The First Time I Beat the AI#
I recently experienced a small but deeply satisfying moment while building Recruiter Bot.
I spotted a potential problem before ChatGPT did.
And I cannot adequately explain how pleased I was with myself.
😂
During one working session, ChatGPT and I were figuring out how to test part of an n8n workflow.
It suggested the next step.
And something about that suggestion bothered me.
I started thinking through what might happen if we followed its advice.
Then I realised:
Hang on. Could that roll us back and potentially undo half a day’s work?
So I explained my concern to ChatGPT.
Not just that I didn’t like the suggestion, but why I thought it might cause a problem.
And ChatGPT essentially responded:
Oh. Yes. That could happen.
Well, well, well.
😂
We didn’t run the risky test.
We found another way to test the workflow, and it worked.
So I can’t claim I’d definitively proved ChatGPT wrong.
But that’s not what made the moment exciting.
When I started this project, I needed ChatGPT to explain practically everything about how the different components worked together.
Now, here I was, recognising a potential consequence of a technical decision before following the instructions.
I was actually starting to understand what I was building.
And that felt brilliant.
Because my whole reason for building Recruiter Bot was to learn how these systems work, rather than end up with a black box I couldn’t explain.
There’s also a funny contradiction here.
My preferred learning methodology is essentially:
Click around and find out.
Yet one of my proudest moments so far was recognising that perhaps, on this particular occasion, we should not click around and find out.
Apparently, I’m developing risk-management skills alongside my technical ones.
😂
It’s also been a useful reminder about working with AI.
ChatGPT can explain concepts, suggest approaches, and help me work through problems.
But it doesn’t always have the full context of what’s happening on my screen.
And a reasonable-sounding suggestion can still overlook something important.
Which is why understanding the system matters.
Not because I need to know more than AI.
But because I need to know enough to question it.
I started Recruiter Bot because I wanted to learn how to build something with AI.
I hadn’t expected one of the most satisfying milestones to be the day I questioned its advice.
But there it was.
A potential problem avoided.
A little evidence that all this experimenting was paying off.
And one very pleased, slightly out-of-practice tech nerd sitting at her computer thinking:
HA! I beat the AI!
😂