VERTEBRATE.ai

Twice in the span of a week, Waymo's driverless cars ran into the limits of their own competence, in ways both tragic and faintly absurd. In Austin, a family's cat, Tommy, described by his owners as "a badass" who ruled the neighborhood, was struck and killed by one of the company's robotaxis on a residential street. Days later, and roughly 900 miles west, another Waymo vehicle got stuck fast in a muddy puddle that had pooled after a storm swept through the Phoenix area, sitting immobilized until it could be retrieved.

Neither event will show up in Waymo's safety statistics as anything more than a footnote. The company, which operates paid robotaxi service in Austin, Phoenix, San Francisco, and a handful of other cities, has spent years arguing — with data to back it up — that its cars crash far less often than human-driven ones. But statistics are cold comfort to a family grieving a pet, and they say nothing about a car that simply doesn't know what to do when the pavement disappears under a sheet of water.

That's what makes this particular week instructive rather than merely bad luck. The two incidents sit at opposite ends of the autonomous-vehicle problem. Tommy's death is about perception and prediction in the chaotic, low-stakes-for-humans but high-stakes-for-someone's-pet environment of a residential block — the kind of place where a cat can dart from behind a parked car with no warning, and where the margin for a sensor suite to get it right is measured in fractions of a second. The stuck car in Arizona is about something more basic: whether the vehicle can even correctly read the ground beneath it when weather turns a road into something unrecognizable to its cameras and lidar.

For an industry still trying to convince a skeptical public that robots can be trusted behind the wheel, both scenarios are corrosive in different ways. A dead cat is the sort of story that spreads through a neighborhood and then a metro area, reinforcing the sense that these cars, for all their processing power, don't yet have the instincts of an attentive driver who might swerve, brake, or simply notice movement in the dark. A car marooned in a puddle is almost comic, but it points to something regulators care about deeply: how these vehicles behave in edge cases that don't show up often in sunny California test loops but happen constantly in the real world, from monsoon season in the Southwest to black ice in the Midwest.

Waymo has expanded aggressively, betting that scale and accumulated road experience will smooth out these rough edges over time. Investors and city officials weighing whether to welcome more driverless fleets are watching closely, because the calculus for permitting this technology rests on a public perception that is built one viral incident at a time. A stuck car gets towed and forgotten. A grieving family's story lingers, and it is exactly the kind of anecdote that shapes how quickly, or slowly, autonomous vehicles are allowed to become an unremarkable part of American streets.