How Legacy Automakers Can Solve Their Software Problems - Autoline Exclusives
ฝัง
- เผยแพร่เมื่อ 8 ก.พ. 2025
- Legacy automakers do not have the organizational structure to write software quickly and robustly. They run into problems that lead to product delays, higher costs and unhappy customers. Jan Becker is the CEO of Apex.ai. He has common sense advice of how automakers can change their approach and get much better results.
INSTAGRAM: / autolinenetwork
TWITTER: / autoline
FACEBOOK: / autolinenetwork
WEBSITE: www.autoline.tv
Jan Becker is correct, small teams of rock star developers kick butt. I've been a software developer for over 40 years involved in small teams and our successful projects made the customer happy. We once had a client that got fed up with a software project off-shored to India that never produced anything in 3 years, we started the development from scratch and had it done in 3 months with 4 people on the team.
Yah, this was a good episode. You hit the nail on the head more or less... legacy auto doesn't have the ecosystem, doesn't have the expertise in-house, doesn't have the mechanisms or corporate governance required to retain quality programmers, doesn't have the necessary level of loyalty to its employees to keep the ones they do get. They've spent decades cost-optimizing-out all of those high-end employees in favor of mediocrity. So many of the pieces have been farmed out solely to try to isolate profit centers that Humpty-Dumpty can't be pieced back together again.
Legacy auto doesn't even know how to talk to the type of people they need to hire without getting laughed out of the room. They are just too-used to treating their employees as disposable parts.
Remember what happened when Herbert Diess tried to fix VW, at this point several years ago? He knew exactly what had to be done but all the individual fiefdoms inside VW were too involved in trying to hold onto their individuial little kingdoms they couldn't see the writing on the wall. Not listening to Diess has set VW back at least 10 years. At least. American legacy-auto is in an even worse situation.
-Matt
Spot on Matt. You nailed it.
I think you hit the nail on the head. Only Tesla understands the importance of software and computer capabilities in their vehicles, since Elon, bless his tormented heart, came from Silicon Valley.
From his description, Apex is doing traditional software design and development - a good thing. It doesn't matter whether the OEM or the suppliers do any particular software implementation so long as it is written to the standards of the OS for the vehicle. Apps for phones are nothing more than software written to standards. The same goes for computers running Unix, Linux, Windows, Android, z/OS, whatever. There is no requirement that the same type of hardware be used. We were doing distributed peer to peer workload back in the 70's with computers that didn't have a single shared hardware characteristic. What was kind of touched on but not emphasized is culture. Software engineers will not tolerate micromanagement, highly vertical org charts or OG management saying this is the way its always been done. Legacy auto needs to learn that the way they plan, approve and manage is easily 40+ years out of date compared to the software industry. The software industry certainly won't go backwards, so if auto wants to stay competitive, and thus in business, they are the ones that must adapt. For some, it would seem good luck with that.
You don’t have to integrate everything under one roof. You just need a standard CPU, OS, and interfaces and force all the suppliers to use it. Then you integrate everything at the OEM. It can be done but not without a standard.
Top talent is key. VW chucked a big pile of cash at building a large software group and it failed. They should have gotten a small team top talent to make things happen.
It makes sense to keep as much software under one roof. The outsourcing mindset that many have been doing for the past 30 years adds too much complexity.
They hired 10,000 developers as if the amount of workers being 5 figures guaranteed success.
In a country where you can't really fire people just because you don't need them anymore.
🤷♂️
As I used to work for JLR for 10yrs & bmw 20yrs , I think I know. In fact the only thing propping these companies up are the service centres, guess what EVs require very little service 😮
Legacy auto: "How do we cost cut our way to better software?" Answer: You don't, you have to spend money.
"Lets do it on the cheap!"
Legacy automakers have been ripping the public off for years,via servicing and repairs!
Wrong!
