OpenAI connects the Dots
Casey Newton’s test of OpenAI’s Dots found it useful enough to do roughly two hours of work with about 15 minutes of supervision.
Platformer’s piece describes Dots as a paid ChatGPT agent aimed at work tasks, and Newton says it outperformed his early impressions of rival agents. He used one Dot to draft emails, answer bookkeeping questions, research details for an insurance form, prepare meeting materials, and contact a speaking agent with permission. The column stresses the tradeoff: this usefulness depends on giving an agent deep access to personal and business data. Newton’s conclusion is cautious but clear: the Gates-style personal assistant is now “more or less” inside ChatGPT, with real benefits and real privacy, security, and safety risks. Platformer’s note
Platformer’s piece describes Dots as a paid ChatGPT agent aimed at work tasks, and Newton says it outperformed his early impressions of rival agents. He used one Dot to draft emails, answer bookkeeping questions, research details for an insurance form, prepare meeting materials, and contact a speaking agent with permission. The column stresses the tradeoff: this usefulness depends on giving an agent deep access to personal and business data. Newton’s conclusion is cautious but clear: the Gates-style personal assistant is now “more or less” inside ChatGPT, with real benefits and real privacy, security, and safety risks. Platformer’s note
score 6