An assessment of electric vehicles, driver assistance, and the limits of automation
In the early 1980s, when I began researching electric vehicles at the Electric Power Research Institute, they were widely regarded as little more than glorified golf carts. Few people imagined that electric vehicles would eventually compete successfully with gasoline-powered cars.
Since June 2019, I have driven two generations of the Tesla Model 3 for more than 60,000 miles. My experience has been overwhelmingly positive when it comes to the electric vehicle itself. But my recent experience with Tesla's Full Self-Driving (FSD) system has been much less satisfactory.
The distinction is important. A great electric vehicle is not necessarily a great self-driving car.
The First Five Years
In June 2019, I traded in my beloved 2005 Porsche Boxster for a Tesla Model 3 Long Range. The Porsche was a beautiful convertible with excellent handling, an attractive interior, and a driving experience that I had enjoyed for many years. One of my fondest memories was driving it through Yosemite Valley with the top down, surrounded by granite cliffs rising thousands of feet above us.
The Tesla offered a very different experience. It had extraordinary acceleration, low maintenance requirements, and the convenience of charging at home. Its touchscreen displayed an impressive amount of information, including how much electricity its consuming per mile. The number is shown as watthours/mile. I wish it showed miles/kWh, which would be analogous to the mpg information shown in gasoline cars.
Tesla’s Supercharger network is extensive and helps to mitigate range anxiety, a key barrier to driving an EV.
Over the next five years, I drove the Tesla approximately 40,000 miles, achieving an average energy efficiency of roughly 3.8–3.9 miles per kilowatt-hour. Regenerative braking reduced wear on the conventional brakes so effectively that they never needed replacement.
Maintenance expenses were remarkably low. There were no oil changes, spark plugs, or engine tune-ups. Charging during off-peak hours was generally less expensive than buying gasoline, although rising electricity prices in California gradually reduced the savings. Rooftop solar panels and a home battery helped manage those costs.
The Tesla had shortcomings. Its suspension was stiff, freeway wind noise was noticeable, and the interior lacked the refinement of my Porsche and Mercedes-Benz. The central touchscreen could also be distracting.
Nevertheless, in an August 2024 article for Energy Central, I awarded my first Tesla a score of +7 on a scale ranging from -10 to +10. I said I would buy it again without hesitation.
The Second Generation: A Remarkable Improvement
In November 2024, I purchased a redesigned Tesla Model 3 Highland, manufactured a few months earlier. The Tesla salesman explained that the same vehicle was being marketed as a 2025 model by October 2024.
The improvement over my original Model 3 was substantial. The suspension was much smoother, wind and road noise were significantly reduced, and the overall driving experience became more refined.
The new car also had dual motors and all-wheel drive, providing greater confidence when traveling to mountainous destinations such as Lake Tahoe and Yosemite.
Its advertised range was 341 miles. More importantly, its energy efficiency was noticeably better than that of its predecessor. All that information is displayed on the screen.
During the first year, I achieved approximately 4.4 miles per kWh. After more than 23,000 miles, my lifetime average stands at approximately 4.2 miles per kWh. Compared with the 3.8–3.9 miles per kWh of my previous Model 3, that represents an efficiency improvement of roughly 8–11%, despite the newer car having all-wheel drive.
For someone who has spent much of his professional career studying energy efficiency and electricity pricing, that is an impressive technological achievement.
The Economics of Electric Driving
The economics of replacing my first Tesla were unusually attractive.
Tesla offered a 10% discount on the new vehicle, a $500 loyalty discount, and three months of free Supercharging. I also qualified for a $7,500 federal tax credit.
CarMax offered $21,000 for my old Tesla, compared with Tesla's trade-in offer of $18,900. After accounting for these incentives and the sale of the old car, my net outlay was approximately $19,600.
During the first year of ownership, I drove 11,901 miles without incurring maintenance expenses. I charge almost entirely at home, unless I am on a long trip.
At an electricity price of 23 cents per kWh and an efficiency of 4.2 miles per kWh, the energy cost was approximately 5.5 cents per mile.
By comparison, a gasoline-powered car achieving 35 miles per gallon would cost approximately 14.3 cents per mile with gasoline priced at $5 per gallon.
