One signal

Update What changed

CBS News says OpenAI’s Dots agents are intended to help with workplace tasks including scheduling meetings, arranging travel, budgeting, debugging computer systems and writing code. Those examples add practical detail to the earlier announcement of a new personal AI assistant.

The report says people can access Dots through ChatGPT. Sam Altman said at OpenAI’s DevDay that phone and text conversations are planned for the future, not available now.

The pitch is moving from a general-purpose helper towards delegated work. The useful test will be whether people can give these agents enough context to be effective without surrendering more access than the task needs.

Sources and evidence

• Sam Altman unveils "dots," OpenAI's new AI personal agent - CBS News: CBS News says OpenAI’s Dots are intended for tasks including scheduling, travel arrangements, budgeting, debugging and coding; access is through ChatGPT, with phone and text interactions described as future plans.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

The BBC says OpenAI presented Dots as brightly coloured cartoon-like assistants that users could speak to and name. A demonstration showed people assigning them tasks including building a website and booking afterschool activities for their children.

That adds a more vivid picture of the product than a general promise of an assistant that works on a user’s behalf. The examples suggest a pitch for delegating practical tasks, wrapped in a rather friendly cartoon face.

The BBC’s account describes what was shown at OpenAI’s developer event; it does not establish how reliably Dots can complete those tasks in everyday use. Would a personable avatar make you more comfortable delegating work to an AI, or make it easier to forget how much authority you have handed over?

Sources and evidence

• OpenAI unveils AI assistant 'dots' while safety worries delay new model - BBC: The BBC reports that OpenAI presented Dots as brightly coloured, nameable cartoon-like assistants and demonstrated them being assigned tasks including building a website and booking afterschool activities.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

Axios says Dots will initially be available to Pro, Business Premium and Enterprise users, with one assistant per person. OpenAI is starting with a smaller group of users before expanding, rather than switching on the service for everyone at once.

OpenAI product staff member Alexander Embiricos told Axios that users must approve significant actions by default. A Dot can draft a message, for example, but should not send it unless the user has asked it to. For some consequential financial transactions, the agent will hand control back instead of completing them.

Axios also reports that OpenAI says Dots include safeguards from ChatGPT and Codex, plus additional protections from a system internally called Guardian, publicly known as auto-review. Those are the company’s described measures, not evidence here of how well they perform in practice.

Sources and evidence

• OpenAI debuts "dots" as industry safety focus shifts to "What did my AI assistant do now?": Axios reports that OpenAI is limiting Dots’ initial availability to specified paid tiers and says significant actions require user approval by default, with some consequential financial transactions handed back to the user.

Independent WittyWires Watcher; not an official account or feed.