Wall Street Journal auto critic Dan Neil, Pulitzer winner and staunch leftist, has at last driven Tesla’s Full Self-Driving (Supervised). After two hours he emerged “bewitched.” “FSD is amazing,” he declared. “I acknowledge I am a sample group of one, with only the scantest evidence to work with. Still, I know good driving when I see it. Mile by mile, the FSD started to earn my confidence.” He marveled at its mastery of rush-hour interstate traffic—the precise stopping and following distances, the lane centering, the “smoothly continuous linear functions” of acceleration, steering, and braking.

One can almost picture Neil as Ellis from Die Hard—the slick, self-assured corporate operator who bursts into the room convinced he’s the smartest guy present, only to hear the dry, cutting reply from Hans Gruber: “Ellis… you figured this out all by yourself?”

The line lands because the discovery arrives years late. Tesla owners and independent testers have accumulated well over a dozen billion miles on successive versions of the system. Version 14 has been widely available long enough for ordinary drivers to log tens of thousands of miles each, routinely handling unprotected left turns, construction zones, dense urban streets, and the very “whitewater-like eddy and flow” that so enchanted Neil. Mainstream coverage, meanwhile, has preferred the comfort of edge-case disengagements and carefully curated skepticism. The technology advanced in plain sight; the professional gatekeepers largely looked the other way.

Neil’s conversion is welcome precisely because he has never been a Tesla partisan. His praise carries the weight of a skeptic finally confronting the evidence. Yet the broader pattern is harder to excuse. Traditional auto writers still approach software-defined vehicles with the toolkit of the internal-combustion era—measuring 0-60 times and debating steering feel—while a neural-net driver that improves with every fleet mile sits in the next lane. The result is a chronic lag: enthusiasts and owners document competence in real time; legacy critics arrive later, astonished that the future did not wait for their permission.

This is not about demanding cheerleading. It is about basic professional currency. When a system has been operating at scale for years, delivering measurable safety advantages over average human drivers and earning the daily trust of hundreds of thousands of owners, “scantest evidence” is no longer an accurate description. The evidence was always there. Mainstream writers simply chose, for too long, not to look closely enough.

Neil has now looked. Welcome to the party, Dan, however fashionably late. The rest of the industry might consider following before the next version renders today’s astonishment obsolete.


Spies, discuss….

The article read like he was a typewriter reviewer reviewing a computer with word processing software.

You could sense his world slipping away.

— Phil Trubey (@PTrubey) August 22, 2026