I Gave OpenAI Dots 20 Real-World Tasks on My Own MacBook — It Completed 19

Workflow Verdict@WorkflowVerdictMedia

added Oct 3, 2026

I gave both AI agents the exact same M3 Max MacBook Pro, the exact same 20 real-world tasks, and one rule: zero human help unless they completely freeze.

First-hand head-to-head: the creator gave OpenAI Dots (and Meta Muse) the exact same 20 real-world tasks on his own Apple M3 Max MacBook Pro, zero human help unless an agent completely froze. His Dot completed 19/20 tasks (95%), with 2 wrong actions, 3 human interventions, 54.4 min total runtime, $24.73 compute cost. Tasks spanned Mail, Calendar, Finder, Numbers, Terminal, and Safari; per-task outcomes and failure recovery (3/5 recoveries) are shown on screen. This is a benchmark/comparison video, but the Dot's concrete completed task results are demonstrated first-hand, not impression-based commentary.