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)
Dot will soon support iMessage interaction, but this feature has not yet launched.
Dot 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.
AI-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.
A Reddit user reported that a single Dot agent they ran consumed approximately 1.6 billion Astra tokens per day, equivalent to approximately 48 billion tokens per month.
1.6 billion tokens per day converts to approximately 18,500 tokens/second.
An engineering design package generated by Dot contains 1,000+ individual components, 6,000+ entities in a complete STEP assembly, and nearly 300 pages of assembly drawings and documentation.
Dot performed multi-level optimization on a real-time 3D combine harvester model, progressively reducing approximately 200,000 active triangle faces to approximately 152,000, 84,000, 56,000, and 50,000.
'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.
The author roughly estimated that equivalent work via API calls would cost approximately $15,000 to $20,000 per day, approximately $450,000 to $600,000 per month, with a median of approximately $540,000.
According to the author's recollection, Tibo once posted that Dot's economic model was a promotional period lasting one month, and the author suspects that the quota or pricing will be significantly adjusted.
21 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 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%Unverified1.6 billion tokens per day converts to approximately 18,500 tokens/second.
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%UnverifiedAccording to the author's recollection, Tibo once posted that Dot's economic model was a promotional period lasting one month, and the author suspects that the quota or pricing will be significantly adjusted.
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%UnverifiedThe author roughly estimated that equivalent work via API calls would cost approximately $15,000 to $20,000 per day, approximately $450,000 to $600,000 per month, with a median of approximately $540,000.
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%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%UnverifiedDot's cloud tasks continue running after the user closes their laptop, while local tasks require the computer to remain on.
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%UnverifiedCurrently, Dot does not consume ChatGPT usage quota and is nearly unlimited, unless it delegates tasks to local Codex or Codex Cloud.
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%UnverifiedDot is a personal agent that takes initiative, and its name, appearance, and communication style can all be customized.
50%UnverifiedCodex is an on-demand agent, while Dot is a proactive agent, but both are built on the same underlying Codex agent.
50%UnverifiedTasks can be dispatched to the user's own computer, Codex Cloud, or Dot's own cloud computer for execution.
50%UnverifiedCurrently, each person can only create one Dot.
50%