Writing the Tesla Code That Helped Spur a Revolution

Craig Carlson helped to define one of the first software-defined vehicles—the Tesla Model S. It's been more than 10 years since the Model S hit the market, the vehicle that not only jump-started the auto industry's current electric-vehicle revolution but also accelerated the transition to SDVs. As with many groundbreaking developments in the technological revolution, this one happened in Silicon Valley, though at the time many were betting the spunky EV startup would go the way of dozens of other unsuccessful automotive ventures.

But Tesla had a not-so-secret weapon that the legacy automakers either didn't possess or underestimated. Not battery technology (though it had a lead there as well), performance prowess, or even design flair, but rather, the use of software and a bunch of experienced and dedicated coders to create it. People like Carlson, the vice president of firmware and electrical integration at Tesla from 2007 to 2014 and current CTO at the Silicon Valley 3-D printing startup Carbon3D. Carlson and others helped midwife the Model S by leading the software and electrical architecture design for the landmark EV.

All About the Software

Carlson credited Tesla's software-focused approach to vehicle development for setting a standard most competitors are still trying to reach. "We didn't leverage the way other people built cars," Carlson said. "But we were also kind of ignorant about how they'd been developed in the past."

This led to the coding of all of the software for Tesla's vehicles in-house—a move Carlson described as somewhat naïve but genius in retrospect. That Tesla created and controlled all the software for the Model S—and started from scratch—turned out to be an important advantage and differentiator for the company. "We were trying to make the Model S in two-and-a-half years from a standing start, and it was a pretty hairy undertaking," he said. "It allowed us to have a clean-sheet approach to solving really, really hard problems."

The Silicon Valley Advantage

Looking back, Carlson believed that Tesla had an advantage by being based in Silicon Valley, especially given many of the people involved had experience in building software-centric organizations. Carlson himself started an accounting software company that was acquired by Intuit, the makers of QuickBooks.

"I didn't know much about how accounting was done, but I also didn't try and do it the way it had always been done," Carlson said. "The same is true about how we built that car. Because the software in the vehicle was built by a Silicon Valley organization, we wouldn't have to consider building it any other way."

See All 4 Photos4

He admitted, however, that he and his colleagues didn't see building Tesla's software in-house as being hugely beneficial as they were doing it. "We just thought, 'Well, we have a challenging technical problem to solve, and software is a powerful tool for doing that,'" he said. "Writing our own software was a significant undertaking. But we didn't really think about it at the time as such an unusual thing until I started interacting more with other OEMs."

Realizing Musk's Unconventional Vision

Carlson also credited Tesla CEO Elon Musk's unconventional vision for the company's approach to building cars. "We were blessed with a really strong team and with a unique leader," he said. "Elon's got his strengths and weaknesses, but he's got some really unusual and important strengths that I think made the company so it could work. He made a decision to not use people from the automotive industry to a very large degree, but instead have people that were passionate about the space and knew how to engineer."

The working environment Musk created also proved to be a significant factor in the overall development of the Model S. "Elon set an incredibly high bar for what he wanted the vehicle to be able to do," he said. "He set it high enough that most of us who led the engineering organizations thought there's no way we can do what he's asking. But innovation flourished in this way throughout the engineering organization, not just in the software."

Adding More Ingredients

While software was the secret sauce that largely led to Tesla's success, Carlson says there were other significant software-related ingredients. "One was that we wrote almost all of the software for all the ECUs in the car," Carlson said. "The second is that we installed a modem in every car so we had access to data from the vehicles. The third was we used powerful, modern computing capabilities in the car, which was unheard of at that time."

Taken together, this gave Tesla complete control of the software in the vehicle as well as incredible flexibility in how the car operated and communicated. "It made it so that when we wanted to make changes to the way the vehicle behaved, we could do that no matter if it touched two, three, or more ECUs," Carlson said.

This is in contrast to traditional automakers, which typically use numerous separate ECUs supplied by third-party suppliers. That can lead to long lead times and lots of back and forth to get everyone on the same page if changes need to be made. "Instead of taking months, we could make changes in days or even hours," Carlson said.

Over-the-Air Importance

According to Carlson, enabling over-the-air (OTA) software updates for the Model S was another huge leap for the brand. While several legacy automakers had used software updates before the Tesla sedan was introduced, they were primarily limited to powertrain fixes and had to be performed at a dealer.

Even when software updates could be performed by car owners starting around 2010, they were only for infotainment and done via USB drives. Starting with the Model S, Tesla's OTA updates were an easy and frequent way to add new software features to a vehicle after it rolled off the assembly line. "The Model S I bought when I left the company in 2014, about a year later started parking itself—it didn't have that feature when I bought it," Carlson said.

He credits OTA as not only a differentiator for Tesla but also as critical to its very existence. "If we had everything else from a technology perspective but couldn't have updated the car over the air, I don't think we would have survived," he said. "I don't think we could have done it in an era when OTA updates weren't possible. It made the cars much more reliable than they otherwise would've been."

Part of this was because Tesla was under the gun to quickly produce vehicles. "We shipped the car before it was done in many ways in order to stay alive," Carlson said. "We shipped it without features that people would say are required." For example, there wasn't a radio in the Model S at the outset of production. It had the electronics necessary for one, but there was no user interface, which Carlson said was added six months later.

OTA updates also allowed Tesla to fix problems in cars that were shipped but before customers bought them. "We were able to just keep moving forward," Carlson said. "I think it was a huge part of why we were able to succeed."