Thus, the Tesla's energy cost was approximately 62% lower.
Of course, the comparison depends heavily on local electricity and gasoline prices. California's high electricity rates make EV ownership less financially attractive than in states with cheaper electricity.
Two Often-Overlooked Advantages: Silence and Clean Air
There are two additional benefits of electric vehicles that deserve greater attention.
The first is the absence of combustion-engine noise. An EV has no engine idling at a traffic light, no engine revving during acceleration, and none of the vibrations associated with an internal combustion engine. The result is a remarkably smooth and quiet driving experience. Tire and wind noise remain, particularly at freeway speeds, but the absence of engine noise makes driving more relaxing.
The second benefit is the absence of tailpipe emissions. An EV produces no exhaust fumes while idling, accelerating, or traveling at freeway speeds. This is particularly important in congested urban areas, where vehicle emissions contribute to air pollution and respiratory illnesses.
Of course, electric vehicles are not entirely emissions-free. Electricity generation can produce emissions, depending on the generation mix, and manufacturing batteries has environmental consequences.
Nevertheless, eliminating tailpipe emissions is an important environmental and public-health achievement. As electricity generation becomes cleaner, the environmental advantages of EVs should increase.
These benefits extend beyond the individual owner. Reduced noise and cleaner urban air benefit society as a whole.
Freeway Driving Versus City Driving
One of the more interesting lessons from seven years of EV ownership concerns energy efficiency. Gasoline-powered vehicles are often more fuel-efficient on highways than in stop-and-go city traffic. Electric vehicles can exhibit the opposite pattern.
Regenerative braking allows an EV to recover some of the kinetic energy that would otherwise be dissipated as heat during braking. Consequently, urban driving can be surprisingly efficient. At freeway speeds, aerodynamic drag becomes increasingly important.
The force of aerodynamic drag is proportional to the square of vehicle speed, while the power required to overcome it increases with the cube of speed. The energy required per mile to overcome aerodynamic drag therefore increases approximately with the square of speed.
This explains why driving at 75 mph can consume substantially more electricity per mile than driving at 60 mph.
On two recent trips, I recorded energy consumption of approximately 196 and 210 watt-hours per mile. The second trip involved a combination of freeway and city driving, with freeway speeds sometimes reaching 75 mph but frequently falling to 60–65 mph because of traffic.
These observations have made me more conscious of the relationship between speed, driving conditions, and energy consumption.
When Innovation Becomes Irritation
Not every change in the redesigned Model 3 represents progress.
Tesla eliminated the conventional turn-signal and gear-selector stalks. I regard this as a misguided innovation. Turn signals are now activated using buttons on the steering wheel. They are less intuitive than conventional stalks, particularly when negotiating turns and roundabouts. Selecting forward or reverse requires interacting with the touchscreen. The central display offers an enormous amount of information, but it can also divert attention from the road.
The seats and interior comfort remain inferior to those of my Porsche and Mercedes-Benz. Yet the Tesla's acceleration, handling, and overall driving performance are excellent.
This illustrates a recurring theme in my experience: technological innovation is valuable when it improves the driving experience, not simply when it replaces a familiar mechanical control with software.
Enter Artificial Intelligence
My experience with Full Self-Driving has been very different. I first experimented briefly with FSD in my original Tesla in 2019 but did not continue using it. With the newer Model 3, equipped with more advanced cameras and computing hardware (HW4) and software (14.3.12), I decided to give the technology another chance.
Several friends had spoken enthusiastically about FSD. Some said it made long-distance driving less stressful. Others believed it drove better than they did.
Initially, I was impressed. The car negotiated intersections, followed curves, changed lanes, responded to traffic signals, and sometimes parked itself. Watching the steering wheel turn without my intervention was fascinating.
But the limitations soon became apparent. The car sometimes selected inappropriate parking spaces, including a handicapped space and restricted parking areas. It ignored directional arrows in parking lots and frequently stopped in my driveway instead of entering the garage.
On freeways, it sometimes moved out of the fast lane several miles before an exit. In city traffic, it occasionally waited too long to move into the correct lane for an upcoming turn.
