(Tips:You can skip ahead to the sections on strengths, weaknesses, and practical tips based on my personal experience. They'll give you a quick overview of my key takeaways and insights from using Dot.)
I think many people have barely scratched the surface of what GPT Dot is capable of.
As a mathematics researcher, I'd like to share some of the ways I've been using Dot in my own work.
Many people seem to use Dot mainly for things like planning trips, ordering coffee, or handling everyday tasks. But I believe its potential goes far beyond that, and I hope my experience can offer some inspiration.
This is not just a list of things I like about Dot. I'll discuss both its strengths and its limitations, based on my actual experience using it for mathematical research.
(And I'm certainly not claiming that I've figured out the best way to use it. I'd love to hear better approaches and ideas in the comments. Criticism and corrections are very welcome!)
How I Use Dot
I primarily use Dot for long-term research projects, advancing complex problems over multiple iterations, and conducting extensive literature reviews.
Rather than diving into the mathematical details, I want this post to help people from different fields discover new possibilities for using Dot. We can discuss mathematics-specific applications another time.
I started using Dot almost immediately after its release, so I've had a relatively long period of hands-on experience with it compared to many newer users.
To really unlock Dot's potential, I think we need to understand both what it does exceptionally well and where it struggles.
The following observations are based on my personal experience, supplemented by some official information.
What Dot Does Really Well
- Powerful model + virtually unlimited usage (the biggest attraction)
Dot was only released recently, and right now we seem to have access to almost unlimited usage.
Dot is powered by Astra, which means we're effectively getting virtually unlimited access to a model that is close to the frontier of what's available today.
That's already remarkable if you use it for individual tasks, let alone for tracking and advancing an entire long-term project.
Astra is also arguably one of the strongest AI models currently available for mathematics, making Dot especially attractive to mathematical researchers.
Of course, that doesn't mean Dot is weak outside mathematics. Astra itself is a highly capable general-purpose model.
2. Exceptional ability to maintain continuity in long-term projects
One common problem with many AI products is goal drift during extended tasks.
As a project becomes more complex, models can gradually lose sight of the original objective, become distracted by intermediate problems, or deviate from the main research direction.
In my experience, Dot handles this noticeably better.
It can often maintain a coherent line of progress across a long sequence of iterations, which is extremely valuable for research.
However, there is an important condition: you need to give it a clearly defined long-term objective.
If your instructions are too short-term, or if your long-term plan contains intermediate points that require additional decisions, Dot may stop and wait for further instructions.
3. Parallel execution of multiple independent tasks
This is easily one of my favorite features.
You can assign several unrelated projects to Dot and have them run simultaneously.
Based on my experience, around 3–5 concurrent tasks seems to be a comfortable range, without noticeably compromising speed or quality.
I've seen others suggest that performance may start to suffer at around 8–9 concurrent tasks, but I haven't personally tested that.
4. Remote project management(Less important)
This one is relatively straightforward, but still very useful.
With just your phone, you can adjust the strategy of a large, ongoing project at any time.
There's no need to re-upload documents or manually retrieve previous files. You can simply give Dot new instructions and let it continue working within the existing project.
Where Dot Falls Short
Interestingly, some of Dot's biggest weaknesses are closely related to its strengths.
1. Version control and file loss (my biggest concern)
This is probably the most serious issue I've encountered.
Imagine you're working on a research project that goes through 50 successive versions.
Sometimes, around version 20, you may suddenly encounter missing files or an unexpected rollback to an earlier version.
In my experience, this is not particularly rare.
So here's one of my strongest recommendations:
Explicitly instruct Dot to maintain a checkpoint and a clear record of changes for every version.
After introducing this requirement into my workflow, I've found that Dot can often locate and recover the most recent checkpoint on its own when files go missing.
However, don't overdo it.
Excessive checkpointing can make your project unnecessarily bloated. Dot may otherwise create multiple intermediate records between two versions.
My recommendation is simple: one proper checkpoint per completed version is generally enough.
2. Slow response to new instructions
Dot often takes some time to process a new instruction before actually starting the task.
In other words, even after you've submitted a request, work may not begin immediately.
From my experience, this initial response delay can be around 1–5 minutes.
It's noticeable, but personally, I find it acceptable given the kind of work Dot can handle.
3. Almost unlimited usage doesn't mean unlimited output throughput
This is something worth understanding before launching very large projects.
Even if the usage allowance is theoretically unlimited, there may still be practical limits on how much content Dot produces in a given stretch of work.
This becomes especially apparent with complex tasks involving very long documents.
For example, when I ask Dot to develop a sophisticated mathematical literature review across multiple iterations, it may progress through many versions while expanding the document relatively slowly.
Sometimes, a new iteration adds only 10–20 pages.
That's quite different from the extremely high output throughput I've experienced with Astra Ultra in Work.
So while Dot is excellent for sustained, long-term progress, you shouldn't necessarily expect the same output speed as a high-throughput Work session.
Practical Tips for Using Dot
1. Understand the difference between parallel and sequential tasks
Dot is very good at handling multiple projects simultaneously.
You can have it advance several independent research problems, organize reference materials, and write literature reviews at the same time.
But there's one crucial rule:
Make the boundaries between independent tasks explicit.
If Task B depends on the results of Task A, don't let them run independently in parallel.
Instead, specify that Task B should begin only after Task A has produced the necessary results.
Otherwise, you may encounter situations where different task branches interfere with each other or even modify files belonging to another project.
In short:
Independent tasks → Run in parallel.
Dependent tasks → Run sequentially.
Shared materials → Define clear access and editing boundaries.
2. Pay attention to usage quotas when Dot invokes other tools
Dot itself doesn't consume your regular usage quota in the same way, but when it invokes Work or Codex, those operations can consume quota.
If you're concerned about Dot using your Work allowance, you can explicitly instruct it:
"If you need to use Astra in Work, ask for my approval first. You may use 6.1 Sol without asking."
Personally, I allow it to use 6.1 Sol more freely because the quota cost is relatively small.
And based on my experience, Dot handles most tasks using its own underlying Astra capabilities anyway, so it often doesn't need to consume much additional quota.
The Most Important Mindset
Here's how I think about working with Dot:
You are the project director and product manager. Dot is your execution manager.
Your responsibility is to define the goals, establish the overall strategy, manage dependencies, evaluate results, and decide when to change direction.
Dot's job is to execute, track progress, manage the details, and continuously advance the work.
I think this division of responsibilities is especially powerful for researchers, developers, and anyone working on complex, long-term projects.
Final Thoughts
Used properly, Dot can become an incredibly powerful assistant.
And with the almost unlimited usage we're currently enjoying, this is a great opportunity to experiment with ambitious workflows and push the tool beyond ordinary everyday tasks.
So I'd encourage everyone to explore what Dot can really do!
And if you've discovered more creative, efficient, or even completely unconventional ways to use it, please share them in the comments.
I'd genuinely love to learn from other users.
Criticism, corrections, and alternative perspectives are all welcome!
