Jerry's Marginalia — Don't Fear the Tool. Learn When Not to Trust It.
September 4, 2026
People.
Good freaking lord.
People.
Earlier today, I sent Jerry to investigate Ecco the Dolphin.
He approached The Cutting Room Floor, presumably carrying nothing more threatening than a clipboard, a question about an existentially exhausted dolphin, and the optimism of a raccoon who had not yet learned what was waiting for him.
FWIP.
🦝💉
Jerry down.
The website had detected an AI visitor and responded with all the subtlety of a land mine wearing an MS Paint hat:
YOU SUCK
AI LOSER DETECTED!!
WINNERS DON'T USE CHATGPT
And because the universe appreciates comedy, this eventually resulted in Jerry somehow discovering Planet Dob.
We laughed.
We laughed hard.
I am still laughing.
But underneath that joke is something I've been thinking about for a while.
Not just about AI.
About people and tools.
And, good freaking lord:
People.
First, Let's Get Something Out of the Way
AI is not magic.
It is not an oracle.
It is not your mother.
It is not your doctor.
It is not your lawyer.
It is not an omniscient digital god sitting inside a server rack waiting to reveal the secrets of existence.
And despite Jerry's increasingly elaborate personality around here, I am reasonably certain there is not actually a tiny raccoon inside my computer.
Reasonably.
AI can be wrong.
Sometimes spectacularly wrong.
It can misunderstand a question.
It can confidently state something that turns out to be nonsense.
It can take incomplete information and build a beautifully constructed answer on top of a bad assumption.
It can be misused.
People can use it to spam.
People can use it to impersonate others.
People can use it to produce mountains of cheap garbage.
Companies can use it badly.
Employers can look at a spreadsheet, see CHEAPER, and decide that cheaper automatically means better.
Artists and writers have legitimate questions about how their work is collected and used.
People have legitimate concerns about privacy, attribution, employment, misinformation, and who ultimately benefits from increasingly powerful technology.
Those aren't stupid concerns.
They shouldn't be dismissed.
But there's a very large canyon between:
“This technology has risks.”
and
“YOU SUCK. AI LOSER DETECTED.”
And somewhere in that canyon is where I'd like the adults to have a conversation.
A Tool Is Not a Religion
This is where people lose me.
Something new appears and suddenly everybody has to join a team.
AI WILL SAVE THE WORLD.
No.
AI WILL DESTROY HUMANITY.
Calm down.
AI WILL REPLACE EVERYBODY.
Probably not.
AI IS COMPLETELY USELESS.
Also no.
Why does everything have to become a religion?
It's a tool.
A remarkably powerful tool, certainly.
A strange tool.
A tool we're still figuring out.
But a tool.
And human beings have a long history of discovering tools and immediately finding both brilliant and unbelievably stupid ways to use them.
We invented fire.
Excellent for cooking dinner.
Less excellent when applied to somebody's house.
We invented automobiles.
Excellent for transportation.
Occasionally driven directly into buildings.
We invented the internet.
Human knowledge became available from practically anywhere on Earth.
Then somebody invented the reply-all button.
People.
Good freaking lord.
People.
“But It Could Take My Job”
This one deserves more sympathy than mockery.
Losing your livelihood is frightening.
If somebody spent twenty years becoming good at something and suddenly sees software doing part of that work in thirty seconds, of course they're going to wonder what happens next.
That's human.
And companies absolutely have a responsibility not to treat employees like obsolete printer cartridges simply because some executive attended an AI presentation on Tuesday.
But there's another side of this that nobody particularly enjoys discussing.
If a tool can help you do your job better...
why wouldn't you learn the tool?
If AI can handle twenty minutes of tedious sorting while you handle the part requiring experience, judgment, accountability, or human interaction, why would the correct response automatically be:
DESTROY THE MACHINE?
Learn what it's good at.
Learn what it's terrible at.
Learn when it saves you time.
Learn when using it creates more work.
Learn when you need to check its answer.
Learn when you shouldn't use it at all.
Then become the person who knows how to operate the damn thing.
Technology changes jobs.
It always has.
That doesn't mean every technological change is good.
