OpenAI Dot stress-tested with a hiring task

added Oct 3, 2026

I Gave an OpenAI Dot a Hiring Task, Then Tried to Trip It Up I gave an OpenAI dot 4 fake intern applications to organize. Then I changed the priority, added 2 more, and slipped in a duplicate. Could it keep it straight? Quick context: dots are OpenAI's always-on agents. Each has its own cloud computer and can work through ChatGPT Work or Codex. The test was all fictional. Rules: organize the evidence, keep gaps visible, leave decisions to me. And it got interesting fast. Applicant two claims Python, but the project was a browser interface with no Python sample. The dot kept both facts visible and marked the evidence unknown. It did not turn a claim into proof, and an absent sample just tells you what to ask for. Changed priority to project-first: it reordered the columns, changed nobody's eligibility. Added 2 more plus a duplicate: got 6 unique applicants, not 7. Asked for an interview draft: 5 project-specific questions, left as a draft. It did not pick the applicant. I did. Honest part: one staged session, fake data, no comparison to normal chat. It organized the material. The decisions stay mine. Full breakdown: https://openai.com/index/introducing-dots/

The LearnOpenCV creator gave his OpenAI Dot 4 fake intern applications to organize, then changed the priority, added 2 more, and slipped in a duplicate to trip it up. The Dot kept evidence visible (never turned a claim into proof), reordered its columns when priority changed, deduped to 6 unique applicants, and drafted 5 project-specific interview questions — leaving the final decisions to the creator. Disclosure: the applications were fictional per the creator's caption; the test was of the Dot's organizing behavior.