OpenAI Dots + Suggix: From Handling Feedback to Actually Fixing Problems

Yesterday, September 29, 2026, OpenAI officially introduced Dots, a new kind of AI assistant.
In simple terms, Dots is quite different from the Codex you may already be using.
Codex is more like an AI developer living on your own computer. When your computer is offline, it can't keep working. Dots, on the other hand, is more like a 24/7 personal assistant living on a cloud computer — closer to the idea of an AI employee.
You can give it a long-running task and walk away from your computer. Dots can continue working in the background until it has a result or needs you to make a decision.

1. How Does Dots Work?
   Dots isn't just a chatbot. It's more like an AI Agent that can work on tasks over an extended period of time.
   Where can you give Dots instructions?
   The most direct way is to talk to Dots inside ChatGPT. You can assign tasks through the ChatGPT desktop app or the web, and where supported, continue checking in and responding from your phone.
   Dots can also connect to communication channels such as Slack and Microsoft Teams. For example, you can @Dots directly in Slack and ask it to perform a task or report its progress. OpenAI also supports interacting with Dots through voice.
   This means you don't necessarily have to sit in front of your computer to give it instructions.
   For example, you could simply tell it:
   "Check customer feedback from the past week and identify recurring bugs. If you find a clear, small issue, have Codex try to fix it and run the tests."
   Then you can walk away.
   What can Dots do after receiving an instruction?
   Dots has its own cloud computer, browser, files, and working environment. It can access applications you've authorized, such as Gmail, Google Drive, and GitHub, as well as other connected tools.
   For development tasks, Dots can also create and manage Codex Cloud tasks and work inside preconfigured repositories and development environments.
   It can:
   Read code and related files
   Investigate GitHub Issues
   Modify code
   Run tests
   Analyze test results
   Create Pull Requests
   Hand the results back to you for review
   It can even work on multiple tasks at the same time and continue running them in the background. You can check its progress, provide additional instructions, or step in when it reaches a decision that requires your input.
   When it needs to interact with a regular website or graphical interface, Dots can also use its own cloud browser and computer instead of relying only on APIs.
   So a typical Dots workflow looks more like:
   Set a goal → Dots gathers information → Breaks the task down → Uses tools → Works in the background → Pauses for key decisions → You approve → Continues → Delivers the result
   That's one of the biggest differences from a traditional ChatGPT conversation:
   You're not giving it a question. You're giving it a piece of work that can continue executing.

2. Which Plan Can Use Dots?
   At the moment, Dots requires ChatGPT Pro.
   According to OpenAI's documentation, Pro or Business Premium users can get their first Dots without paying an additional fee.
   If you're on the Free Plan, you'll currently see:
   You’re on a Free plan. Upgrade to Pro 200 to access dots.
   So if you want to try Dots, you'll need to upgrade to Pro at least.

3. OpenAI's Own Example Is Actually Quite Interesting
   OpenAI gave an example when introducing Dots that is especially relevant to developers.
   Dots monitors customer feedback for recurring requests, determines the scope of small improvements and bug fixes, handles the development and testing, and then hands the completed PR to a developer for review.
   That's worth paying attention to.
   Because this is no longer simply:
   "AI helps you write code."
   It's:
   "AI starts with user feedback and goes all the way to code changes and testing."
   And there's one important piece missing in that workflow:
   A system that can continuously provide real user feedback.
   That's where Suggix can fit in.

4. What Happens When You Combine Suggix + Dots?
   Suggix is built around the workflow of:
   User Feedback → Product Decisions → Development → Release
   For example, one user might submit:
   "I'd like to see GitHub Issues synchronization."
   Another user might submit a very similar request.
   Suggix can bring these pieces of feedback together, helping you see which problems keep coming up, which requests receive more votes, and which issues may deserve attention.
   Dots can then take over the work that comes afterward.
   An ideal workflow could look like:
   User Feedback → Suggix → Dots analyzes feedback → Determines priority → Creates development task → Codex modifies code → Runs tests → Creates PR → You review → Release
   At this point, Suggix is no longer just a tool for collecting feature requests.
   It can become an entry point for AI Agents to access real user needs.

5. Suggix MCP Can Connect These Two Worlds
   Suggix already provides MCP, allowing MCP-compatible AI tools to directly read and interact with feedback stored in Suggix.
   For example, an AI can:
   Query recent user feedback
   Get the full details of a specific request
   Create new feedback
   Update feedback status, priority, and other fields
   Continue working on tasks based on that feedback
   This means Dots doesn't need to rely on copy-and-paste to understand what users are asking for.
   It can use Suggix directly as its product feedback data source.
   You could give Dots a long-running task like:
   "Continuously monitor Suggix for recurring bugs and small feature requests. For clear, low-risk issues, have Codex implement the fix and run the tests. When it's done, create a PR and notify me for review."
   From there, each component has a clear role:
   Dots coordinates the work.Suggix provides real user needs.Codex writes the code.
   Humans only need to make decisions at key points.
   This may be one of the most interesting things to explore with Dots + Suggix.

6. From Feedback Management to an AI Product Agent
   The traditional software development workflow usually looks like:
   User → Feedback → Product Manager → Developer → Code → Release
   With AI Agents, this chain is getting shorter:
   User → Feedback → AI Agent → Code → Test → PR → Human Review
   Suggix can become the Feedback Layer in this workflow.
   Dots keeps the work moving.
   Codex writes the code.
   And Suggix tells the AI:
   What do users actually want?
   Of course, this doesn't mean that every product decision should eventually be delegated to AI.
   People should still decide which feedback is worth addressing and whether a request fits the product direction.
   But for a large number of clear, repetitive, low-risk issues, letting AI take a piece of feedback all the way to a PR could be a very interesting way to work.
   That's also the direction we're exploring with Suggix MCP:
   Turn user feedback from a list of requests into development work that AI can actually execute.