Best of LinkedIn: Next-Gen Vehicle Intelligence CW 13/ 14

00:00:00: Traditionally, you know when you build a car.

00:00:02: You start with the chassis.

00:00:03: it is It's physical its steel and it's entirely measurable

00:00:08: right?

00:00:08: You bolted engine in you attach to the

00:00:09: wheel.

00:00:09: Yeah exactly.

00:00:11: And that structural foundation?

00:00:13: it just never fundamentally changes.

00:00:15: The vehicle rolls off the line and mechanically speaking That is the best it will ever be.

00:00:20: but I mean, when you step into the world of software-defined vehicles that chassis basically becomes invisible.

00:00:28: We're looking at a foundation that is entirely digital.

00:00:30: it's constantly shifting and right now It is the site of this massive industry wide power struggle.

00:00:37: absolutely

00:00:38: we aren't just changing how The car drives anymore were changing.

00:00:40: who actually owns its nervous system?

00:00:43: Navigating That Power Struggle Is Exactly What Todays Deep Dive Is About.

00:00:47: But really quick before we jump in, brought to you by Thomas Allgaier and Frennus.

00:00:52: This edition highlights key LinkedIn posts on Next Gen Vehicle Intelligence In Weeks Thirteen & Fourteen.

00:00:57: Frenness supports Tier One automotive suppliers with early stage market validation for their R&D efforts By combining secondary research direct OEM expert interviews And facilitated customer meetings.

00:01:08: You can find more info in the description

00:01:10: Awesome.

00:01:10: Well, welcome to the deep dive.

00:01:12: like if you are a mobility professional navigating this transition right?

00:01:16: You already know The landscape is incredibly noisy.

00:01:19: Yeah

00:01:19: There's so much marketing fluff out there

00:01:21: exactly.

00:01:21: So our mission today Is just cut through that noise.

00:01:23: we're looking at the top next-gen vehicle intelligence trends Curated directly from the conversations happening across the industry specifically in calendar weeks thirteen and fourteen.

00:01:34: We are stripping away the high-level PR stuff and looking at what the engineers, the strategists, and executives are actually debating on the ground.

00:01:41: And you know... The first big thing we have to talk about is how the era of the isolated ECU is just over.

00:01:49: I mean, we've spent decades building cars where literally every function.

00:01:53: From the windshield wipers to anti-lock brakes... Right!

00:01:55: They all had their own dedicated little computer box.

00:01:59: but now The New Battle Ground is consolidating all of that.

00:02:03: It's about figuring out who actually gets to own the underlying software platform.

00:02:09: That runs a whole show

00:02:10: Yeah and the biggest shockwave on that front over last couple weeks definitely came from Google.

00:02:15: They officially launched Android Automotive OS for software-defined vehicles, which they're calling AAOS SDV.

00:02:23: Wow!

00:02:24: Okay... Yeah and Hinchim Simuti and Jack Dunkley both pointed out a really critical shift in Google's strategy here because they are no longer just aiming to control your infotainment screen

00:02:32: Right..they want the whole thing

00:02:33: Exactly ..They want the entire stack.

00:02:35: We were talking body functions climate control lighting systems The core compute architecture And even over the air updated infrastructure.

00:02:44: So they're moving from the dashboard directly into the vehicle's actual internal software fabric.

00:02:49: Yeah, and the crazy part is that did it by open sourcing the OS?

00:02:53: Benjamin Herman had a brilliant perspective on this.

00:02:56: he called it The Android Playbook.

00:02:58: Oh I like

00:02:59: that!

00:02:59: Right because if you look back at say twenty fourteen Google just handed every car manufacturer A simple free USB cable to get Android Auto onto their screens.

00:03:09: Yeah and some OEMs resisted, you know they tried to build their own clunky alternative?

00:03:14: Totally.

00:03:15: but Google just waited!

00:03:16: And now here in twenty-twenty six...they aren't offering a cable.

00:03:20: They're offering a completely free open source nervous system for the car which immediately commoditizes what legacy automakers have literally spent billions trying to build independently.

00:03:31: It's basically an existential crisis for the traditional automaker at this point.

00:03:34: I mean, Augustin Friedl brought up The OEMs are getting squeezed from both sides.

