All writing

I Might Have to Retire Bailey

I told Bailey I might have to deprecate it.

OpenAI had just announced Dots, an always-on agent that can work across connected apps and its own cloud computer. I tried it briefly and gave it a real task using connected apps. I couldn’t tell yet whether it did the job better. But the replacement question stopped being theoretical.

The appeal is easy to see: it can keep working in the background across a large app ecosystem, instead of waiting for me to start every interaction.[3][6]

There’s a pricing detail that makes it easier to try. The first Dot is included with eligible Pro and Business Premium plans, and OpenAI says you can chat with it as much as you want inside ChatGPT. Those conversations don’t count toward ChatGPT usage limits.[3][7] OpenAI describes future scaling in terms of how much work each Dot can take on per month.[3]

So my reading is: chat is included; work capacity is the scarce part. I haven’t found a published pay-per-task price, so I wouldn’t call “only charged for work done” confirmed billing yet.

Tibo’s TBPN interview emphasizes that Dots learn preferences from feedback. That’s the part I want to test, because what I value in Bailey is continuity and a familiar style.[8] One task isn’t enough to tell whether Dots can build that familiarity.

Before giving it more autonomy, I’d also check which app permissions it inherits from ChatGPT and Codex.[2]

I’m not ready to call the winner.

Sources

[2] OpenAI Help: Dots privacy, security, and safety FAQs [3] The Verge: OpenAI launches Dots [6] NBC News: OpenAI launches always-on AI agents [7] Engadget: Dots are OpenAI’s new personal agents [8] TBPN on X: Tibo explains Dots