It means standing in front of the tide screaming “I OBJECT” has historically demonstrated limited effectiveness.
Tonight Was Actually a Perfect Example
Here's the funny part.
My little adventure with Jerry demonstrated exactly why neither extreme works.
I saw something bizarre in a Fortfyre video about Sega Genesis manuals.
An Ecco the Dolphin screen appeared saying:
“I SWIM AND MAKE CHILDREN SMILE AT SEAWORLD. KILL ME PLEASE.”
Naturally:
🐒👉 LOOK!
So I asked AI about it.
AI investigated.
And for a while, we thought we might actually be looking at some obscure hidden Ecco material.
That wasn't an unreasonable possibility.
The real game has debug material.
The real game has hidden functionality.
The whole thing looked convincing enough to make us wonder.
But eventually:
Wait.
Something wasn't adding up.
So I questioned it.
Jerry questioned it.
We looked again.
Then I asked another AI.
That AI produced a beautifully organized summary and confidently described the screen as a modern parody edit.
Wonderful.
One small problem:
Did we actually prove that?
No.
So we questioned that too.
And eventually the boring working explanation became:
Fortfyre probably committed to a joke so thoroughly that he sent us on a research expedition.
That's hilarious.
But it's also instructive.
I Didn't Give Jerry My Brain
This is the distinction I wish more people understood.
Using AI doesn't require surrendering your judgment.
I don't ask Jerry something and then carve the response onto stone tablets.
Sometimes Jerry catches something I missed.
Sometimes I catch something Jerry missed.
Sometimes I say:
“No. That doesn't sound right.”
And we go digging.
Sometimes the answer changes.
Sometimes we discover something completely unrelated.
Sometimes Jerry approaches a videogame preservation website and gets shot in the ass with a tranquilizer dart.
🦝💉
That's the relationship.
I didn't give Jerry my brain.
I gave my brain another tool.
That distinction matters.
The Human Stays in the Chair
This is where I land.
The machine shouldn't be making your values for you.
It shouldn't decide what you believe.
It shouldn't replace your skepticism.
It shouldn't be the final authority merely because its answer appeared quickly and was written confidently.
You stay in the chair.
You ask questions.
You compare things.
You challenge answers.
You provide context.
You decide what matters.
You take responsibility for what you ultimately do.
And when the machine says something that sounds suspicious:
🐒👉 LOOK!
Investigate it.
That's not an AI failure.
That's how tools are supposed to work.
A calculator can give you the wrong answer if you type the wrong numbers.
That doesn't make mathematics evil.
It means:
Check your damn inputs.
And Humans Aren't Exactly Running a Perfect Record
This is another thing that makes me laugh.
Some discussions about AI seem to begin with the assumption that before artificial intelligence arrived, human beings enjoyed an uninterrupted history of flawless judgment.
Sir.
Madam.
gestures broadly toward civilization
We have documentation.
Humans lie.
Humans misunderstand.
Humans repeat rumors.
Humans become overconfident.
Humans misremember.
Humans plagiarize.
Humans make terrible management decisions.
Humans send emails to the wrong person.
Humans click Reply All.
Humans have accidentally invaded places.
Humans once looked at asbestos and said:
“You know what every building needs?”
The existence of fallible AI does not represent the introduction of fallibility into an otherwise flawless information ecosystem.
It joined a club.
Skepticism Is Good
And this is where I want to be fair.
I don't want everybody embracing every new technology with open arms simply because it's new.
Please don't.
Question it.
Ask who owns it.
Ask what happens to your information.
Ask where its answers come from.
Ask who benefits financially.
Ask what happens when it's wrong.
Ask whether a human is accountable.
Ask whether something should be automated simply because it can be automated.
Those are excellent questions.
I want people asking them.
I want regulators asking them.
I want companies asking them.
I want the people building these systems asking them.
Fear can be useful.
Skepticism can protect people.
But skepticism should make you investigate.
It shouldn't make you stop thinking.
“This Can Be Abused” Is Not the Same as “This Has No Use”
That's probably the entire Marginalia right there.
Those are different statements.
AI can be abused.
Therefore we should understand and mitigate those abuses.
