The Tesla Model 3, Long Range: an Objective Technical Assessment

Early in the Cold War, a reporter asked Albert Einstein about his recipe for world peace. He answered, “Why ask me? I’m just a physicist.” Or words to that effect.

Not so Elon Musk. Whatever the question, he has an answer. Often his answer reveals more ego than thought. Apparently many people—including an old friend of mine—evaluate his cars by assessing the man himself. They note his extreme libertarian politics and his inability to keep his mouth shut. Not surprisingly, their assessment is poor.

But Tesla cars are not Musk’s alter egos. They are physical products, as innovative and consequential as Henry Ford’s Model T in its day. So I thought maybe interested buyers might like an objective assessment by a guy who spent the first thirty years of his life immersed in science and engineering, and much of the next forty in and around the computer/software industry. I’m just a driver, not a competitor or media person.

Background. My wife has owned her long range-Tesla Model 3 for six months. During that time, we’ve driven it in just about every road and weather condition except pouring rain. We drove some 1,100 miles from Berkeley, CA, to Santa Fe, NM, in January, encountering some snow. We drove it back in June. We drove it all around Northern NM, on high-speed freeways and winding mountain roads. We’ve driven it around crowded Berkeley, at speeds seldom exceeding 30 MPH, and on interstate highways at speeds above 110 MPH and at sustained speeds near 90. Just this week we returned from weeklong “electric-car pilgrimage” to visit friends in Southern California, a region designed more for cars than for people.

Neither I nor this blog is new to electric cars. I started writing about them in May 2007—over fifteen years ago. I dithered for eleven years in trying a Chevy Volt, before leasing one for three years in 2018. Last year I bought it off the lease. We still drive it around Santa Fe, NM, where we live about half the year. I love it: it’s easier to use, but not quite as fast or as solid as the Tesla. For general thoughts on how economical electric driving can be, even before the recent gas price surge, click here.

In college in the sixties, I owned a 1955 stick-shift Chevy sedan. So I’m no stranger to good ICE cars. I loved that car as only a twenty-something can do. I overhauled its engine myself, with some help in grinding the values. Eventually, its engine blew up, due to my failure to replace the main bearings in the overhaul. So I ended up with no car in my senior year and learned some lessons about ICEs and their maintenance.

What follows is my personal evaluation based on all this experience.

The Car and How it Drives. Let me cut to the chase: the Tesla driving experience is the best I’ve ever had. Unlike the basic Model 3, the long range Model 3 has two electric motors, not just one. It’s rated at 4.2 seconds from zero to sixty. So it delivers impossible acceleration, both from the line and from speed, without any rumble or engine noise.

If you’ve driven ICE cars most of your life, as I had before 2018, you have to experience this to believe it. The rocket power and complete absence of engine noise reminded me of the scene in the movie Independence Day where the Will Smith character gets hold of an alien space fighter. He jerks its control stick back and forth. The ship moves accordingly, instantly and effortlessly, as if Newton’s First Law of Motion (the one about inertia) had somehow been repealed.

That’s how our Tesla feels to drive. Once you experience that for a few days, you’ll never look at an ICE car the same way again. You’ll appreciate in your soul how klugy is a machine that turns successive fuel explosions into rotary motion, as compared to an electric motor that is wholly symmetrical and intrinsically rotary.

The car’s curb weight is 4,250 lbs, nearly twice the weight of most similar compact cars. But most of that weight is in the battery, and most of the battery sits near or below the wheel hubs. So the car has an impossibly low center of gravity. It corners like a panther.

On a winding mountain road (NM State Highway 14, through the East Mountains north of Interstate 40), driving at around 70 MPH felt like driving at 25 MPH down a residential street. The steering tracks precisely, no matter what the speed. This despite the tires being a little harder, and consequently the ride a little harsher, than in the average ICE road locomotive.

The final joy of driving the Tesla (before we get to the bad stuff) was “single pedal driving.” This is a software-selectable option, which slows and even stops the car whenever you lift your foot off the gas pedal. When you do that, the car automatically—seamlessly, noiselessly, and jerklessly—turns the motor into a generator, which then charges the battery from the car’s forward inertia. This process is known as “regenerative braking.”