00:03:46: So at the bottom layer, you've got the Silicon Giants providing this sheer hardware compute power.

00:03:52: and then on top of that there's a massive cloud hyperscaler managing all their off-board data.

00:03:58: And right in the middle is really thin layers of the stack.

00:04:02: It's an application layer which interacts with the driver

00:04:05: That interaction layer where the assistant lives?

00:04:07: Where the brand actually gets to speak for consumers?

00:04:10: Exactly.

00:04:11: But Fridl warms that OEMs risk losing that specific layer entirely because if you seed the interaction model to Google or, You know another tech giant.

00:04:21: You lose the monetization opportunities.

00:04:23: You loose control over the user experience too

00:04:25: Yeah and most importantly you lose the continuous data loop.

00:04:29: your historic automotive brand just disappears behind someone else's digital interface And you're left acting as a low-margin hardware provider.

00:04:36: yeah but Let me push back a little here.

00:04:40: Are OEMs really just rolling over and accepting this?

00:04:42: I mean, i don't see how legacy OEM's survive this.

00:04:45: honestly Google has infinite capital And basically the best software engineers on the planet

00:04:51: right.

00:04:51: so why shouldn't company like vw or BMW Just take the free operating system save The billions they're currently burning on software r&d and just focus On building phenomenally good hardware?

00:05:03: Handing over the digital keys kind of seems inevitable if they want to stay profitable.

00:05:07: We'll see, taking The Free OS is really tempting in short term but it's fatal long-term and Andrew Bullen shared some crucial insights on how leading players are actively fighting this vendor lockin'.

00:05:19: Ok so How Are They Fighting It?

00:05:20: Well Leading automakers companies like BMW, Mercedes & Volvo aren't rolling over.

00:05:25: They are deploying what's called a layered architecture strategy.

00:05:28: Break

00:05:28: down the mechanics of that.

00:05:29: for me, how does a layered?

00:05:35: So instead of handing the entire vertical stack over to one provider, they isolate the layers.

00:05:42: At the very base that use a neutral highly secure safety certified foundation something like QNX.

00:05:49: Then for heavy lifting bring in NVIDIA or Qualcomm.

00:05:55: But crucially, the OEM builds a proprietary firewall around the very top layer.

00:06:01: So they maintain absolute control over the brand interface?

00:06:04: Exactly!

00:06:05: The brand interface...the customer data and overarching vehicle intelligence.

00:06:10: It's essentially strategic supplier management.

00:06:13: They extract best pieces of technology from giants without actually surrendering their long-term leverage.

00:06:19: That makes a lot of sense.

00:06:20: And look, if you're an engineer or product manager listening to this right now... ...you are standing at very real strategic fork in the road.

00:06:26: You really have to ask yourself Are you designing your component to live seamlessly and open standardized Google Ecosystem?

00:06:32: Or are you preparing your product integrate with these fiercely guarded proprietary OEM layers?

00:06:37: Yeah because those two paths require fundamentally different engineering & compliance

00:06:42: strategies Exactly!

00:06:44: And that's strategic divergence leads directly into next massive challenge we saw trending Because if an OEM successfully defends and builds that bottom software layer, well the immediate problem becomes managing this sheer volume of code running on it.

00:06:58: Right.

00:06:59: Traditional deterministic code where a human programmer writes specific rule for every single scenario is just failing to scale.

00:07:05: It

00:07:05: IS!

00:07:06: And that's exactly why industry is aggressively pivoting to AI.

00:07:10: We are officially moving from software-defined vehicles to AI-defined vehicles or AVs?

00:07:15: Yes, eight EV's.

00:07:17: Dr.

00:07:17: Marina Zebian and Prené Agrawal broke down Qualcomm Snapdragon Digital Chassis recently And it just illustrates this pivot perfectly.

00:07:25: I mean A.I is no longer a localized feature like you know voice assistant that you layer on top of the software at end development cycle.

00:07:32: No

00:07:32: not all!

00:07:33: A.Is becoming the architectural foundation itself.

00:07:35: Exactly

00:07:36: In an AI defined vehicle The car basically operates as network cooperating agents.

00:07:42: Yeah.

00:07:42: You have agents processing the ADS, you know?

00:07:44: The advanced driver assistance systems...you have different agents managing that digital cockpit and others handling telematics

00:07:51: Right!

