Lately, I've really been feeling a bit overwhelmed by AI. A few days ago, I wanted to create a workflow for "automatic data organization + automatic content generation." Now, every day online, people are talking about AI Agents, automation, and the future of low-code. At first, I thought this stuff should be easy to set up and run.

But once I started actually working on it, I immediately fell into endless debugging mode.
This prompt isn't right, try a different approach. Claude outputs too generic, switch to GPT.
GPT's logic is correct, but the format isn't good enough.
Then I keep changing tools, adding rules, modifying workflows. Sometimes I can't even tell if I'm "creating content" or "training AI."

And this kind of thing really messes with your mindset.
The same tool, someone can finish in ten minutes, but I sometimes spend two hours tweaking a single result.
All those highly starred skills online look impressive, one after another, but when you try to apply them to your own scenario, many just don't run smoothly.
You copy what others do, but in the end, you still have to modify it little by little yourself.

Later, I even considered hiring someone specifically to build a workflow for me.
But thinking carefully, small teams are actually in a tricky spot.
Hiring someone who understands AI isn't cheap; explaining needs, clarifying logic, back-and-forth revisions—sometimes it's more exhausting than doing it yourself.

Gradually, I realized that the biggest problem with AI isn't actually that it's "not smart enough."
It's that ordinary people need to learn way too much just to use AI well.
You have to study prompts, understand model differences, learn skills, figure out workflows, and find the right tools for each task.
Work only takes up part of your time; most of your energy goes into "making AI work properly."

So recently, when I tried @dappOS_com的 and @xbubble_xyz, I felt their approach was quite different.
Many AI products now teach users: how to write prompts, how to set up workflows, how to tune Agents.
But xBubble seems to be doing something else:
"AI learns from AI, AI uses AI."

From my own experience, the biggest takeaway is that I no longer have to constantly worry about "which model to use for this step."
I just tell it what I want.
Bubble Pilot will automatically recognize the task type and then distribute it to the appropriate SOP and execution path.
If there's no existing SOP, it will fallback to a general Agent.

The key is, its backend Bubble Engine keeps learning.
Which models are suitable for which tasks, which tool combinations are more stable, which workflows have higher success rates—these previously tedious things are now being handled by AI itself.

This experience is actually pretty awesome.
Because in the past, it wasn't that AI couldn't do the work, but that users had to learn to program just to get AI to do anything.
Especially with modes like Bubble Computer, I feel deeply.
In the past, to complete a full task, I had to open several windows: research, organize, write content, proofread, then output.

Now, it can run the entire chain on its own.
Including local modes like Bubble Personal, which can directly operate files and browsers on the machine, but without users needing to set up environments themselves.

I'm increasingly convinced that the future of truly good AI shouldn't make ordinary people more exhausted.
Instead, AI should learn how to use AI on its own.
Users just need to tell it the goal, and the rest is handled by the system.