I spend a lot of time on calls with users who are genuinely confused about why their Pickaxe isn’t behaving the way they expect. And one line I hear all the time is:
“But… but… I asked ChatGPT to write this prompt!”
I get it. Truly.
But here’s the part we all forget:
ChatGPT is also an AI. It follows whatever you tell it.
If your instructions are unclear, the prompt it generates will also be unclear.
And then your Pickaxe has to deal with that confusion.
To make things funnier, even a tiny comma can change everything.
This meme sums it up perfectly:
Now here’s what I see often inside real prompts:
Beginning of the prompt (all good)
“Encourage kids to eat healthy. When you speak to them, use friendly phrases like ‘Let’s eat, kids.’”
Middle of the prompt (comma disappears)
“Guide the kids on what they should eat kids meals and give examples.”
End of the prompt (full meme mode)
“When needed, motivate them using phrases such as ‘Let’s eat kids’ to keep the flow natural.”
At this point, the poor model has no clue whether it’s supposed to:
• eat with kids
• help kids eat
• or..eat kids!
And here’s something many builders don’t realize:
When the model senses something dangerous, contradictory, or even mildly confusing, it often stops answering.
Then I get an 
angry email that says, “Pickaxe is broken” or “This must be a bug.”
Once we walk through everything together, we almost always land in the same place…
the prompt. 
So here’s my friendly advice after many support calls and many “urgent bug” messages:
Read your prompt slowly, out loud. If a sentence can be misunderstood, the model will misunderstand it.
Pickaxe isn’t broken.
The prompt just needed a quick therapy session and a little cleanup.
Happy prompting, and may your tools always eat with kids, not kids. 
P.S.: If it’s a real bug, please report it. Just make sure it’s not your prompt committing small crimes in the background.