← All Episodes
Based on Lenny's Podcast data
Lenny's Knowledge Sketch

Building Trust in Self-Driving
Cars at Waymo

Shweta Shrivastava
Senior Director of Product, Waymo
FEATURED
The Challenge

Fully Autonomous:
No Driver, Full Trust

WAYMO vs TESLA
"We started by solving the problem of fully autonomous driving without a human driver at the wheel from the get-go. It's not a driver assist system that relies on the human taking over in complex situations."
  • L4 autonomy: fully autonomous without human at wheel, no expectation of takeover
  • Not driver-assist like Tesla; no mental load on rider to stay ready to take control
  • Built for complete delegation of trust, not conditional assistance
  • Every interaction designed to feel natural, predictable, and credible
Design Philosophy

Building Credible
Autonomous Behavior

HUMAN-LIKENATURAL FEELTRUSTin rider
5min
until feels natural
then uneventful
intended design
  • Train on human driving data: use real driving behavior, discard bad habits
  • Deep learning for intent: read pedestrians' body language, gestures, eye direction
  • Learn social norms: understand local street cultures (SF jaywalking vs other cities)
  • Small design details: slowing on slopes, respecting speed limits, predictable lane changes
The paradoxDon't realize how sophisticated, interactive, and social driving really is until you have to teach a machine to do it naturally.
Measuring Progress

KPIs That Matter:
Safety & Scaling

  • Commercial metrics: trips per week, daily/weekly active users, funnel conversion
  • Operational cost: cost-per-ride to operate the fleet sustainably
  • Safety benchmark: collisions per 100k miles vs human driver baseline
  • Progress & assertiveness: unduly stops, rescue assists, traffic impact on others
  • Compliance: adherence to road rules, speed limits, local traffic patterns
The benchmark

Drive safer than humans. No single human benchmark exists, so Waymo gathers data on what human drivers actually do—collisions, patterns—and compares against it.

The balance

Safety if you never move. Progress if you're reckless. The goal: assertive enough to get riders to destination on time, but safe at all times.

Long Game

Keeping Leaders
Bought In

  • Show meaningful progress: not just tech breakthroughs, but commercial deployment working
  • Accelerate milestones beyond expectations—beat your own forecasts
  • Focus on customer value first, investors second—build a real business
  • Don't chase short-term brownie points; investors see through it
  • Let results speak for themselves; progress = confidence
Alphabet's advantageBacking a moonshot like Waymo requires patience, but it also requires showing momentum. Hard to cut something that's winning.
Contrarian

What PMs Get Wrong About Autonomous Vehicles

L5 autonomy is the goalINSTEAD →L4 in structured environments solves the problem. L5 (off-road, no maps, unstructured) may stay niche.
MVP means cutting safetyINSTEAD →MVP bar for safety is extremely high. Can't iterate corners on trust—safety is non-negotiable.
More features = better productINSTEAD →Natural, predictable, boring driving builds trust. Fancy behavior erodes confidence in the system.
Tech breakthroughs drive adoptionINSTEAD →Commercial deployment and real riders prove the tech works. Results matter more than announcements.
𝕏︎ X / Twitterin LinkedIn📸 Instagram🔗 Copy link
0:00