Regenerative braking is not a new technology. Every electric car has it, as do many older hybrids, such as Toyota’s Prius. Cars with a full-electric mode, including my Chevy Volt, use it routinely whenever you slow down. But not every car has integrated it as seamlessly into the driving “feel” as has Tesla. (If you want my Volt to stop fully, for example, it’s best to use a separate paddle on the steering wheel to step up the regenerative braking.)

What Tesla has done is make regenerative braking uniquely powerful, and uniquely easy, using single-pedal driving. With a little practice, you can stop the car fully using only the gas pedal, by moderating your pushing and releasing it. More than that: using only regenerative braking, you can stop the car entirely. It then stays stopped, even when facing up or down a hill. So you can learn to use the brake for panic stops only, and occasionally when you’ve failed to use the gas pedal correctly.

Once you get used to this mode of driving, it becomes relaxing and precise. If you’ve driven other cars for years, as I have, you never forget how to use the brake when necessary. But I worry about kids who learn to drive this way: how will they react in cars that require other-pedal braking? I have no idea.

These are the unambiguously good things about the Tesla. They are very good. But before you run out and buy one, be sure to read the rest of this post.

There are two important things you need to know first. The first is price: my wife’s fully-loaded Tesla Model 3, Long Range, with non-standard color, wider tires (for safety), and all the options except Autopilot, cost about $60K out the door, including California taxes and license. That’s far from the $35K that the Model 3 was reported to cost, at base, in early days. After some heated discussions with Tesla reps, we have reason to believe that our willingness to buy a premium model helped get us early delivery of the car—about a month after our order date—despite then-serious supply-chain problems.

The second thing you need to know is that our Tesla has issues: serious and annoying issues. The most serious involve software, because Teslas rely on software more than any other car you’re likely to have owned. In theory, these might disappear with software upgrades. But our experience suggests that may take some time. Other issues involve what I call “body design”—things like how the doors, trunks and glovebox open. Some of these you may have to live with forever, because they involve hardware.

Tesla Software. I first programmed a computer in machine language in summer 1961, at The Summer Science Program, a science camp for high-school kids. Since then, I’ve programmed computers in Basic and Fortran, and (if you consider that programming) I do my own HTML on this blog. As a lawyer, I spent eight years practicing “high-tech” law in San Francisco and Silicon Valley, including some of the first-ever licensing of so-called “expert systems” software, i.e., rudimentary AI. Since the Internet began, I’ve spent hours per day online.

So I’ve thought a lot about how “engineering” software differs from producing hardware, and the consequences of those differences. This prior post of mine discusses how bad software apparently killed 346 airline passengers, and how software “engineering” differs fundamentally from engineering mechanical stuff. This post and this one discuss how “communication” software often fails to communicate well. This post explains why I believe that modern hype about artificial intelligence (“AI”), including especially Elon Musk’s, is wildly premature.

All this matters to Tesla drivers because the primary “human interface” between car and driver, apart from the steering wheel and standard pedals, is a big computer screen located midway between the driver and passenger. In other words, your relationship as driver to your Tesla is much like your relationship to your cell phone or laptop.

Based on six decades of experience with computers and software, I consider the Tesla’s software abysmal. In my view it’s ill-considered, ill-planned, ill-executed, and poorly and inconsistently updated. If I were Musk, I would pay whatever it took to hire a top software manager from Google, Amazon or Apple (in that order), put all existing software on hold, and redesign its user interface from scratch. And I would do it all with constant and consistent user input—something that I’ve been unable to find anywhere in Tesla’s universe, despite persistent effort.

The rest of this essay lays out the ways, both small and large, in which Tesla’s software and body-hardware design decisions fall short. Sometimes they can lead to driver’s rage, although I hope not road rage.

The very most annoying thing about my wife’s Tesla is the absence of practically useful cruise control. The first modern automotive cruise control appeared in the 1958 Chrysler Imperial, about 64 years ago. But I can’t use (or find) anything resembling cruise control on my wife’s Tesla because it’s not there, not reliable and/or, in my view, not safe.