00:07:51: And they use on-edge language & vision models combined with cloud intelligence to share context.

00:07:58: So..the vehicle stops following rigid pre-programmed rules and begins reasoning adapting an actually learning from entire fleet's data in real time.

00:08:06: It is incredible But relying on AI to write and manage code introduces a really severe hidden danger.

00:08:14: And Robert Faye highlighted the concept that I think is critical for any engineering team to understand right now, which is debugging collapse.

00:08:21: Debugging collapse?

00:08:22: That sounds ominous!

00:08:23: It is.

00:08:24: A lot of organizations are super thrilled about using AI to generate software tests and you know on paper it looks like massive progress.

00:08:31: You get exponentially more tests higher code coverage faster execution time.

00:08:35: Right

00:08:35: everybody loves faster execution.

00:08:37: But Faye points out that AI does not actually remove The Engineering bottleneck.

00:08:41: It just relocates it.

00:08:43: It moves the bottleneck from creation of tests to human comprehension of failures.

00:08:47: The math he provided is a pretty stark reality check.

00:08:51: So imagine your AI generates and runs fifty thousand automated tests overnight.

00:08:57: Forty-eight thousand eight hundred of them passed perfectly.

00:09:00: Okay, but twelve hundred fail.

00:09:02: Yeah.

00:09:02: And here is where the system just totally breaks down because if it takes a human engineer say Just one to two hours to unpack a complex AI generated failure You know trace the logic back through the neural net and actually debug It right.

00:09:14: that is up to twenty four hundred hours Of manual debugging generated from a single overnight test run.

00:09:20: It's insane.

00:09:21: It really is!

00:09:22: it's like cranking up a factory assembly line to ten times its normal speed, but then firing all the quality control inspectors.

00:09:28: at the end I mean you aren't scaling your software quality You are literally just staling your confusion.

00:09:32: That is perfect analogy.

00:09:34: Humans simply do not scale linearly with AI output.

00:09:37: So if your system cannot automatically explain its own failures The AIs just amplifying your uncertainty.

00:09:42: And

00:09:42: when you were engineering two-ton machine moving at highway speeds Uncertainty is unacceptable

00:09:48: Exactly which is why Imran Khan and George Pasca weighed in heavily on the functional safety side of this AI explosion, specifically looking at the ISO-to-six-two sixty two standard.

00:09:59: Yeah, Khan made a vital point that Functional Safety is really the quiet discipline holding us entire transition together.

00:10:04: I mean In complex AI architectures faults are going to inevitably occur.

00:10:09: Right it's unavoidable.

00:10:10: And when they do, the vehicle must transition to a safe state predictably.

00:10:16: You cannot just bolt functional safety on later through testing.

00:10:19: it has to be architected into the bedrock from day one.

00:10:22: Yeah and PASCA highlighted really specific danger in this integration process which is the SEOE or Safety Element out of context.

00:10:30: Oh

00:10:30: yeah!

00:10:30: The SEOC

00:10:32: To put that in perspective Building software out of context, it's kind like buying this state-of-the art pacemaker.

00:10:36: Right?

00:10:37: But then trying to wire into a patient whose heart has completely different voltage requirement.

00:10:41: The pacemaker works perfectly in a vacuum but the actual integration will be fatal.

00:10:46: So in software terms A COC is basically when a tier one supplier develops safety related component based entirely on assumptions.

00:10:54: They don't actually know specific final vehicle architecture It'll inhabit

00:10:59: Exactly.

00:11:00: You might engineer your component, assuming it will live in an ASIL-D environment.

00:11:04: Yeah

00:11:05: And you know...ASIL D is the highest automotive safety integrity level thing.

00:11:09: steering or braking systems that require massive dual core redundancy.

00:11:13: Right so if one processor fails The other instantly takes over.

00:11:17: Exactly.

00:11:17: But then the automaker takes your ASIL D component and integrates into a subsystem That only supports ASIL B. That's much lower safety like cluster display Which just lacks deep redundancy.

00:11:29: So the component assumes a safety net exists that simply isn't there.

00:11:32: Yep, and when those assumptions clash during late-stage integration... ...the architecture breaks.

00:11:37: The engineers panic And under massive deadline pressure they write what Pasca calls glue code just to force the mismatched pieces to communicate.