@@brunoheggli2888 Legacy Auto is SCREWED Because they have far to long Screwed their customers who now want other automakers and cheaper Chinese BEVS😂
Servicing and repairs are the primary profit centers for all vehicle dealerships. The sale of cars is little more than advertising for the service department.
@@brunoheggli2888 How can you lose your warranty if you refuse to do scheduled maintenance? How is that not a fucking ripoff?
Just stop calling software defined vehicle, is the first step.
It’s not that hard really. My years ago I worked at a legacy company and they are still running their 30 years old system. And I worked at Amazon, and Bezos gave the business heads 3 years to move from their database (mostly Oracle) to AWS! And it is done! It just takes leadership and determination!
I just test drove Xiaomi SU7 Max in China and it blew me away. Chinese EV makers are leading this shift toward software centric automobile dev. Tesla is the only US based car company that is competitive I am afraid. Detroit is so far behind by now and I seriously doubt that they can ever catch up. If one day EV does displace ICE no amount of tariffs can protect legacy car manufacturers. On a positive note, I notice the vehicle I rode runs on Nvidia chip for much of its AI functionality. At least in this area US still has some leverage.
Back to the original question, does vertical integration (in this case having in house SW dev) is the critical path for auto makers to succeed? The SU7 Max is actually built in the factory owned by Beijing Automotive, a traditional car company. Maybe this arrangement provides a path of salvation for legacy car makers.
Even though Tesla is years ahead with in house software it still is not perfect. I have a model Y in Sydney Australia. A great car BUT the wiper auto control is not good and manual use requires a two step fiddle while driving which is annoying and distracting. Enhanced auto pilot which borders on dangerous does not work properly at all in cross city tunnels and still phantom brakes on open roads. It does not disengage when the accelerator is pressed which because of regenerative braking can allow you to forget it is on and it can unexpectedly accelerate when going through a round about or turning a corner. With enhanced auto pilot engaged the distance from another vehicle is also mysteriously inconsistent. When set to 7 car lengths from the car in front it can vary all over the place. There is another problem with a centralised "computer on wheels" system. With one over the air update or an implanted bug an entire fleet of vehicles can be disabled. Tesla could do this tonight so what could a Chinese car company do? Perhaps the Chinese Communist Party could just march in to BYD one day and say " we don't like the U.S.A. anymore turn off all their cars."
Cybertruck software is still missing paid for functions seven months after product release.
I believe you’re pretty on track with your remarks. Thank you. As noted, Tesla vehicles have a powerful central computer, but for good reason does not run all vehicle functions on the one computer. Tesla still has ECUs for running firmware which must have a deterministic response time between sensor input and actuator output. What Tesla does better than most is to provide the ability to fine tune the behavior of its ECUs by over the air updates managed by the central computer. Legacy makers such as GM usually require ECUs to be manually re-flashed by the dealer service network, and that means warranty charge-backs and wasted time for customers.
@@wtmayhew Yes over the air updates, their super chargers and overall commitment to excellence was why bought the Model Y. A fantastic car and I know why the wipers and EAP are a pain. It's because they are relying on their vision system to eventually resolve these issues. I have learned to be wary of EAP and the wipers but the rest of the experience is fantastic. So many benefits rarely talked about like how fast they can brake. Before you can even get your foot to the brake pedal regenerative braking is slowing you down and when you do have to emergency brake the brake lights flash, warning the drivers behind. All cars should have emergency flashing brake lights. Then not how fast from 0 -60 but how very fast they are from 40 to 80 when overtaking. IMO the less time on the wrong side of the road the better. Anyway they have a big lead and I hope they solve it all and as a consequence reduce the deaths and injury on U.S. roads.
@@markumbers5362 Thanks for sharing your perspective, it is much appreciated.
The phrase I did not agree with is “that worked well in the past”. No, it did not. How many cars did you buy that had navigation built in only to find out it was so out of date it was useless and any update you could buy was also years old and out of date. Bugs of all kinds went unfixed and all you could expect is the way the car worked when you bought it. Cadillac said in 2013 the ATS was be different and updatable but again it was not.
