I saw my first robotaxi in San Francisco two years ago. It stopped for a squirrel. Then it stopped again. Then it just… sat there. The human safety driver inside looked like they were trying not to cry. I laughed. Then I got a text from a friend: "They're gonna fix this. It's just software."
Turns out, software doesn't fix rain.
Waymo's latest pause — pulling its fleet from Atlanta, Dallas, Houston, San Antonio, Austin, and Nashville — isn't a glitch. It's a pattern. Every time it rains hard, the AI panics. Not because it can't see. Because it can't decide. It doesn't know when to stay off the road. And that's not a sensor problem. That's a judgment problem. And judgment, as any human driver knows, isn't learned from data. It's learned from living.
The company's own recall notice last week didn't mention "flooded streets." It said "perception degradation." That's corporate-speak for: "We don't know what to do when the world gets wet."
And now? They've also pulled robotaxis off freeways in San Francisco, Los Angeles, Phoenix, and Miami. Why? Construction zones. Not because the cars can't detect cones. Because they can't predict what a human worker will do next. Can't guess if someone's going to step out from behind a barrier. Can't trust the flick of a wrist.
This isn't a bug. It's the architecture.
Autonomous systems are trained on billions of miles of clean, dry, predictable data. Real roads? They're messy. They're chaotic. They're human. And the moment the weather changes, the rules change — and the AI has no playbook. No instinct. No fear.
I've driven through a Texas downpour. You don't slow down because the map says "slow." You slow down because the car ahead of you is fishtailing. Because the puddle looks deeper than it is. Because your gut says, "Don't go." That's not data. That's embodied knowledge. And Waymo's fleet? It doesn't have a gut.
The media keeps calling this "progress." But progress implies forward motion. What we're seeing is a company running in place — launching in new cities, hitting a wall, retreating, then launching again. It's not scaling. It's circling.
And here's the kicker: no one's asking why.
We've built this narrative that autonomy is inevitable. That it's just a matter of time. But time doesn't fix uncertainty. Experience does. And if you're building a system that can't handle a thunderstorm, you're not building transportation. You're building a very expensive, very slow, very dangerous toy.
I don't hate Waymo. I respect the engineering. But I'm tired of the PR spin. "We're making progress." Tell that to the driver in Houston who got stuck behind a robotaxi for 45 minutes while rain poured down and the car refused to move. Tell that to the pedestrian who had to wave their arms like a lunatic to get it to yield.
The real question isn't whether robotaxis will work. It's whether we're willing to accept them as they are — flawed, conditional, and deeply, dangerously incomplete.
Because right now? They're not ready for rain. And until they are? We shouldn't let them drive.
The Other Robotaxi Players: Who's Still in the Game?
Waymo isn't alone. But they're the only one with the scale to make these retreats visible.
May Mobility? They're quietly partnering with Ecarx — Geely's tech arm — to build thousands of purpose-built robotaxi platforms. No flashy launches. No press releases. Just steady, methodical deployment in controlled urban corridors. No freeways. No floods. Just predictable routes. That's not sexy. But it's smart.
Nuro? They're not even trying to carry people. They're delivering groceries. And they're profitable. Why? Because delivering a box of milk doesn't require the same judgment as delivering a person. The stakes are lower. The consequences of error? Less catastrophic. They've built a business on the margins — and they're thriving.
And then there's Tesla. Oh, Tesla. They're still pushing Full Self-Driving (FSD) in Lithuania, as if Europe's narrow streets and erratic cyclists are just another dataset. Musk's $1 trillion pay package is tied to hitting 10 million FSD subscriptions by 2035. That's not a product goal. It's a financial hostage situation. And the users? They're the collateral.
Lyft and Uber? Both still insist robotaxis and human drivers are a "hybrid" future. That's not a vision. That's a survival tactic. They know if they alienate their human drivers, they lose their entire workforce overnight. So they hedge. They say "both." They say "gradual." They say "not yet." It's corporate doublespeak. But it's also honest. Because they know: the tech isn't ready.
And Stellantis? They've partnered with Wayve to bring hands-free driving to Jeeps and Rams by 2028. Not full autonomy. Not robotaxis. Just hands-free. Which means: the human is still in charge. The car just helps. That's not a revolution. It's a compromise. And maybe that's the real future.
For a broader look at how Waymo and its competitors stack up globally, see our TechCrunch Mobility: New Robotaxi Scorecard Reveals China's Competitive Edge, which breaks down the latest rankings from Autnmy AI's Road to Autonomy Index.
The Nissan Leaf That Got Me Hopeful
I know. I'm supposed to be talking about robotaxis. But I got back in a 2026 Nissan Leaf last week — the Platinum+ model — and I was surprised.
It's not perfect. But it's human. The interior is clean, quiet, and feels like it was designed by someone who's actually driven a car. The 14.3-inch screen? It doesn't scream at you. The adaptive cruise? It doesn't jerk like a robot. The backup camera? It works. No lag. No blur.
And here's the thing: it's $42,635. That's not cheap. But it's not a luxury. It's a car. A good one. And it doesn't need a cloud to operate. It doesn't need a server farm to see the road. It just… works.
I've spent years writing about autonomous tech. And I've never felt this… calm.
Maybe the future of mobility isn't about removing the human. Maybe it's about making the car better at being a tool — not a replacement.
I don't know. But I know this: I'd rather drive a Leaf than ride in a Waymo when it's raining.
Final Thought: We're Not Building a Future. We're Building a Fantasy.
We keep calling this "the future of transportation." But what we're really building is a fantasy. A fantasy that machines can replace human judgment. That data can replace intuition. That algorithms can handle chaos.
We've built a world where we believe if we just throw enough compute at a problem, it will solve itself.
It won't.
Robotaxis aren't failing because they're too slow. They're failing because they're too perfect. They don't make mistakes. And because they don't make mistakes, they don't learn.
A human driver sees a puddle. They slow down. They might even laugh. Then they go faster. They adapt. They improvise. They learn.
A robotaxi sees a puddle. It stops. It calculates. It waits. And then it waits some more.
The future of mobility isn't about autonomy. It's about humility.
And until we admit that — we're just driving in circles.