00:11:44: Wow!

00:11:45: Writing rushed glue code for a Safety Critical AI system sounds like an absolute recipe of disaster.

00:11:50: It is Without rigorous gap analysis early in process….

00:11:54: …the entire vehicle's safety argument basically falls apart.

00:11:58: And I think this friction between elegant software theory and messy physical integration exposes the core vulnerability of the entire industry.

00:12:07: You can design the most sophisticated layered architecture, you can train advanced AI models but if your organization lacks a structural discipline to actually build it then strategy is completely worthless.

00:12:18: Oh totally!

00:12:19: We saw pretty brutal contrast in execution capability just over two weeks alone.

00:12:24: Yeah Perzanth Garvahalla wrote about the cancellation

00:12:28: which

00:12:29: was the highly anticipated joint venture between Sony and Honda Mobility.

00:12:33: Right, there's supposed to be a flagship software-defined vehicle packed with all these ambitious infotainment in gaming features...

00:12:40: But they shut it down!

00:12:41: And that lesson is that SDV success requires robust scalable underlying architecture.

00:12:48: A wish list of high tech feature just cannot compensate for weak structural foundation

00:12:53: Absolutely.

00:12:53: Yeah, and then you contrast that O'Fila cancellation with the update shared by Brian Carlson in Watson-Benset.

00:13:00: because Volkswagen and Rivian just successfully completed rigorous winter testing of their new Zonal architecture

00:13:06: which is huge.

00:13:07: it proves It works across both mass market and premium vehicle lines.

00:13:11: Yeah, and just for some context here.

00:13:13: Traditional vehicles use a domain architecture meaning they have over one hundred different computing boxes scattered throughout the car connected by miles of heavy copper

00:13:21: wiring.

00:13:23: But zonal architecture rips all of that out.

00:13:25: It groups the compute power by physical location, or zones and routes everything back to a few central supercomputers.

00:13:33: so it saves weight cuts costs and it centralizes software.

00:13:37: but actually rewiring vehicles entire physical layout is just incredibly difficult to execute

00:13:42: Exactly.

00:13:43: And because VW and Rivian achieved technical proof point in harsh conditions for winter testing.

00:13:48: it unlocked an additional one billion dollar investment from VW into their joint venture.

00:13:54: Wow,

00:13:54: a billion dollars!

00:13:56: Yeah... In the mobility sector today PowerPoint presentations just no longer secure capital.

00:14:01: Discipline execution and tangible technical milestones are essentially the only currency that matters.

00:14:07: So why is it so notoriously difficult for legacy automakers to actually execute on this level?

00:14:13: I mean Marcus Rettstedt in Venetius.

00:14:15: Tadeuzein argued that failure-to-scale STVs not caused by talent gap at all.

00:14:21: No Not At All.

00:14:22: Automakers have hired thousands of brilliant software engineers.

00:14:25: Right,

00:14:25: the failure is actually caused by organizational politics.

00:14:29: Rettstat describes it as a

00:14:32: mechanical hangover.

00:14:33: I love that because traditional auto teams are basically trapped by Conway's law.

00:14:37: conways

00:14:38: law That's the one that states that an organization will design software systems.

00:14:41: They mirror its own internal communication structures.

00:14:44: Exactly

00:14:44: legacy OEMs have spent a century operating in these really rigid physical silos.

00:14:49: The powertrain team sits on one building, the infotainment team sits at another and they just rarely speak.

00:14:56: If you're a software engineer at a tier one supplier, You know exactly what this feels like.

00:15:24: made a profound observation here architecture is management.

00:15:28: wow

00:15:29: yeah.

00:15:29: an automotive company only truly becomes software defined when its technical architecture finally overrides it's internal politics.

00:15:36: and this organizational paralysis is precisely why everyone in the west is terrified of china speed.

00:15:41: oh absolutely.

00:15:42: martin schleicher analyzed whether the european automotive industry can actually move at the velocity of chinese e v manufacturers.

00:15:52: He says European OEMs absolutely can move at China's deed, but currently they only do it when they enter what he calls task force mode.

00:16:00: Task Force mode.

00:16:01: So that's what happens when a massive crisis hits, right?

