Suggested tags: work, limits, usage, performance, agent
Hello OpenAI team!
First, thank you for ChatGPT Work. I really enjoy using it for larger and more complex projects.
That usefulness creates a slightly funny problem:
the better Work becomes, the more we want to use it — and the faster the available usage can disappear.
A single advanced workflow may involve:
- opening many files;
- researching several sources;
- editing documents;
- revising outputs;
- interacting with websites;
- performing multiple sequential actions.
Higher limits would certainly help, but I think there is another opportunity as well:
make Work more efficient.
For example, Work could:
- reuse information already gathered during the same task;
- avoid reopening unchanged files;
- cache intermediate results when appropriate;
- use lighter models for simple subtasks;
- reserve expensive reasoning for genuinely difficult steps;
- avoid duplicated web searches;
- reduce unnecessary agent actions.
I would also love an optional Work efficiency/reasoning selector:
Efficient
Prioritize lower usage and faster execution for routine tasks.
Balanced
The normal default.
Deep Work
Use more reasoning and resources when maximum quality matters more than consumption.
It would also be useful to see approximately how much Work allowance a task consumed.
This would give users more control over the trade-off between:
- quality;
- speed;
- reasoning depth;
- usage.
Of course, higher overall Work limits would still be very welcome.
But the solution does not necessarily have to be making everything unlimited.
It could also mean:
making every unit of Work usage accomplish more.
The best Work session is one that ends because the task is finished — not because the quota is.
Thank you for hearing out a very enthusiastic Work user. 
All the best,
Merpherus