then proactively changes the route and the cabin lighting.

00:12:09: It's acting on your behalf.

00:12:10: Looking at the last cluster of posts, it seems pretty clear.

00:12:13: no one company can build all of this alone.

00:12:16: The not invented here syndrome is finally dying out of necessity.

00:12:19: The complexity is just too high.

00:12:21: We saw three really key partnerships announced in these weeks.

00:12:25: Let's run through them quickly.

00:12:26: First up, Tata LC and Autolink.

00:12:28: Yeah, Sachin Padreshet highlighted this.

00:12:30: This is a big structural move to create a proper SDV base layer, the plumbing, so OEMs don't have to start from scratch.

00:12:38: Okay, then ZF and Qualcomm.

00:12:40: Jack Dunkley covered this.

00:12:41: This is about building a scalable, open, AI-driven ADS solution.

00:12:46: The keyword is open, moving away from the supplier, black box.

00:12:50: And finally, Electrobit with Vector and QNX.

00:12:53: This one.

00:12:54: covered by Ramki Krishna and Afiram Thiraga, is all about standardizing the non-differentiating software with something called Alloy Core.

00:13:02: Non-differentiating.

00:13:03: That's the key.

00:13:04: It is.

00:13:05: No customer buys a car because of its awesome bootloader.

00:13:08: So why is every OEM spending millions to reinvent it?

00:13:11: This is about stopping that.

00:13:12: So if I were to sum up these two weeks, it feels like the bureaucracy is fighting the technology, the suppliers are fighting for their place in the value chain, and the AI is demanding we prove it works.

00:13:21: It sounds chaotic.

00:13:23: And it is.

00:13:24: But I actually see it as a sign of maturity.

00:13:26: We're not arguing about what to build anymore.

00:13:28: We're finally arguing about the hard truths of how to actually get it done.

00:13:32: The industry is getting real.

00:13:34: And that means facing some uncomfortable truths.

00:13:37: A great place to wrap.

00:13:39: for this deep dive.

00:13:40: It's going to be a fascinating year to watch.

00:13:42: If you enjoyed this episode, new episodes drop every two weeks.

00:13:45: Also check out our other editions on electrification and battery technology, future mobility and market evolution, and commercial fleet insights.

00:13:54: Thank you for listening.

00:13:55: Thanks for your time and don't forget to subscribe for the next deep dive into the source material.