To Invent Waymo, They Had to Reinvent PM

In this episode, Saswat Panigrahi, Waymo's CPO, joins me for the third conversation in the "Inside PM" series. Saswat joined Waymo in 2016, back when it was still Google's self-driving car project and the entire product team was fewer than five people, and has spent all ten years since on the same product. In this episode, we discuss how his product org had to invent its own way of measuring good when there was no dashboard to check, why he hires and promotes for competence, courage, and compassion, and what it takes to be an executive builder at Waymo.

–

Key topics:

• Cutting severe injuries "sixteen-fold"

• Why measuring good driving became its own product, staffed by dedicated eval PMs

• Why specifying what good looks like is AI's hardest problem

• Injecting rain and fog into simulated scenes instead of waiting a decade for the real thing

• The living room pitch to Waymo's first driverless riders

• Avoiding "Brownian motion": channeling vision into tracks a team can actually execute

• Spotting curiosity and grit in candidates in the age of AI

• Training to the point of failure

–

Timestamps:

00:00 Introduction

03:25 Joining Google's self-driving car project before it was Waymo

07:15 Why "feeling the pain" of failures builds real product leadership

08:56 Motivating a team a decade away from a real product

14:38 Was there ever a moment you thought this wouldn't work

16:15 Why measuring "good" driving became its own product, staffed by dedicated eval PMs

18:55 The living room pitch to Waymo's first driverless riders

21:11 Confronting the 40,000 deaths US roads accept every year

23:46 Cutting severe injuries "sixteen-fold"

27:59 Training to the point of failure

33:27 Spotting curiosity and grit in candidates in the age of AI

37:14 Integrity, and sleeping at night over what you approved

41:02 A tree of metrics, not a union list of 37 dashboards

47:14 Competence, courage, and compassion: what builds followership

53:53 What AI has democratized for product discovery and research

57:41 Why specifying "what good looks like" is AI's hardest problem

–

Brought to you by:

• Firecrawl—the web data API for AI agents. Use code THESKIP for 10K free credits

• Jira—Do the right work with Jira

–

Referenced:

• AMD

• Edwin Markham's poem on preparedness

• Google X

• Inside PM at Meta

• Inside PM at Stripe

• MMLU

• NASA

• Piccadilly Circus

• Transformers

• Waymo

–

Where to find Nikhyl:

• Twitter/X

• LinkedIn

–

Where to find Saswat:

• Twitter/X

• LinkedIn

–

Join The Skip:

• Skip Coach

• Skip Community

–

Find The Skip:

• Website

• Substack

• YouTube

• Spotify

• Apple Podcasts

–

Don't forget to subscribe to The Skip to hear me coach you through timely career lessons. Access exclusive sessions from 100+ top product leaders at skip.coach. If you're interested in joining me on a future call, send me a note on LinkedIn, Threads, or Twitter. You can also email me at [email protected]

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit theskip.substack.com