Walking Devs 001 Interviewing in the Age of AI
Feb 10, 2026
They unpack surprise model releases and the rush of simultaneous launches. They test a new Codex app and debate sandboxing and orchestration for agents. They compare strategies for running multiple agents and tiny, fast coding assistants. They argue about using AI during interviews, redesigning interview formats, and flaws in traditional hiring practices.
AI Snips
Chapters
Transcript
Episode notes
Benchmark Arms Race Shapes Releases
- Model release cadence now focuses on marketing comparisons rather than apples-to-apples benchmarks.
- Rapid back-to-back launches push labs to rush deployments and create noisy benchmark narratives.
Uncomfortable With Codex App UI
- Nikolas tried the Codex app and disliked its opaque sandboxing and UI compared to a terminal workflow.
- He felt uncomfortable not being able to inspect running tasks or easily interrupt processes with Ctrl-Z.
Sandbox Agents To Prevent Interference
- Use sandboxes for agents to avoid long-running, uncontrolled processes touching your local environment.
- Consider remote sandboxes or copy-on-write repos to isolate concurrent agent changes safely.