When AI writes the code, review becomes the bottleneck | Rahul Sanghavi, Twilio Inc.

When most of your code is written by AI, the job shifts from writing to reviewing, and reviewing is where the work piles up. Rahul Sanghavi, Director of Engineering at Twilio, describes one engineer who generated 60 to 70 PRs in a few days, more than anyone could review.

Recorded with Denis Scott at ELC Annual 2026, the conversation covers how his team uses tests as the gate for shipping safely, how AI now runs inside their on-call rotation, and how they are capping per-engineer spend as AI costs climb.

Topics discussed:

• Shifting from writing code to reviewing code

• Code review as the new engineering bottleneck

• Using tests as the gate for shipping AI-written code

• Writing tests deep enough to catch edge cases

• Running AI inside on-call to debug and mitigate incidents

• Setting per-engineer AI spend limits as costs climb

• Planning 2027 AI budgets after a free-hand Q1

• Weighing in-house LLMs to control rising AI spend

  
  

Listen to more episodes: 

Apple 

Spotify 

YouTube