OpenAI announced dots on September 29: assistants that work on a cloud computer and use connected apps. The rollout starts on selected paid plans, not every ChatGPT account. OpenAI and 9to5Mac’s event coverage describe the launch.
The useful part is obvious: less time chasing unfinished work. Our suggested first test is a comparison of three repair quotes, using copies with personal details removed. Ask for differences, exclusions and questions to ask the contractors. Check the result yourself. That is a proposed test, not a capability UOD has independently validated.
The dangerous part is confusing permission to read with permission to act. OpenAI says unsolicited background research uses read-only tools; assigned tasks can do more, subject to permissions and approval rules. Your computer is separate unless you connect it. Those are company descriptions, not a guarantee against mistakes.
UOD judgment: potentially useful. Start with one bounded job, not your whole life. Require review before anything is sent, purchased or changed. Inspect actual app permissions as well as your written instructions. Keep financial accounts, private family conversations and sensitive records out of the first experiment.
We have not tested dots. A product announcement tells you what the company offers, not how reliably it will handle your household.