I have had CUE in 2013 ATS, 2014 XTS, and current 2018 CT6 2.0E plug-in. They all worked fine, no need to update. That's the same with most of GM Intelilink systems.
I hope legacy auto can fix the Uconnect dumpster fire software they put in the Pacifica. So many critical bugs, half thoughts, partial implementation. I wonder if the same crap is in the Grand Wagoneer for 6 figure car with garbage software.
Our 2021 PacHy is working fine.
Reliability is key. You do not want to have a poor system that you need to a windows reset every so often.
Tesla continues to struggled with releasing Cyberflop software, many of the paid for functions are still missing seven months after release.
What really surprises me is how long the Cybertruck was in development, but so many paid-for functions are incomplete upon delivery to customers. Tesla comes closest to the development model suggested in this video, yet Tesla still has had trouble making features promised available to customers. Tesla’s performance makes me question some of the premises put forward in the video. One point where I agree is that manufacturers need to be better at providing necessary amenities for engineers and software developers. Flexibility in work location and hours should be a given in these days of pervasive Internet and wide band communication capacity. Workers may be productive any time anywhere and team with their cohorts without physically being in office buildings. Attracting and retaining exceptional employees is key to a durable and lasting corporate culture.
Wait, don't help me, I'll get there . . . Vertical Integration!
Its time for a change! Welcome Chinese fabulous, reliable,honest auto 🇨🇳😊
Boston Dynamics, Tesla, Nvidia and others are showing mind blowing strides on human like robots. How does an automobile differ from robots. Even ice cars go into safe mode for system failures. I can do without FSD!
The difference will be the number of vehicles in service. There could be potentially millions of semi autonomous road vehicles in operation giving manifold greater opportunity for errors to manifest themselves. In its current state Tesla FSD makes fairly serious errors, especially when it has to coexist with non-smart vehicles piloted by humans. Tesla fans like to point out the 2.5 redundancy (it isn’t triply redundant as the third channel votes, but doesn’t control) of Cybertruck steer by wire, and compare aircraft fly by wire. The difference is Airbus fly by wire is a triply redundant system of multiply redundant systems which can fall back to electrically connected manual control if need be. It makes economic sense to do that on an aircraft, but would be too expensive for a consumer vehicle.
Tesla could have had an electrically clutched steering intermediate shaft for fallback to direct manual control on the Cybertruck, but packaging of the instrument panel may have made it impractical. At least two manufacturers do steer by wire with mechanical coupled fallback if the electronics fail.
US build cars with software, China build computers with powertrain and wheels.
Software without Vision leads to boring garbage. The reason SDVs are NOT what legacy wants to do is because legacy stops development once its ships. There’s no feature updates. There’s no post-sales improvements that aren’t safety related.
Without a centralized team, nothing works together so no changes.
Everyone are now copying Tesla. 15 years after…. Elon is so far ahead🤣
Software defined cars are 💩👎
Guess who writes the software for brakess, ESP and brake booster in Tesla? Bosch. Guess who writes the software for suspension in Tesla? Continental. Guess who writes the software for all the collission systems in Tesla? Autoliv. Guess who writes the software for the LED-matrix lights in a Tesla? Hella. Guess who writes the software for the steering in a Tesla? ZF. Dont belive the hype..
That’s not what he said in the video. He said legacy needs to write the Base software with SDKs for Bosch and all the others to write their code into. They used Teslas SDKs to write into Teslas software. Legacy auto is using their software and computers and have hundreds of computers and software that can’t talk to each other and is not upgradable.
@@andrewvercillo7584 It doesent really matter what he says. Just open up a Tesla and see. The computers i mentioned are bought from theese suppliers, and Tesla are not even allowed to write a single line of code to those computers. Tesla must provide the parameters they want, and the suppliers write the code and program the controllers. The basic arcitechture of the electronics in a Tesla or a not very different from any other vehichle. That is a myth created by some people on TH-cam, including this channel.
