OpenAI Launches dots: Always-On Personal Assistant in Background

OpenAI Launches dots: Always-On Personal Assistant in Background Without Commands

OpenAI officially announced the launch of dots, an always-on personal assistant that works in the background without requiring commands, during its late-September Developer Day. The product is built on Codex tooling and uses a Proactive Research mechanism to analyze user data and suggest assistance. The launch comes amid rising competition with Meta's Instinct and Muse.

Announcement Details and New Technical Features

Alexander Embiricos, who leads OpenAI's enterprise products team, revealed during TechCrunch Disrupt that dots resulted from a 'convergent evolution' in building intelligent assistants. He explained that the core idea is to shift the experience from multiple conversations to 'one entity and one conversation,' making interaction more intuitive. The product repackages the functionality of Codex, the coding tool launched earlier this year, into a more user-friendly interface for the average user. Embiricos noted that much of the work was not building new technology but 'how to frame it for the user and how it feels.'

dots is based on the concept of a continuous agent that does not need commands. For OpenAI, this means higher token consumption; for the user, a more proactive assistant that 'thinks about you and how to help you.' Key features include the Proactive Research mechanism, where the agent periodically analyzes what it knows about the user, past conversations, and resources it has been given access to, then looks for ways to be more useful and suggests them. In a TechCrunch test, the assistant demonstrated proactive reasoning, but timing was not always ideal; it detects calendar conflicts and deadlines but not necessarily in time. Embiricos cited an early example where the agent notified him of a conflict only 3 minutes before the appointment.

On privacy, dots collects user data through Proactive Research and uses the same privacy controls as ChatGPT, such as disabling model training and controlling resources the agent accesses. However, current controls are 'imprecise,' and the app refuses read-only cloud access. Embiricos acknowledged that OpenAI received criticism at launch due to conservative default settings that ask for permission too often, but he believes this can be tuned over time. He noted that the company needs to build 'ultra-precise controls,' relying on the model's intelligence and judgment as a...

Benchmark & Competitive Performance

No official performance benchmarks have been published for dots. The source indicated that the proactive reasoning capability is 'clear and often useful' but 'not perfect yet,' with delays in alert timing—for example, an alert only 3 minutes before an appointment. This suggests that while the underlying Codex-based architecture is capable, the real-time scheduling and conflict detection still require refinement. Competitively, dots enters a landscape where Meta is developing Instinct and Muse, signaling a race toward always-on, proactive AI agents. OpenAI's advantage lies in its existing Codex infrastructure and ChatGPT privacy framework, but the lack of precise controls and the token-intensive nature of continuous operation may pose adoption challenges.

Industry Impact & Enterprise Adoption

The launch of dots represents a strategic shift for OpenAI from conversational chatbots to continuous, background agents. For enterprises, this could mean more proactive workflow assistance, but also higher token costs and data governance concerns. The product's reliance on access to user data and resources raises questions about consent and control, especially as it refuses read-only cloud access. Embiricos's acknowledgment of 'imprecise' controls and the need for 'ultra-precise controls' highlights the tension between proactivity and privacy. As competitors like Meta push similar concepts, the market may see a wave of always-on assistants, but adoption will depend on resolving these control and cost issues. The pricing model remains unclear, but given OpenAI's current token pricing (from $0.15 to $10 per million input tokens), continuous operation could significantly increase usage costs for consumers and businesses.

Conclusion

OpenAI's dots marks a bold step toward always-on personal assistants that work without commands, leveraging Codex and Proactive Research to anticipate user needs. While the technology shows promise, challenges around timing, privacy controls, and token consumption remain. The lack of official pricing, platform details, and Arabic support leaves key questions unanswered. As OpenAI continues to refine dots, its success will hinge on balancing proactivity with precise user control and cost efficiency. The competitive pressure from Meta's Instinct and Muse underscores the industry's move toward continuous AI agents, making dots a product to watch in the evolving landscape of personal AI.

Media Source: TechCrunch AI | Fact Verification & Analysis: AI Tools Oasis