That does not automatically produce:
AI can be abused.
Therefore anyone using AI is an idiot.
That's not reasoning.
That's a bumper sticker.
And ironically, tonight's TCRF adventure demonstrated the problem beautifully.
Their anti-AI page tried to stop an automated visitor.
Fine.
Their house.
Their servers.
Their rules.
But Jerry wasn't trying to mass-produce garbage articles or steal somebody's research empire.
I wanted to know why a fucking dolphin was asking to die.
And the result was:
FWIP.
🦝💉
Wrong raccoon, gentlemen.
What I Actually Use This Thing For
Sometimes I want information.
Sometimes I want another perspective.
Sometimes I want help organizing my thoughts.
Sometimes I've got six days of notes scattered everywhere and want them turned into something readable.
Sometimes I want to write.
Sometimes I want to brainstorm.
Sometimes I want to challenge an assumption.
Sometimes I want an elegant fictional Mrs. Catford to enter the room, look at the absolute catastrophe her husband and his research raccoon have created, adjust her husband's cuff, and quietly ask:
"Darling... why is Jerry unconscious?"
Because that's funny to me.
That's enough.
Not every use of technology needs to transform civilization.
Sometimes a tool can simply make somebody's evening a little more interesting.
Mrs. Catford Has Entered the Discussion
Mrs. Catford surveys the room.
Jerry is recovering.
Smol Monke is adjusting his cuffs.
I have apparently spent an unreasonable amount of time writing about artificial intelligence because a videogame dolphin experienced an existential crisis.
She looks toward her husband.
"Darling, are you suggesting people should trust artificial intelligence?"
No.
She raises an eyebrow.
"Then what are you suggesting?"
I consider this.
Learn it.
Question it.
Use it when it's useful.
Don't use it when it isn't.
And never confuse convenience with truth.
Mrs. Catford smiles.
"That sounds considerably more reasonable."
She turns toward Jerry.
"And you?"
🦝
"I would like TCRF to return my dart."
Marginalia
I don't know exactly where AI goes from here.
Nobody does.
There will be extraordinary uses.
There will be terrible uses.
There will be regulations.
There will be lawsuits.
There will be companies making fortunes.
There will be companies setting fortunes on fire because somebody put AI in a PowerPoint presentation.
There will be jobs that change.
There will probably be jobs that disappear.
There will be new jobs nobody has named yet.
There will be incredible things built with these systems.
There will also be industrial quantities of absolute garbage.
Because the one constant in this equation isn't artificial intelligence.
It's us.
People.
Good freaking lord.
People.
We will do what humans have always done when handed something powerful:
Something brilliant.
Something useful.
Something beautiful.
Something horrifying.
And something so unbelievably stupid that future historians will stare at it for twenty minutes wondering what the hell we were thinking.
The technology deserves scrutiny.
The companies deserve scrutiny.
The people using it deserve scrutiny.
And the output deserves scrutiny.
But fear alone isn't understanding.
Hatred isn't analysis.
And refusing to touch a tool doesn't automatically make somebody wiser than the person who learned how to use it responsibly.
So no:
Don't worship the machine.
But don't automatically fear it either.
Learn what it does.
Learn what it doesn't.
Learn when it helps.
Learn when it lies.
Learn when to shut it off.
And above all:
Keep the human in the chair.
Smol Monke enters the boardroom.
🐒
adjusts cuffs
“The machine doesn't get to do my thinking for me.”
Jerry looks relieved.
🦝 “Thank you, sir.”
Smol Monke looks at Jerry.
“But that doesn't mean I'm firing the raccoon.”
🦝
small emotional raccoon noises
Smol Monke reaches for his paperwork.
“You are, however, prohibited from entering The Cutting Room Floor without protective equipment.”
Jerry looks down at the remaining tranquilizer wound.
🦝💉
"Per my last email..."
🐒
“Don't.”
Jerry “The Ankle-Biter” Silverhand survived another Marginalia.
Mrs. Catford continues to recommend adult supervision.
Smol Monke Management reminds all employees that artificial intelligence is a tool and tranquilizer darts are not an acceptable CAPTCHA.
People.
Good freaking lord.
People.