Should you let interviewees use AI? A chat with the CTO of Convex

Описание: James, the CTO of Convex, breaks down what hiring looks like when AI tools are everywhere and why Convex still optimizes interviews to reveal how candidates think, not how fast they can prompt. From his MIT PhD in distributed systems (and lessons on abstractions from Barbara Liskov) to scaling Dropbox infrastructure, he explains how culture, values alignment, and clarity of thought shape engineering teams.

The conversation covers Convex’s hiring loop (phone screens, onsites, architecture and deep-dive interviews, and a decision-centric debrief), why rubric-driven scoring fails in high-bar environments, and why “let them use AI in the interview” can add noise when you’re trying to evaluate fundamentals. On the workplace side, he argues against AI mandates: engineers are accountable for output quality, whether work was written by hand or assisted by tools, and strong teams don’t need policies to enforce productivity.

Timestamps

[00:28] Should developers use AI, and should candidates get it in interviews?
[01:32] MIT PhD, Barbara Liskov, and why abstractions matter
[03:09] Dropbox infra at scale: exabytes, transactions, high ownership
[07:04] Culture is decision-making, not perks: values, “why,” integrity
[13:54] Conflict resolution: stop the meeting and go “up the stack”
[20:26] Convex hiring loop: phones, onsite, coding, architecture, deep dive
[21:08] No scorecards: debriefs force real decisions and train taste
[27:25] AI in interviews adds noise: the Michelin-chef “chopping machine”
[33:12] AI at work: no mandate, just accountability for output
[39:18] Advice for juniors: love training, build daily, contribute to OSS

Hashtags

#softwareengineering #hiring #engineeringleadership #aidevelopment #systemdesign #architecture #distributedSystems #typescript #developerproductivity #startupculture #interviews #management