Is ChatGPT Agent Mode Falling Behind Its Competitors?
Table of contents 3 items
What it actually feels like to use
When you start a task, ChatGPT Agent Mode first sets up its virtual desktop, and you wait.
After a few minutes, it switches to “thinking.” This stage can drag on too.
Then the summarization tasks finally kick off.
After 37 minutes, the deck is finished.
And here is the slide it produced.
The verdict on real usage
About 40 minutes for a slide that, frankly, isn’t very polished.
You’d get further by combining an AI search tool with Marp — both the speed and the content quality would be better.
Manus has been ahead in this space for a while now, and its slide generation is genuinely impressive. The catch is that Manus is hard to deploy inside a typical company because of its operational and security constraints. So for now, the realistic option is to wait for Agent Mode to keep improving.
Current state
There’s nothing strikingly new here, and overall it lands below the existing competition. You’ll also feel the friction of repeated permission prompts on tasks that require login. Worse, scheduled (task) execution seems buggy — it didn’t run at the time I specified.
If those gaps close, you could imagine handing off meeting minutes or data aggregation overnight while you sleep. We’re not there yet, and there’s still a lot of road to cover.
Was this article helpful?