00:16:03: Like a critical recall or a bug That's going to delay a major launch

00:16:07: exactly.

00:16:07: suddenly the organizational chart just evaporates The silos vanish the OEM and the suppliers put all their engineers in one single room executives make localized decisions In real time without needing committee approval And they achieve this incredible velocity.

00:16:22: But the catch is the moment the crisis is resolved.

00:16:25: if the company just reverts Right back to its standard operating procedure which Is Massive bureaucracy optimized entirely for risk avoidance rather than value creation.

00:16:37: It's basically like cramming a college final the night before.

00:16:41: Sure, you might pass this test but it is an exhausting unsustainable way to run multi-billion dollar business.

00:16:47: You just cannot base your long term software strategy on permanent adrenaline.

00:16:53: Time is rapidly running out to fix that operating model.

00:16:56: Pedro Pacheco outlined the immediate stakes here, pointing to upcoming launches of BMW iX-III and Mercedes CLA in Chinese market.

00:17:05: Right!

00:17:05: These specific vehicles are designed to be Europe's definitive technological response...to China's overwhelming EV & SDV dominance.

00:17:14: Exactly!!

00:17:15: And Pacheko warns that these models simply must win in China.

00:17:18: I mean historically a German automaker could command significant price premium strictly based on their heritage badge.

00:17:25: Yeah But Chinese consumers today have a vast array of domestic options that excel in software, AI integration and digital safety.

00:17:34: And they do it all at lower price point

00:17:35: Exactly.

00:17:36: The badge alone just no longer justifies the cost.

00:17:39: So if the IX-III and the CLA fail to capture market share It sends really chilling signal That traditional OEMs might lack structural breakthrough required To fend off Chinese competitors on global scale.

00:17:50: Yeah!

00:17:51: The war for this software defined vehicle is literally being fought...and won in China right now.

00:17:55: Absolutely, and you know when you synthesize all of these massive shifts the battle for platform ownership The transition to AI native intelligence And a desperate need to overcome These internal silos To achieve execution speed

00:18:09: it

00:18:09: points toward A fundamental transformation In the life cycle Of vehicle itself

00:18:13: Which actually brings us to a final really provocative trend that is Quietly reshaping a massive sector of the industry.

00:18:21: Yeah, and that is the aftermarket.

00:18:23: Oh yeah Poblotelleria Basadone dropped some staggering insights regarding this.

00:18:27: I mean The automotive Aftermarket Is A five Hundred and Twelve Billion Dollar Global Industry And Right Now It Relies Entirely on Things Breaking Down & Humans Physically Fixing Them.

00:18:37: But AI Defined Vehicles Are Radically Altering That Dynamic.

00:18:40: Just Consider The Shear Data Volume.

00:18:43: A Modern Connected Vehicle Is now Generating Up To Twenty-Five Gigabytes Of Driving and System Data every single

00:18:49: hour.

00:18:50: That is

00:18:50: massive!

00:18:51: It is, and Bastadon points out that if OEM successfully leveraged the continuous data stream for predictive maintenance we could see a fifty percent reduction in unexpected vehicle breakdowns.

00:19:01: Furthermore utilizing this connected data can allow regulators to detect seventy-percent of U.S.

00:19:06: vehicles recall far earlier than they do today which will prevent massive physical service campaigns...that's incredible.

00:19:13: So If you're building the future Here's the final thought for you to consider.

00:19:19: As these AI-defined vehicles become capable of self diagnosing their own systemic issues in real time and deploying What actually happens to the physical infrastructure we rely on?

00:19:35: Right.

00:19:35: Will the traditional dealership service center even exist in its current form five years from now?

00:19:41: or will your vehicle simply fix it's own nervous system while It sits parked and you're driveway overnight,

00:19:47: it's wild think about.

00:19:48: obviously The mechanical chassis will always require Physical maintenance.

00:19:52: You know tires wear out.

00:19:53: brake pads degrade sure but the digital foundation that invisible constantly shifting architecture we started this discussion with.

00:20:00: This is going to require a completely different ecosystem of care, and the companies that figure out how to service that digital chassis will ultimately own the next

00:20:13: era.

00:20:20: Thank you so much for joining us on this deep dive.

00:20:23: Keep building, keep questioning the underlying assumptions in your own architecture and make sure to subscribe.