AI agents used in interviews
A social post describes a candidate who reportedly used AI agents (Codex, ChatGPT 5.4) to build a HubSpot mini‑app in 10–15 minutes and then discussed agent constraints in the interview — prompting directors to value AI fluency and logic over manual typing. That anecdote signals interviewers may probe AI-agent design and governance as part of technical screens. (x.com)
OpenAI launched the Codex desktop app as a macOS “command center for agents” that runs multiple coding agents and parallel workflows, bringing a dedicated developer UI to agentic coding. (openai.com) OpenAI also released GPT‑5.4 mini and nano into the API, Codex and ChatGPT in mid‑March 2026, marketing the smaller variants as faster, agent‑friendly models for high‑volume coding and subagent tasks. (community.openai.com / 9to5mac.com) HubSpot’s AI stack (branded Breeze/Copilot/AI Agents) exposes agent and app integration points in its CRM, with HubSpot advertising quick setup for AI‑driven workflows and the platform reporting 288,000+ customers across 135+ countries. (hubspot.com / hubspot.com) Interview prep guides and hiring blogs published since 2025 list “agent design,” multi‑agent orchestration, tool‑calling, and safety/constraint design as explicit system‑design topics candidates should expect in technical screens. (datainterview.com / blog.promptlayer.com) Recruiting coverage shows firms adjusting assessments after widespread candidate AI use; Karat was cited by HR reporting as suspecting more than 80% of some candidate pools used AI tools, and outlets documented companies expanding policies and detection for AI on coding tests in 2025. (hrexecutive.com / time.com) Survey and HR playbook publications report hiring managers elevating practical AI fluency and creative problem solving over rote typing: a March 2026 Resume.org survey found roughly 6 in 10 hiring managers now prefer creative thinkers to pure coders, while firms like Zapier and assessment vendors publish rubrics for assessing “real” AI fluency and responsibility in interviews. (marketwatch.com / disco.co / testgorilla.com)