Apparently Tesla’s version of cruise control is inextricably linked to features of Musk’s much-hyped AI “Autopilot.” It consistently produces instances of so-called “phantom braking,” i.e. sudden onsets of hard regenerative braking for no good reason. During our trip to Santa Fe in January, it braked hard, unnecessarily and often. The only apparent causes were such things as temperature-inversion mirages on desert roads, trees’ shadows on the road, changing roadway colors (common in AZ), shadows thrown by trucks in the lane to the right, and even (in one case) a sharp and unusually ramped exit to the right.

Some time after that trip, online and media complaints about phantom braking exploded. References to “cruise control” (or similar features) disappeared from the ever-inscrutable and badly organized user interface. Yet just recently I discovered, solely by accident, that I could invoke cruise control by pressing the right-hand steering-wheel lever down (the same lever also used for drive, reverse and park). Then I could take my foot off the gas, and the car would continue at a set speed, noted on the display’s home screen. I could even adjust the set speed, up or down, by using the right vernier wheel mounted on the steering wheel.

Nothing on the display had notified me of this, far less any user update. I found it by accident. But fair enough. It seems to work well on divided highways; at least I encountered no problems while using it there. On two-lane and four-lane highways with no center divider, I suffered phantom braking within minutes, so I had to turn the cruise-control off (by pushing the same lever up).

So six months out, I’m still reluctant to use the AI-based cruise control, except carefully on not-too-crowded divided highways. Sudden phantom braking seems too big a risk to take with a teenager or pickup truck tailgating you at 75 MPH.

Body-Design Issues. The second most annoying thing about Tesla’s software user-interface is what I call the “let software do it” prejudice. The glovebox, for example, opens only with software. Often you have to hit two onscreen software buttons to open it, and only one of them is closest to the passenger. When you’re trying to retrieve binoculars, sunglasses or a Covid mask from the glovebox using the passenger-side door, that’s inconvenient. As for the economy of production, I can’t see why the kind of push button to open the glove box that cars have had for decades would cost more than a software-driven electronic actuator.

The trunks are similarly annoying. As I had suggested generally in an early blog post, the Tesla has two trunks, an ordinary rear one and a much smaller front trunk called the “frunk.” Both open with software only, although the rear trunk has a “close” button accessible only when it’s open.

Imagine this. There you are, your hands full of stuff to pack in the trunk or frunk. You have to fumble in your pocket for your cell phone to open the orifice, trying to hit the “open” button while not dropping the phone on the pavement. Or you have to put all the stuff down, on the ground, the top of the car, or the other trunk, so you can open the door, get the home screen going, and hit the software button there.

No doubt the Tesla’s rear trunk is uniquely secure. You can hear a big screw turning home when you shut it electronically. So no one is going to steal stuff from your trunk without a crowbar, a powerful drill, and a lot of noisy effort. But packing the trunk is another story—an exercise in patience and advance planning that, on car trips, can become a chore.

The doors are a similar annoyance. Yes, it’s a good idea, for reasons of safety and reduced air resistance, to have the handles flush with the car’s surface when closed. But why are the parts you push with your thumbs to the right, so that your hand is palm down when it grasps the protruding part of the handle? I don’t know about you, but my hand pulls more weakly in that position, especially when fully opening the doors against what seems like Tesla’s stronger-than-usual restraining springs.

So just opening the doors can be a chore, especially on the way to packing things. Why didn’t someone consult anatomical specialists before designing these pesky handles? Why aren’t they vertical, or slanted, the better to comport with human anatomy? Beats me. I think some hardware engineer didn’t do his or her human-anatomy homework.

The Software-User Interface. Things like the door handles and trunk openers can’t be changed, except in subsequent models. But almost everything about the software can. That’s its beauty.

Yet so far, Tesla software has suffered, big time, from the evils of apparently hasty user-interface design. For a while, in January, much of the GUI tool bar seemed devoted to various for-profit subscription services. Now they are mostly confined to a single button, which opens up a whole screenful of them. Fair enough.

But the phone-interface software needs work. When my wife and I travel together, especially on trips, both of our phones rest in their little nests on the forward console, where the car charges them inductively and you can see their screens. But the software lets only one of them be designated as “primary.”

