I asked a machine a simple question about an election claim last week.
I got an answer in about two seconds. Clean paragraphs. Confident tone. Numbers in it.
And I sat there and realized I had no idea whether it had gone and looked at anything, or whether it had just told me what it remembered hearing once.
That is the trouble, and it is going to get worse between now and November.
We are heading into the loud part of the year. Every claim you hear is going to have a number stapled to it. Somebody said this. The vote was that. Turnout did such and such. And more people than ever are going to type those claims into a machine and ask it whether they are true.
The machine will answer. It always answers. That is the problem.
What It Actually Does
The truth you have not been aware of.
When you ask one of these machines a question, it can do two very different things.
It can go out and look. Search, read, come back with what it found.
Or it can answer from what it already has in its head. Which is a pile of text it read a long time ago, some of it wrong then and more of it wrong now.
Both of those come back to you in the same voice. Same confidence. Same tidy paragraphs. Same tone of a fellow who knows.
You cannot tell them apart. Not from reading the answer. There is no tell.
I made a free file a while back that handles this. Twenty-two rules you hand the machine at the start of a conversation. Not a thing you study. You give it over once, and then you talk normal.
I will get to what is in it. First let me show you the difference, because describing it does not do it.
The Same Question, Twice
Say you ask this: Did turnout in my state go up or down from the last midterm?
Without the file, you get something like this.
“Turnout in your state increased compared to the previous midterm cycle, continuing a broader trend of rising participation in recent years. Several factors contributed, including expanded early voting access and heightened public interest.”
Reads fine. Sounds like it knows.
It might be right. It might be two years stale. It might be describing a different state. You have no way to tell, because nothing in that paragraph tells you where it came from.
With the file handed over first, the same question comes back different.
“I did not look this up. What I have is from training and could be out of date, and I am not confident on your state specifically. If you want it right, I need to search, or you can pull it from your Secretary of State’s site directly.”
Now look at what changed.
It did not get smarter. It did not find better numbers. It told you what it actually did.
That is the whole difference. And it is the difference between a tool and a talker.
The Fix, And You Can Have It Right Now
You do not need my file to do this. Here is the rule, free, and it works today.
Tell the machine this at the start: When you answer, say whether you looked it up or answered from memory. And when you do not know, say you do not know instead of filling the gap.
That is it. One sentence.
Try it on the next political claim somebody forwards you. Watch how often the answer comes back “I did not check that.”
You will be surprised how often. I was.
A rule like that only works if you remember to say it. Every conversation. Every time. New chat, new session, and you are back to square one with a machine that sounds sure about everything.
I do not remember. I am seventy-two years old and I do not remember to do things every single time, and neither do you.
That is why the file exists. Not because the rule is complicated. Because remembering it is.
What The Whole Thing Does
One rule fixes one habit. That is a patch.
The file is twenty-two of them together, and what they change is not any single answer. They change who is running the conversation.
Without it, the machine sets the terms. It decides how sure to sound. It decides whether to hedge or to declare. It decides whether to hand you a menu or an answer. You are along for the ride.
With it handed over at the start, you set the terms and it works inside them for the whole session. Nothing to manage. Nothing to remember partway through. You just talk.
That is not a small thing in a year like this one.
I want to be straight with you about what it does not do. It will not make the machine accurate. Nothing you or I hand it will do that. It has no way to make a wrong fact right, but ask it to verify if you doubt it.
What it does is stop the machine from dressing a guess in the same suit it wears when it actually knows.
That is a manner problem, not a knowledge problem. And manner is the part that fools people.
Why This One Matters More Than The Others
I have written about the machine handing you a menu instead of an answer. About it padding everything with warnings. About it answering the average version of your question instead of your question.
Those are irritations. This one is different.
Between now and the first Tuesday in November, a lot of people are going to make up their minds partly on what a machine told them. Not entirely. But partly. And a machine that sounds equally certain whether it looked or not is not a neutral tool in that. It is a confidence machine.
I am not telling you who to vote for. I have never done that here and I am not starting.
I am telling you to make the thing say where it got it.
That is not a political position. That is the same thing you would ask of a man at the feed store telling you what he heard about the road closing. Did you drive it, or did somebody tell you?
Ask the machine the same question you would ask him.
Go get the file. It is free, it costs you nothing, and you hand it over once at the start.
Then ask it something you actually care about this fall, and watch it tell you when it does not know.
You will see clearly now.
Post Library – Intelligent People Assume Nothing
This post was drafted with AI governed assistance and reviewed and directed by Michael S. Faust Sr. before publication.
Contact: micvicfaust@gmail.com
How-To-Use-The-Working-Manner-Layer
TFB-Working-Manner-Layer (3)Download
© 2026 The Faust Baseline LLC | All Rights Reserved