WRONG! - Tell me you don’t know what you’re talking about without telling e you don’t know what you’re taking about.
These other OEMs you speak of aren’t “writing the software” they do have boards and chipsets (MCU’s) which have underlying firmware but firmware, strictly speaking, isn’t Software in the way you’re speaking of.
So, yes, Tesla DOES write as much software as you say they don’t. Do they have to hook into the MCU’s? Sure, but no one ever said they didn’t.
What Tesla is known for is the overall design of the software interface the users interact with and the underlying code base for things like Autopilot, etc.
Your argument is like saying Microsoft doesn’t do sh*t because other manufacturers like Nvidia for GPU’s also run their own software to integrate with the Windows OS and PC hardware…So what??
@@Av-vd3wk No. Theese supliers that I mentioned sell their components as a whole package hardware and software. Have always done, and always will.
@@GTO33 you don’t get it old timer…
Expect software engineers to transition out of irrational (fast buck) Tesla into legacy respectful employment. Tesla isn't into auto anymore.
Whatever you think about Musk we should thank him for disrupting Legacy Auto and moving their cheese 🧀
Tesla didn't make money for almost two decades and still relies on ZEV carbon credits...
Scintillating. 🤦🏽♂️
It’s easy to add features after babysmurf9000 fired all the SQA engineers!
LOL those SQA engineers will just desolve into that rats nest of corporate culture never to be heard from again. Its still pretty wild they are still struggling to do what Tesla can do in their Sleep. you will start seeing software collaboration with Tesla in the future. Its only a matter of time.
@@dclpgh Tick tock to June 13th…
@@bobbbobb4663 Tick tock to 8/8....
@@bobbbobb4663 Tick tock to 8/8.......
@@dclpgh Guess who writes the software for brakess, ESP and brake booster in Tesla? Bosch. Guess who writes the software for suspension in Tesla? Continental. Guess who writes the software for all the collission systems in Tesla? Autoliv. Guess who writes the software for the LED-matrix lights in a Tesla? Hella. Guess who writes the software for the steering in a Tesla? ZF. Dont belive the hype..
👍
Legacy automakers should contract with Tesla to solve software defined electric vehicles. Tesla Full Self Driving is revolutionary and consumers will demand autonomous new cars before we know it. Tesla engineers understand how to integrate casting, 48 volt architecture, drive by wire, get control of 100% of a vehicles controllers, etc. Tesla supply chain understands how to source materials from cooperative suppliers, etc.
Remember how GM bought the excellent Electronic Data Systems corporation in 1984 to solve its badly broken IT organization and how the acquisition didn't exactly work out? I was an GM-EDS employee in Detroit and it was a cultural nightmare.
Legacy automacers has integrated big castings for 30 years, has used 48v systems for 11 years and has bought its drive by wire systems form ZF in germany. Controllers for brakes, ESP, brake boost, suspension, lights, airbags etc are also bought from traditional suppliers in Europe.
@@GTO33 Ok, then how come legacy auto isn't able to profitably sell EV's?
@@cathyk9197 That is a totaly different discussion not related to this topic, but actually most of them do make money.
@@GTO33 Rivian, Ford, and GM aren't making a profit selling EV's. Neither is any Japanese automaker. BMW, Mercedes, and/or Hyundai/Kia possibly?
@@cathyk9197 How come it took Tesla 18 years to become profitable?
Ford's software in the Mach-e and Lightning is better than my Tesla.
Lol!
Jim Farley talked about how Ford and other legacy automakers could never have software as good as Tesla and other EV companies. You should go to the Ford headquarters and tell him he’s wrong.
@@juliahello6673 He is wrong. Guessing he hasn't owned a Tesla.
@@juliahello6673 he’s probably too busy listening to his Zune.