I Gave OpenAI's AI Agent 2 Real Tasks. Here's What Happened
added
I gave OpenAI's new always-on agent two real tasks. Task one: find AI videos beating their channel's average — it took three and a half minutes and warned it couldn't verify live view counts. Task two: build a morning briefing — I checked every story, two were right, one overstated who gets access.
First-hand YouTube Short: The AI Frontier gave his Dot two real tasks and audited the results. Real handle: The AI Frontier. Task 1 — find AI videos beating their channel average: the Dot took 3.5 minutes and honestly warned it couldn't verify live view counts. Task 2 — build a morning briefing: he checked every story — two were right, one overstated who gets access (Enterprise is beta, admin must turn on). Fast and honest about its limits. Concrete completed results with the creator's own verification.


