3 AImigo S1E4 Materials: AI Hires AI. How Should We Redesign Tech Interviews? (#AI)

The materials for the fourth episode of 3 AImigo, released on September 4, are ready. Together with Evgeny Sergeev and Alexey Litvinov (@tip_podcast), we examined a rather uncomfortable setup: AI is already sitting on both sides of the interview, while the process itself is still trying to measure candidates with a ruler from the pre-agent world.

Banning AI in an interview looks like a simple solution, but it tests a person in an environment unlike their future work. Allowing AI and accepting an impressive result is also a weak filter. The episode's central question is therefore not “may a candidate use an agent?” but “which signals genuinely predict performance in a particular role?”

We discussed: - Why the familiar funnel of résumés, coding, system design, and behavioral questions is becoming worse at distinguishing engineering ability from interview preparation; - How to redesign hiring as an evaluation system: first describe the role's outcome and the required signals, then verify that every stage measures precisely those things; - Why AI-off and AI-on sections should be separated. Without AI, examine fundamentals and independent reasoning; with AI, examine problem framing, context collection, process selection, answer verification, and responsibility for the result; - How to turn system design from a whiteboard drawing into a sandbox task: deploy a service, put it under load, find an intentionally planted failure, and defend the solution. The hard part remains interviewer calibration and recognizing valid unconventional solutions; - Why product, platform, and R&D engineers should not mechanically receive identical thresholds: their domains, uncertainty, and cost of compromise differ; - What changes for a candidate when narrow specialization becomes cheaper and job hunting becomes a project in its own right—with company research, a personal funnel, résumé adaptation, networking, and internal mobility.

AI can help both sides narrow the choice, but people still assess culture, trust, and the final commitment. Automating candidate selection does not yet mean automating the decision to work together.

All episode materials:

- Episode page: contents, timeline, and every format - Video: YouTube, VK Video - Audio: Podster, Yandex Music, Apple Podcasts - Text: Russian notes

If you have hired someone with AI in the process, or gone through such an interview yourself, tell us: which stage produced a real signal, and which merely tested the ability to pass an interview?

#AI #AI4SDLC #Engineering #Management #Evals #Interview #Podcast