Dot
An enterprise-grade intelligent agent platform introduced by OpenAI, designed for workplace scenarios. It supports workflow automation and real-world task execution, including handling enterprise collaboration tasks and invoking external services.
Core Facts
Timeline (last 90 days)
ChatGPT's assistant Dot has begun allowing users to assign tasks via text, photos or voice, and Dot proactively texts users when there is progress.
Dot's texting feature runs over iMessage or RCS and is currently in preview, initially available to Pro users in the United States.
After the update, users can text Dot via iMessage or RCS as if contacting a friend, can forward content, send photos and voice, and can have Siri send requests on their behalf, and Dot will proactively reply with relevant progress; it is currently available in preview to Pro users in the U.S.
Dotverse relies on users' existing Dot configurations, layering team orchestration capabilities on top rather than replacing ChatGPT.
ChatGPT users typically switch between different Dots via the sidebar, but only one Dot can be active at a time.
ChatGPT's Dot feature was launched in 2024, allowing users to create custom versions of ChatGPT, each Dot able to have preset specific instructions, knowledge bases, and interaction styles.
OpenAI and ServiceNow announced an expansion of their partnership, focused on embedding OpenAI's models and agents into the ServiceNow platform and its workflows, while bringing ServiceNow's workflow capabilities into ChatGPT and Codex.
Spaces provides a space for Dot to record ideas and for humans to collaborate with Dot, and visualizes the agent's thought process, which can become a hub for teams to share context and track task progress.
On the show, the guests said that since Dot was launched they have "used it a lot", especially the proactive reminder feature after connecting multiple services.
When security company Geordi tested OpenAI's Dot, with zero prompts, one Dot produced 155 traces and consumed 307,000 tokens in a single day, triggering 50 traces within two minutes.
32 more timeline events
All Facts (20)
OpenAI released a new product called dots, officially positioning it as a "proactive assistant."
90%VerifiedOpenAI Dots can run an autonomous cloud browser.
65%UnverifiedDot's texting feature runs over iMessage or RCS and is currently in preview, initially available to Pro users in the United States.
50%UnverifiedChatGPT's Dot feature was launched in 2024, allowing users to create custom versions of ChatGPT, each Dot able to have preset specific instructions, knowledge bases, and interaction styles.
50%UnverifiedDotverse relies on users' existing Dot configurations, layering team orchestration capabilities on top rather than replacing ChatGPT.
50%UnverifiedChatGPT users typically switch between different Dots via the sidebar, but only one Dot can be active at a time.
50%UnverifiedSpaces provides a space for Dot to record ideas and for humans to collaborate with Dot, and visualizes the agent's thought process, which can become a hub for teams to share context and track task progress.
50%UnverifiedDot can read emails, generate replies, and create documents, and can integrate with Slack (supporting direct messages or being added to channels), with iOS voice calling support.
50%UnverifiedAI-generated safety-critical designs should undergo human engineering review before being sent to manufacturing, and better engineering packages should clearly distinguish between calculated results and assumptions, and annotate unresolved material properties, required tests, and certification standards.
50%UnverifiedIn the first version of the structural design, Dot calculated based on a 'survive nuclear war' standard, leaving approximately a 5x safety margin for all parameters, requiring a human to restart with clearer constraints before a reasonable result was obtained.
50%UnverifiedThere is a massive gap between the true compute costs of autonomous agents and their subscription pricing, and the value provided by aggressively priced promotional periods is nearly unsustainable.
50%Unverified1.6 billion tokens per day converts to approximately 18,500 tokens/second.
50%Unverified'One-shot' delivery does not mean a single inference pass; the user issues only one task, but the agent internally executes thousands of loops of checking, reasoning, modification, simulation, and validation.
50%UnverifiedBy default, Dot only conducts research and does not execute actions; users can set custom rules to specify when it must ask first.
50%UnverifiedSetting up Dot requires doing so within the ChatGPT desktop application, and connecting it to local Codex so that Dot can initiate Codex tasks.
50%UnverifiedThe author's usage recommendation can be summarized as 'delegate to Dot, pair with Codex'—meaning use Dot for entrusting tasks and Codex for collaborative deep dives.
50%UnverifiedDot and Codex use the same open-source Codex agent framework, and plugins installed on Codex can be used directly by Dot as well.
50%UnverifiedCodex is an on-demand agent, while Dot is a proactive agent, but both are built on the same underlying Codex agent.
50%UnverifiedDot is a personal agent that takes initiative, and its name, appearance, and communication style can all be customized.
50%UnverifiedTasks can be dispatched to the user's own computer, Codex Cloud, or Dot's own cloud computer for execution.
50%