Using Data to Fix Teslas

All of that data moving back and forth over the air wouldn't have been nearly as effective had the firm not added substantial computing power to its vehicles, a move that allowed the company to extract and store massive amounts of data from the entire Tesla fleet. "It took a fair bit of storage to be able to capture all the data in the vehicle, and I think a lot of companies would've thought of it as too expensive," he said.

"We were able to get data from all of the vehicles, learn from that data, and then make improvements that affect all the vehicles," Carlson said. "The vehicle could have a problem that the owner might or might not notice. We could learn what that problem is and decide how we're going to fix it—and then fix it in every car with an OTA update before the rest of our owners ever experienced that problem."

It's a much different approach than the way most automakers operate today. Though some have begun using OTA updates to varying degrees, they still typically ship vehicles and then recall them to the dealerships if a problem arises.

Tesla's small size at the outset of Model S production also allowed it the flexibility to do things that bigger organizations weren't capable of, so equating the then upstart EV maker's software processes to that of traditional automakers was an apples-to-oranges comparison. "It was a fast-evolving process and very, very different from the way traditional automakers did it," he said.

See All 4 Photos4

Adapting to a Changing World

While Carlson noted that the traditional auto industry has learned several lessons from Tesla's approach to vehicle software, he admitted "that no one else was gunning for us at the time," during the development of the Model S. "No one particularly cared, frankly," he said of the early Tesla doubters and the general auto industry skepticism about the EV startup from Silicon Valley. "Their reaction at the time was that Tesla is unimportant."

The initial pessimism concerned Carlson. "The way the leadership [at large OEMs] carried themselves at that time scared me because the world was changing and they didn't appreciate it," he said. "It was really easy for people to overlook Tesla, and it's not been until quite recently that people can't ignore it anymore."

Those attitudes began to change, though, once the industry as a whole started to understand how critical the concept of a software-defined vehicle is to the future of automaking. "Several times, senior people came to visit us to see what we were doing from a software perspective—and their eyes would go wide," he said. "That we had such control of the software—it was really clear to them what an asset it was."

Of course, there's far more to automaking than simply having a strong software backbone, and Carlson readily acknowledged that legacy automakers still have a huge advantage when it comes to the holistic production of the automobile. "Software isn't the only thing," he said. "At the end of the day the cars are mechanical devices, and you've got to be good at that part or you're not going to make it."

Searching For the Best Solution

But despite the lessons learned and the general understanding of how critical software will be to the future development of the automobile, Carlson noted that several challenges remain.

"They have to be software companies if they really want to thrive," he said of what today's automakers will need to do in order to truly transform their business. "That changes who and how you recruit, how you motivate engineers, how to embrace change and encourage starting over and maintaining control of all the software," Carlson added. "It's a lot of things that are really different from the world those folks live in—they have to find the courage to do it differently."

Doing it differently for one thing would mean taking ownership of the software in their vehicles. By owning what Carlson said is the biggest asset in the car, it will allow automakers to further define themselves from the rest of the industry, create buzz around their brands, and offer more value to their customers.

"How do you differentiate yourself if you're using the same code as everybody else?" he said. "How do you move faster than anybody else if you don't actually have control of the biggest thing you're trying to move forward? It's a tough road."

But simply throwing money and resources at the problem when you're behind the curve isn't necessarily the answer, either, according to Carlson. "If you hire 500 people to go tackle software, I feel you're already starting at a disadvantage," he said. "We built a whole car from the ground up and without a lick of code with maybe 50 people—and I think that was a huge advantage."

Carlson added that partnering with a tech company skilled in automotive software development may be the best way forward for automakers who are struggling to catch up in the space. No matter how they get there, taking a couple of pages from the playbook Carlson and his team wrote a decade ago certainly couldn't hurt the cause, as success of Tesla's approach to the software-defined vehicle has aptly demonstrated.

Cars and Computing: Carlson Went from Big V-8s to Bytes of Code

Craig Carlson grew up in a family of auto enthusiasts in Southern California, wrenching on his own vehicles. "Big V-8s were a big part of what I loved," he said. He ended up computer crazy as well. Just prior to his senior year in high school, a math teacher acquired an Apple II computer and allowed him to use it, which piqued his interest in programming. That teacher later started a computer class Carlson enrolled in. The syllabus required writing 30 pieces of software before the end of the quarter. By the first weekend of school, Carlson had completed all 30 assignments.

After high school he enrolled at Stanford, where he initially decided to pursue a mechanical engineering degree "because I loved vehicles." He wanted to keep growing his programming knowledge as well, but Stanford didn't have a computer science program at the time. "So I got a double engineering degree; it was like an engineering/computer science degree. That was the closest you could get to computer science at Stanford in those years," he said.

While still in college, Carlson used his coding skills to form an accounting software startup that was acquired by Intuit in 1991. He later led development of Intuit's QuickBooks and QuickBooks Online products. After leaving Intuit at the end of 2006, he went to work for another startup in early 2007: Tesla. The rest, as they say, is automotive history. "We really wanted to move the needle," he said of his early days at Tesla. "We didn't set out to change the course from a software perspective; we set out to electrify transport for most of us."

See All 4 Photos4

FAST FACTS: CRAIG CARLSON

Resume Key Points

• Started small software company (1983-1991), sold to Intuit
• Led development of QuickBooks and QuickBooks Online at Intuit from 1991-2006
• Vice president of software and electrical integration, Tesla from 2007-2014. Responsibility for all low-voltage electronics and software for Model S

Education

• Bachelor of Science in electrical engineering, Stanford University

Present position

• Chief technical officer at Carbon Inc., responsible for hardware, software, and materials development

Hometown

• Los Angeles

Currently Residing

• Los Altos, California