Apparently, this means that only one can receive calls through the car’s electronics, and only one can bluetooth music into its speakers. (At least that’s been our experience.) Would it be too much trouble for the programmers to have the software answer both phones and let the user choose which to answer, and which to decline or put on hold, just as the phones do themselves in receiving near-simultaneous calls? Would it be too much trouble for the software interface to give users an easy choice of phone for both calls and music?

The environmental systems have similar annoyances. When the car is charging or otherwise stationary, the HVAC system turns off. To start it without putting the car in drive, you have to jump through a bunch of software hoops. That’s especially inconvenient during the pandemic, when one or both of us are hopping in and out to stay safe and do errands. Would it be too much to have a single button to turn the HVAC on to its last pre-stopping state, or even to use the air-bag occupancy sensors to keep it turned on whenever someone is in the car, even if stationary? Who knows? The latter feature, if adjusted for light weight, might save a pet’s life on really hot or cold days.

My last significant complaint concerns Tesla’s navigation software. Apparently Musk decided not to license Google Maps or even Apple’s less mature competing product. Instead, he decided, so to speak, to re-invent the wheel. The result is unimpressive.

After six months and about an equal number of annoying and stressful missed directions, we no longer use Tesla’s internal navigation software for anything but Tesla’s own supercharging stations. For all other directions we use Google Maps on our cell phones.

There are several reasons why. First. Google’s visual directions are richer, with arrows showing every turn, lane warnings and indications for multilane intersections, and visual distance countdowns (to key turns and destinations) by each 0.1 mile and sometimes down to feet.

We’re not sure about Tesla’s oral directions, as we turned its voice off early in our experience due to inability to reliably control its often startling stentorian volume. Even Tesla’s directions to its own superchargers often leave us wandering around huge shopping centers, or the backs of hotels or service stations, wondering precisely where the chargers are, until we spot their distinctive colors.

So I’ve reached the conclusion that much of Tesla’s software development has followed the “talking dog” philosophy: it’s not that the dog talks well, but that it talks at all. Tesla doesn’t seem to understand that software—especially the part that communicates with and informs people—is an exercise in human communication. No programmer or manager, no matter how smart or experienced, can possibly anticipate how many ways drivers will use the software, and with how many other systems it might interact in strange ways. That’s why the best software-development firms maintain user focus groups and admit (and heed) constant user feedback, even if the old industry custom of issuing “beta” (development) versions to self-selected users has largely gone the way of all flesh.

As far as I can tell, Tesla software updates are initiated and designed without significant user input. At least, after six months, I’ve been unable to find a way to submit the feedback that most software producers seek from users routinely. So Tesla seems to be bucking the trend of the software industry itself, which offers users every opportunity to criticize and comment, often through links on the very pages of a primary Website.

Perhaps the cause is Musk’s management style. Although brilliant as a (mostly self-taught) hardware engineer, he seems an industrial autocrat in the mold of Henry Ford, Thomas Edison, and Steve Jobs. He seems personally unreceptive to criticism, and not too receptive to advice from underlings or customers. He doesn’t seem to need or get advice from investors, since he has self-financed Tesla with his killings from PayPal.

If so, a change of heart, with a change of inferior management, might make a big difference in the quality and user friendliness of Tesla’s software and improve its several body-design defects. These seem to be the chief impediments to making the Tesla the best car ever. In the meantime, we Tesla drivers can enjoy a superbly engineered automobile whose user friendliness in software and small things needs a lot of work.

We can also hope that Elon Musk will someday grasp the essential differences between hardware and software and/or hire people who do. And while he hypes his precious AI, in the hope of saving all us still-fallible humans from ourselves, we geezers are best advised to avoid it like Covid. Those of us in our late seventies can be pretty sure that Autopilot will not be anything like ready (let alone approved) for autonomous driving until we are too old to drive ourselves. By then we may not care.

What we need now is on-board software that is intuitively easy to use and communicates and works well and reliably, just like the best software now available on the Web. Tesla’s onboard software now falls far short of that standard.

For brief descriptions of and links to recent posts, click here. For an inverse-chronological list with links to all posts after January 23, 2017, click here. For a subject-matter index to posts before that date, click here.

Permalink to this post