Branden Cook

Robot Cars of the Future

From a 1980s magazine story to a 2026 Tesla — a driver’s account of what self-driving actually feels like.

September 29, 2026

I still remember an article in Boys’ Life from the 1980s. A family is on a road trip in the kind of car magazines promised was just around the corner. The self-driving system fails. The parents have to take the freeway themselves. What stuck with me wasn’t the gadgetry. It was how nervous they were. They had already forgotten how to drive in traffic at speed. The future had made them passengers, and then it handed the wheel back.

That scene lived in the same part of my brain as Friday nights with Knight Rider. KITT wasn’t a feature. He was a character. He talked. He thought. He saw the thing you were about to miss and did something about it. For a kid, that was the whole point of a car that could drive itself: not that it followed a painted line, but that it was smart.

Decades later, on October 23, 2023, I bought my first EV: a 2023 Ford Mustang Mach-E. On the way home I turned on BlueCruise. Hands-free on a mapped freeway. I was as tense as those parents in the magazine. Part of me still expected a talking Trans Am. What I got was a competent highway assistant with a lot of asterisks.

BlueCruise only worked on approved stretches of freeway. It dropped out in construction, faded lane lines, odd ramps, and the “problem areas” every owner learns by muscle memory. It seemed more settled at night, when headlights and simpler visual clutter helped it lock onto the world. In daylight it was useful. It was not KITT. I was still the driver. The car was a very expensive cruise control with good manners.

Then I took a test drive in a 2026 Tesla Model Y.

The difference wasn’t subtle. Full Self-Driving (Supervised) didn’t wait for a blue zone. It handled surface streets, unprotected lefts, messy intersections, and the kind of improvisation that used to be the exclusive job of a human. I traded the Mach-E. I bought the Model Y.

FSD is not a parlor trick. It will sometimes flag a pedestrian I haven’t seen yet and change the plan before the scene becomes a problem. It eases out of a bad gap. It waits when a human would have punched it. It is also not perfect. It still needs a person who is paying attention. Supervised means supervised. But each update is visibly less clumsy than the last. The car I drive now is not the car I brought home. That is the part the 1980s article never quite imagined: the software keeps going to school after you leave the lot.

The reason this matters is not novelty. It is the bill we already pay for human driving.

In 2024, NHTSA counted 39,254 people killed on U.S. roads, about 2.42 million injured, and roughly 6.18 million police-reported crashes. That is on the order of one death every thirteen minutes in a typical recent year, and a crash every few seconds if you spread the police reports across a day. Property damage, backups, emergency response, and lost work sit on top of the fatalities. NHTSA has long attributed the large majority of serious crashes to human factors: distraction, impairment, speeding, fatigue, misjudgment. In 2023 alone, alcohol-impaired driving killed more than 12,000 people; speeding was a factor in nearly 12,000 deaths.

No production system has erased that ledger. Tesla’s own fleet reports for supervised FSD claim far fewer collisions per mile than the U.S. average and fewer than Teslas driven manually with active safety on. Independent robotaxi operators publish their own, sometimes striking, reductions in serious injury crashes. Those numbers are not interchangeable, they are not unsupervised “the car is the driver” numbers in every market, and they should be read with the skepticism any manufacturer-supplied statistic deserves. What they point to is the same thing that magazine family was groping toward: most of the wreckage on American roads is not a mystery of physics. It is a mystery of attention, timing, and impulse.

I used to think the future would arrive as a breakdown on a freeway, with two scared parents remembering how to merge. It arrived instead as a quiet handoff on the way home from a dealership, then another handoff a couple of years later that felt less like cruise control and more like a partner who sees farther than I do.

KITT was fiction. The nervous parents were fiction. The crashes are not. The useful question is no longer whether a car can drive. It is how many human mistakes we are still willing to treat as normal.