More troubling were four serious near misses. At a busy roundabout, the car nearly collided with another vehicle. On another occasion, it approached a line of vehicles waiting to enter a freeway and nearly rear-ended the car ahead. In a third incident, it made a right turn when the light turned green and nearly struck a pedestrian in the crosswalk. In a fourth, it attempted a left turn from an exit lane marked for right turns. When I intervened, the vehicle swerved sharply and nearly struck roadside obstacles.
I had to intervene in all four cases.
These experiences fundamentally changed my assessment of FSD.
What Tesla Told Me
I discussed these incidents with Tesla employees at a showroom and service centers.
Several employees suggested that FSD operated correctly approximately 90–95% of the time and that the problems I encountered were not particularly unusual. Tesla reviewed my vehicle’s telemetry and reported that the cameras and other hardware were functioning normally. The company characterized FSD as a work in progress. However, Tesla did not share the underlying telemetry data with me, and the service ticket was eventually closed.
This left me with an uncomfortable conclusion. The problems did not appear to arise from a malfunctioning vehicle. They appeared to reflect limitations of the driving software.
After approximately 900 miles of FSD operation during a month-long trial, I canceled my subscription on October 9, 2026. I found supervising FSD more stressful than driving manually. And, to top it off, I had to pay $99 a month plus tax to Tesla.
It was like supervising an exceptionally talented student driver who occasionally made an inexplicable and potentially dangerous mistake. The difficulty was that I never knew when the next mistake might occur.
Discovering the Benefits of Autosteer
After canceling FSD, I began experimenting with Tesla's simpler Autosteer feature. Autosteer combines Traffic-Aware Cruise Control with automatic lane centering. Traffic-Aware Cruise Control maintains the selected speed and adjusts it to preserve a suitable following distance from the vehicle ahead. Autosteer adds steering assistance to keep the vehicle near the center of its lane.
I initially found Autosteer irritating because it disengaged when I changed lanes, sometimes accompanied by an unpleasant warning sound. I subsequently discovered that using the turn signal before changing lanes made the transition much smoother.
I now signal, take over steering, complete the lane change manually, and reactivate Autosteer. The arrangement works well.
I remain responsible for deciding when to change lanes and which route to follow. The car assists with maintaining speed, following distance, and lane position. I find this much less stressful than supervising FSD. Plus, there is no monthly subscription fee.
Another interesting observation concerns lane centering. I discovered that I tended to drive slightly toward the left side of my lane. When I activated Autosteer, the car made a fairly abrupt correction toward the center. Once it has established that position, however, it maintains the lane smoothly, including on freeway curves.
I have learned to position the car closer to the center before activating Autosteer, reducing the need for an initial correction. The experience has made me more conscious of my own lane-positioning habits.
Autosteer is not autonomous driving. I must remain attentive and ready to intervene. But for my freeway driving, it provides a useful balance between assistance and personal control.
Why Is Tesla Removing Autosteer From New Vehicles?
There is an irony in my discovery of Autosteer's benefits.
In January 2026, Tesla stopped including basic Autosteer as standard equipment on newly ordered Model 3 and Model Y vehicles in North America. Traffic-Aware Cruise Control remains standard, but customers who want the additional lane-centering capability are directed toward Tesla's Full Self-Driving subscription.
At the announced subscription price of $99 per month, the annual cost is $1,188 plus tax. Over five years, that amounts to $5,940 plus tax, assuming the price remains unchanged.
Tesla's decision appears consistent with its broader strategy of increasing recurring software revenue. It is also worth noting that some competing manufacturers include lane-centering assistance as standard equipment.
Existing Tesla owners such as myself retain the Autosteer features included with their vehicles. From a business perspective, encouraging FSD subscriptions may make sense for Tesla. From a customer's perspective, however, the decision is more questionable.
My experience suggests that a relatively simple, predictable driver-assistance system may provide greater practical value than a more sophisticated system that requires constant supervision of unexpected decisions.
More automation does not necessarily mean a better driving experience.
The Difference Between 95% and 100%
My experience also raises a fundamental question about applying artificial intelligence to safety-critical activities.
