Waymo One is now open to everyone in San Francisco

Original Article ↗ Hacker News Discussion ↗

Waymo has opened its driverless taxi service to all riders in San Francisco, prompting reports that the cars feel cautious, smooth, and generally safer—or at least more predictable—than many human-driven Ubers. Commenters debate how “autonomous” the system really is given remote human assistance, whether lidar- and HD-map–based fleets like Waymo can scale economically, and how they compare to Tesla’s camera-only approach. Broader themes include regulatory hurdles, impacts on jobs and public transit, privacy and cleanliness concerns, and whether robotaxis should become core urban infrastructure or remain a niche premium service.

User Experience & Perceived Safety

  • Many SF riders report Waymo drives “like a very cautious grandma”: smooth, defensive, low drama, good at complex 4‑way stops and construction.
  • Cyclists and pedestrians often feel safer around Waymos; cars are described as highly aware of bikes, scooters, wrong‑way cyclists, and pedestrians who might jaywalk.
  • A few commenters report near‑misses and at least some red‑light running; others counter with anecdotes of extreme caution (e.g., stopping on green for a potential jaywalker).
  • First rides feel “sci‑fi,” but quickly become boring — which several see as a sign of maturity.

Autonomy, Remote Assistance & Limits

  • Ongoing argument over whether this is truly “driverless” if remote supervisors exist.
    • One side: not autonomous if humans can intervene; remote staff limit scalability and make private ownership unrealistic.
    • Other side: remote “fleet response” gives high‑level guidance only; no one is driving via joystick, and there’s no human in the vehicle, so it’s L4 driverless.
  • Waymo still geofenced; in SF it avoids freeways for rider‑only service, uses preset pickup/dropoff spots and sometimes requires short walks.
  • Works at night and in rain but not snow; construction, emergency vehicles, and hand signals are explicitly handled but still a known source of edge cases.

Tech Stack: HD Maps, Sensors & “Cheating”

  • Waymo uses lidar, multiple sensors, and high‑definition maps.
  • Some see HD maps as “cheating” compared to human‑like driving with eyes only; others reply that companies are trying to ship a product, not win a purity contest.
  • Debate over scalability of HD maps: critics say constant updates are too costly; defenders say vehicles can auto‑update maps and that maps are a prior, not a crutch.

Economics, Cost & Scaling

  • Per‑vehicle hardware is estimated in the $150k–$300k range today; huge capex on depots, charging, cleaning, and maintenance staff.
  • Not currently cheaper than Uber/Lyft; sometimes slower due to conservative driving and legal stops.
  • Some think private ownership is unlikely; model is fleet robotaxis with depots. Others speculate about subscription remote‑assist for owned cars, but see cost as prohibitive.

Comparisons: Uber/Lyft, Taxis & Cleanliness

  • Many prefer Waymo to Uber/Lyft despite price/wait: no small talk, no driver body odor or heavy perfume, more predictable behavior, no risk of harassment.
  • Others value human drivers for “character,” conversation, and flexibility (e.g., creative pickups, luggage help), but also recount negative experiences (unsafe, sleepy, proselytizing, even assault).
  • Cleanliness: current cars are very clean; skepticism about long‑term standards once cost‑cutting starts. Interior cameras, depot cleaning, and the ability to bill/ban riders are seen as important tools.

Tesla FSD vs Waymo

  • Large sub‑thread argues whether Tesla can “overtake” Waymo:
    • Waymo today: operating L4 robotaxis in multiple cities, driverless within geofences, lidar + HD maps, heavy mapping and infrastructure overhead.
    • Tesla today: camera‑only supervised FSD that still requires constant driver attention and has no true driverless miles.
  • Pro‑Tesla side: rapid improvement with NN‑based FSD, huge existing fleet, low hardware cost, and upcoming GPU clusters; believe a “flip of a switch” could unleash millions of robotaxis once FSD crosses a threshold.
  • Skeptical side: Tesla has zero autonomous operation and still makes basic mistakes (stationary objects, walls); vision‑only seen as a harder path, with unclear route to L4/L5; regulatory distrust of Tesla’s culture is highlighted.

Regulation, Cities & Social Impact

  • Europe is seen as far behind due to stricter regulation; some EU commenters lament economic/tech “self‑sabotage,” others defend safety and privacy laws.
  • Some argue robotaxis should be integrated into subsidized public transport or combined with trains; others insist only transit can solve congestion, with robotaxis at best a complement.
  • Strong concern about job loss for taxi/Uber/delivery drivers and other low‑skill workers; calls for policy “off‑ramps” (e.g., safety net, healthcare decoupled from employment).
  • Counter‑argument: automation raises productivity; harm or benefit is a policy choice, not a technology inevitability.

Vandalism, Misuse & Surveillance

  • Worries about vandalism and “coning” protests; others think novelty will fade and ID‑tied accounts plus cameras will strongly deter abuse.
  • Debate on privacy: interior video monitoring and detailed trip‑tracking feel invasive to some, but others see them as necessary for safety, cleanliness, and billing.