OpenAI rebuilt its coding tools into dots, an always-on AI assistant

From coding tools to assistant

At its developer day in late September, OpenAI released dots, an always-on personal assistant that keeps working in the background even when nobody is prompting it. Speaking to TechCrunch at the outlet's Disrupt conference this month, Alexander Embiricos, who leads OpenAI's enterprise product team, explained that the product is largely a repackaging of capabilities the company had already built for a very different audience: software engineers.

According to TechCrunch, OpenAI's Codex coding app — released at the start of this year — could already accomplish much of what dots does, but only with considerable hand-holding. The company had also previously spun ChatGPT Work out of the agentic coding tools it built for developers. dots takes that machinery and reframes it as a single persistent entity in one continuous conversation, rather than many separate chats, which Embiricos said makes the product feel intuitive.

What dots does

dots lives inside the ChatGPT desktop and mobile apps and requires a Pro-tier subscription or higher, with plans starting at $100 a month, TechCrunch reports. Once connected to a user's calendar, email, computer and other tools, it tries to anticipate what needs doing.

A feature Embiricos calls "proactive research" runs on a cadence: the assistant reviews what it knows about the user, their past conversations and the resources it has been granted, then goes off to research how it could be more helpful and proposes actions.

In practice, the anticipation is uneven. Embiricos recounted that during early development his dot warned him about a calendar conflict three minutes before it began. But he also relayed a story from an engineer at a drone delivery startup whose dot noticed an outage before the human did, analyzed it, and arrived with useful information for resolving it.

Even the name is a product decision — a reference to the negative space inside OpenAI's logo, chosen to feel like something that could eventually be ubiquitous, according to Embiricos. He added that he learned only late that rival entrepreneur Elon Musk owns the dot.com web address.

Privacy trade-offs

According to TechCrunch, dots carries the same privacy controls as ChatGPT itself: users can turn off model training on their conversations and manage what the agent can access. But those controls are described as not particularly granular, and the app can be pushy — it reportedly refuses to accept read-only access to cloud drives.

Embiricos acknowledged the gap, saying the company needs to build "super, super, super granular controls," while pointing to the model's intelligence and discretion as the other side of the coin. He also said OpenAI shipped with deliberately conservative defaults that asked permission frequently, expecting the criticism and planning to tune it over time.

The competitive backdrop sharpens the question. Meta's Muse assistant has drawn criticism for the profiles it builds of users' family and friends, and the startup Hark markets its own assistant on the claim that, unlike OpenAI and Meta, it does not collect data to fuel an advertising business. dots sits in the same always-on wave as Muse and Instinct, the startup TechCrunch describes as an overnight sensation in the category this year.

Can agents scale

The other open question is whether agent interactions can scale into the wider world. Embiricos told TechCrunch he keeps a human assistant because OpenAI's corporate partners want to deal with a person, adding that he personally never intends to have a meeting with someone else's agent. He also said OpenAI is not always the fastest to ship new ways for agents to act on the world, but that it is deep in conversations with payment processors — a step that would let an assistant spend money, and a significant trust ask for consumers.

Why it matters

dots is a case study in how AI products are increasingly won through packaging rather than raw model capability. OpenAI built agentic tooling for programmers, then reframed that same machinery as an intuitive, persistent assistant for a mass audience — and is charging from $100 a month for it. Whether the strategy pays off hinges on two unresolved problems: privacy controls granular enough for users to trust with their inboxes and files, and a world of businesses, websites and payment networks willing to transact with software acting on people's behalf. How those land will determine whether the always-on assistant becomes the default interface for AI, or this cycle's most over-promised category.