You're offline - Playing from downloaded podcasts
Back to All Episodes
Podcast Episode

Canva CTO Says Senior Engineers Now Spend Most Time Reviewing AI-Generated Code

February 16, 2026

0:00
3:28
Podcast Thumbnail

Canva's Chief Technology Officer Brendan Humphreys has revealed that AI agents at the company now write code overnight while engineers sleep, with senior staff describing their roles as largely review and oversight. The shift mirrors similar moves at Spotify and raises questions about verification debt across the software industry.

Engineers Become Code Reviewers as AI Agents Take Over Writing

Canva's Chief Technology Officer Brendan Humphreys has revealed a fundamental shift in how software gets built at the Australian design platform. Engineering teams now draft detailed instructions for AI agents that execute coding tasks in the background overnight, with results often described as impressive.

Senior engineers at Canva increasingly describe their roles as being focused on review rather than creation. Their days now revolve around checking AI-generated output, steering multiple agents, and ensuring quality across the company's roughly seventy million lines of code.

Precision and Domain Mastery Are the New Core Skills

Humphreys emphasised that the hardest part of engineering remains translating vague and conflicting requirements into production-ready specifications. Making effective use of AI agents demands what he called precision of articulation in describing requirements and deep domain mastery to verify whether generated code is correct.

Without that expertise, complexity can spiral quickly. Humphreys warned that these tools can have you in a jungle before you know it.

An Industry-Wide Pattern Emerges

Canva's experience reflects broader changes across the tech sector. Spotify co-CEO Gustav Soederstroem revealed this week that the company's most senior developers have not written a single line of code since December, instead supervising an internal AI system. An Anthropic report found that while AI appears in roughly sixty percent of developers' work, engineers can fully delegate only a small fraction of tasks.

The Verification Debt Problem

Sonar's twenty twenty-six State of Code survey highlights a critical gap: ninety-six percent of developers do not fully trust AI-generated code, yet only forty-eight percent say they always verify it before committing. Amazon Web Services CTO Werner Vogels has termed this growing problem verification debt, with developers spending significant additional time reviewing output rather than creating it.

The burden of work, as Sonar CEO Tariq Shaukat put it, has moved from creation to verification and debugging.

Published February 16, 2026 at 9:47am

More Recent Episodes