Waymo

One-line summary: Alphabet's fully driverless robotaxi service — the operational yardstick against which every other autonomous-driving program is measured.

What it is

Waymo operates a fully driverless, paid robotaxi service using a sensor-fused stack (LiDAR + radar + cameras + HD maps). Unlike tesla-fsd, Waymo has cleared the US regulatory bar for SAE Level 3+ autonomous operation and reports per-trip data publicly.

Why it matters to autonomous-driving

Waymo is the "existence proof" for commercial driverless service. Numbers from Waymo are the reference other programs are compared to.

Key facts

• Scale: ~500,000 paid robotaxi rides per week across 10 cities; 15M rides completed in 2025 2026-04-20-autoresearch-tesla-fsd.
• Austin presence: ~100 vehicles, ~90 sq mi geofence 2026-04-20-autoresearch-tesla-fsd.
• Safety-driver removal threshold: Waymo pulled safety drivers after reaching ~30,000 city miles to critical disengagement — ~37× the ~809 currently observed on tesla-fsd v14.2 per community-tracker data 2026-04-20-autoresearch-tesla-fsd.
• Sensor approach: Redundant LiDAR + radar + camera + HD maps 2026-04-20-autoresearch-tesla-fsd.
• Transparency: Submits detailed per-trip data under California AV permit; Tesla (under TCP/limo permit) does not 2026-04-20-autoresearch-tesla-fsd.

Strengths (from our perspective)

• Operational track record at scale.
• Regulatory posture in good standing — holds full AV permit in California.
• Sensor redundancy mitigates the visibility-failure class of incidents driving nhtsa's EA26002 against Tesla.

Weaknesses (from our perspective)

• HD-map dependence limits expansion velocity compared to a truly generalizable stack.
• Per-city geofence still bounded.
• Per-vehicle cost (sensor stack) higher than tesla-fsd's camera-only hardware.

Operator framings from Karpathy (October 2025)

andrej-karpathy (Tesla AI / Autopilot lead 2017–2022) gives several first-person-adjacent assessments of Waymo's deployment reality. They're held alongside the "operational yardstick" framing above as a structural caveat.

• The 2014 perfect-drive demo data point. andrej-karpathy in 2025-10-17-dwarkesh-patel-andrej-karpathy-summoning-ghosts: "when I was joining Tesla I had a very early demo of a Waymo and it basically gave me a perfect drive in 2014 or something like that. So perfect. Waymo Drive a decade ago took us around Palo Alto and so on... and then still took a long time." Anchors the demo-to-product-gap-march-of-nines argument: a perfect demo in 2014 still required ~12 years to reach the current minimal-but-real deployment, and Karpathy contends the gap to deployment-at-scale is still substantial.

• "Deployments still are pretty minimal" + capex problem. andrej-karpathy in 2025-10-17-dwarkesh-patel-andrej-karpathy-summoning-ghosts: "self driving cars are nowhere near done still. So the deployments still are pretty minimal. So even Waymo and so on has very few cars. And they're doing that, roughly speaking, because they're not economical, because they've built something that lives in the future. And so they had to pull back future, but they had to make it uneconomical. So they have all these, there's all these costs, not just marginal costs for those cars and their operation and maintenance, but also the capex of the entire thing. So making it economical is still going to be a slog, I think, for them." The "500K rides/week, 10 cities" headline number is held against this: per-ride economics are reportedly not breakeven, the capex of the sensor stack + HD mapping + ops infrastructure is substantial, and current scale is bounded by economics as much as by safety.

• Teleoperation-in-the-loop claim (unconfirmed by Karpathy himself). andrej-karpathy in 2025-10-17-dwarkesh-patel-andrej-karpathy-summoning-ghosts: "when you look at these cars and there's no one driving, I also think it's a little bit deceiving because there are actually very elaborate teleoperation centers of people actually kind of like in a loop with these cars. And I don't have the full extent of it, but I think there's more human in the loop that you might expect. And there's people somewhere out there basically beaming in from the sky. And I don't actually know they're fully in the loop with the driving. I think some of the times they are are, but they're certainly involved and there are people and in some sense we haven't actually removed the person, we've moved them to somewhere we can't see them." Karpathy explicitly caveats he doesn't have the full extent. This is a strong claim that needs corroboration — see related question does-remote-supervised-robotaxi-qualify-as-driverless. If true, materially changes how "driverless" should be parsed in Waymo's headline metrics.

• Geofence as signal-quality proxy. andrej-karpathy in 2025-10-17-dwarkesh-patel-andrej-karpathy-summoning-ghosts: "Waymo can't go to all the different parts of the city. My suspicion is it's like parts of city where you don't get good signals anyway. So basically I don't actually know anything about the stack. I mean, I'm just making up stuff." Speculative — Karpathy explicitly disclaims — but worth noting as a possible explanation for the geofence-shape pattern.

Open questions

• When (if ever) does tesla-fsd close the scale and reliability gap?

Sources

• 2026-04-20-autoresearch-tesla-fsd
• 2025-10-17-dwarkesh-patel-andrej-karpathy-summoning-ghosts (multi-context) — Karpathy's 2014-demo / "deployments still minimal" / teleoperation-in-the-loop / capex observations.

Related

• tesla-fsd
• demo-to-product-gap-march-of-nines — operationalizes Karpathy's 2014→2026 gap claim
• mercedes-drive-pilot
• vision-only-vs-sensor-fusion