Apparently, I Have to Tell AI Not to Lie About Me#
Captain’s Log. Stardate 2026.294.
The Recruiter Bot has reached the Human Review Queue.
The captain is exhausted.
Unfortunately, the captain is also Starfleet Command, management, quality assurance, and the person who commissioned this entire ridiculous expedition.
🤣
Today’s mission?
Explain to my AI how I make decisions.
APPARENTLY, “YEAH, NAH” IS NOT A VALID DECISION TREE.
😂
Here’s something I hadn’t really considered when I started building Recruiter Bot.
I can read a job advertisement and make a fairly quick assessment.
Yep, I’ve done that.
Nope, haven’t done that.
Well, I’ve done something similar.
That requirement sounds important.
That one sounds like corporate jargon for something I already know how to do.
And sometimes?
It’s just a gut feeling.
I don’t consciously work through a detailed decision tree.
I just… know what I think.
But AI doesn’t automatically know what I mean.
And apparently, neither do I until I have to explain it.
🤣
So I’ve been turning those instinctive judgments into explicit rules.
What counts as relevant experience?
When is something transferable?
What’s the difference between a genuine skills gap and a gap in my documentation?
How much evidence is enough?
When should the system make a recommendation?
And when should it stop and ask me?
Because there’s a difference between an AI identifying a possible connection and having enough evidence to make a claim.
I don’t want Recruiter Bot inventing qualifications or achievements just because they’d look good on an application.
But I also don’t want it dismissing experience simply because I haven’t described it using the same terminology as a job advertisement.
SOMEWHERE BETWEEN “MAKE IT UP” AND “YOU’VE NEVER DONE THAT” IS A FAIRLY IMPORTANT THIRD OPTION.
Ask me.
😂
Which brings us back to the Human Review Queue.
During testing, the system flagged five items for review.
One was budget management.
I have budget management experience.
But after documenting roughly 40 pieces of career evidence, I apparently hadn’t described it in enough detail for the system to recognise it.
CAREER ARCHAEOLOGY II: THE EXCAVATION NOBODY ASKED FOR.
🤣
And that’s when I realised something.
I wasn’t just teaching AI about my career.
I was trying to explain the decision-making process I’d developed over years of working.
All those little judgments I make without consciously thinking through every step.
And every time I thought I’d documented enough, testing revealed another assumption I’d forgotten to explain.
Apparently, the hardest part of teaching AI to think like me is discovering how many decisions I make without realising I’m making them.
And I’m only on Step 179.
😂
When I started this project, I had a rough idea of what I wanted to build.
What I didn’t know was how many rules, assumptions, and judgments were hiding behind something as simple as reading a job advertisement.
At this point, I’m beginning to think I don’t need another AI assistant.
I NEED TO CLONE THE GITHUB REPOSITORY OF MY BRAIN.
Preferably with the documentation already completed.
Because apparently, even I don’t know everything that’s in there.
🤣