An AI system that produces satisfactory results 95% of the time may be extremely useful for drafting documents, analyzing data, or answering questions.
Driving is different.
A single serious error can have irreversible consequences.
Moreover, a system that performs impressively for long stretches may encourage complacency. Maintaining constant vigilance while a vehicle appears to be driving competently can be psychologically demanding.
Tesla correctly describes FSD as a supervised system. The driver remains responsible for operating the vehicle safely.
The technology is undeniably impressive. But impressive is not synonymous with dependable.
The challenge is not simply to automate more driving tasks. It is to demonstrate that the automation improves safety and reduces the driver's overall workload.
An Unexpected Benefit: Becoming a Better Driver
Despite my disappointment with FSD, the experience taught me some valuable lessons.
It reminded me to stay centered in my lane, maintain an appropriate following distance, stop completely at stop signs, look both ways before proceeding, and signal before every turn.
It also reinforced the importance of paying undivided attention to the road rather than becoming distracted by scenery or technology.
In a curious way, supervising FSD made me more conscious of my own driving habits.
My subsequent experience with Autosteer reinforced those lessons, particularly the importance of lane positioning and deliberate lane changes.
Thus, a technology designed to assume more of the driving task ended up making me a more conscientious driver.
Perhaps that was one of its most valuable contributions to my driving experience.
Tesla Versus Porsche and Mercedes-Benz
My experience with three very different automobiles has also influenced my thinking.
The Porsche Boxster represented the traditional pleasures of driving: elegant styling, responsive handling, and an attractive interior.
My Mercedes-Benz E350 emphasizes comfort, refinement, and familiar controls.
Tesla represents a different philosophy, emphasizing electric propulsion, software integration, minimalist design, and continuous technological innovation.
I admire much of what Tesla has accomplished. Its instant acceleration, quiet operation, regenerative braking, energy efficiency, and over-the-air software updates have fundamentally changed what I expect from an automobile.
But technological sophistication does not automatically translate into better ergonomics or a better driving experience. I miss the conventional stalks and physical controls of my earlier cars.
Tesla's cameras provide useful additional views, particularly when turning my neck is difficult. Yet looking at the touchscreen can also divert my attention from the road. The best technology should complement human judgment rather than complicate it.
My Updated Scorecard
After more than seven years of electric vehicle ownership, my assessment of the Tesla Model 3 as an EV has become even more favorable.
The redesigned Model 3 is quieter, smoother, more efficient, and more capable than the original.
Its acceleration remains exhilarating. Its maintenance requirements are minimal. Its operating costs are attractive, and the convenience of charging at home is difficult to surrender once experienced.
I award the redesigned Model 3 a score of +9 on my scale of -10 to +10, excluding FSD.
That represents a substantial improvement over the +7 I awarded the original Model 3 in 2024.
FSD deserves a separate and much more critical assessment.
Its technological accomplishments are remarkable, but the unexpected failures I experienced prevent me from regarding it as a dependable substitute for human driving.
Autosteer, by contrast, has demonstrated the value of simpler and more predictable driver assistance.
The car has earned my confidence. The self-driving software has not.
Looking Ahead
My journey with electric vehicles began more than four decades ago, when they seemed like a distant technological possibility.
Today, I cannot imagine returning to a gasoline-powered car as my primary vehicle.
The electric vehicle revolution has succeeded in ways that would have seemed improbable when I began studying the subject.
The transition to autonomous driving is a different matter.
Artificial intelligence will undoubtedly continue to improve, and future driving systems may eventually become safer and more dependable than human drivers. But that outcome must be established through rigorous evidence rather than assumed on the basis of technological enthusiasm.
For now, I remain an enthusiastic EV driver without the proverbial seven-year itch.
I am also a satisfied user of basic Autosteer, but a skeptical former subscriber to Full Self-Driving.
The most important lesson from seven years behind the wheel of a Tesla is that technological progress should be measured not by how much a machine can do, but by how much it improves the experience, safety, and well-being of the person using it.
Tesla has largely succeeded in making an excellent electric car. It has not yet convinced me that the car should drive itself.