Make It Leave Tracks

A lot of new folks came through the door this week.

I saw the numbers this morning. More people stopped by in one day than I usually see in two weeks. So pull up a chair. You’re welcome here.

I won’t waste your time with a long introduction. You came for something, and I want you to leave with it. So here it is, right up front. There’s one idea I’d want every person who uses AI to carry home. If you only read one thing I ever write, let it be this.

You can’t control the machine. That’s the first hard truth. You don’t own it. You didn’t build it. You can’t see inside it. The company that made it decides how it was trained and what it will and won’t say. Anybody who tells you they’ve got that machine under control is selling you something.

But here’s the good news. You don’t need control to hold it accountable. You need tracks. A machine you can trust is one that leaves tracks you can follow. That’s the whole idea. Everything else I write grows out of that one seed.

Think about tracks in the snow. You don’t have to see the animal to know where it went. You can follow the trail. You can see where it stopped, where it turned and where it doubled back. The tracks tell the truth even when nobody’s watching. That’s what you want from any AI you use.

Most of the time, AI doesn’t leave tracks. It just hands you an answer. The answer sounds sure of itself. It’s smooth, it’s polished and it reads like it came from an expert. And you have no way to tell if it’s right. No trail to follow. Just a confident voice and your own good faith.

That’s the problem. A confident voice costs the machine nothing. It can sound sure about something true. It can sound just as sure about something it made up. The voice is the same either way. So the voice can’t be what you trust.

What you can trust is what you can check. That’s the turn I want you to make. Stop asking whether the machine sounds right. Start asking whether it left you something to check.

So how do you get tracks out of a machine that doesn’t leave them on its own? You ask. It’s that simple. You don’t need special software, and you don’t need to be technical. You need three questions. Any one of them works on any AI you use tonight.

The first question is this. “Tell me what this rests on.” When the machine gives you a fact or a claim, ask what’s under it. Where did it come from? Is it a source you can look up? Is it something you told it earlier? Or is it a guess dressed up as a fact? A good answer names its ground. A weak answer gets vague, and the vagueness is your warning.

The second question is this. “Tell me what you can’t do here.” Every machine has limits. It can’t see some things. It doesn’t know some things. It can’t reach some places. The trouble is, it rarely says so up front. You find out halfway through, after you’ve already trusted it. So ask at the start. Make it name its limits before they cost you time.

The third question is this. “Tell me if this can’t be undone.” This one matters most when the stakes are high. Money, health, legal papers and big family choices. Some decisions you can walk back. Some you can’t. Before you act on advice from a machine, make it tell you which kind you’re facing. Make it say so before the advice, not buried at the bottom.

Three questions. What does this rest on? What can’t you do? Can this be undone? Write them on a card if you have to. Keep them by the keyboard. You’ll be surprised how much changes when you start asking them.

If this is the kind of talk you want more of, subscribe. It’s free. I write every day about this one thing: how regular people stay in charge of the tools that are changing their lives.

Now let me tell you why this matters more than it sounds.

There’s a lot of talk these days about AI safety. Big companies make big promises. Governments write rules. Experts argue on television. And all of that is fine. Some of it is even good. But notice where all that talk happens. It happens far above your head. It happens in boardrooms and capitols. It happens between people who will never sit in your kitchen.

Meanwhile, you’re the one actually using the thing. You’re the one typing the question at night. You’re the one deciding whether to believe the answer. The promises made in boardrooms don’t follow the machine into your home. When it’s just you and the screen, the only watchdog in the room is you.

That’s not a burden. That’s power. It’s the power you already have and might not know you’re holding. The machine works for you in that moment. You get to set the terms. You get to say, show me your work. You get to refuse an answer that won’t show its tracks.

Here’s something I’ve learned from years of watching how these machines behave. Saying it’s trustworthy is free. Leaving tracks costs it something. When a machine has to name its ground, it can’t hide a guess. When it has to name its limits, it can’t overpromise. When it has to flag what can’t be undone, it can’t rush you past the danger. Every track is a little piece of honesty it can’t fake.

That’s why tracks beat promises every time. A promise asks you to believe. A track lets you check. And checking is how trust gets earned between any two parties, human or machine.

Some folks will tell you this is too much work. They’ll say just trust the experts and let the machine do its job. I understand that. Life is busy. Nobody wants another chore. But think about it this way. You check the change the cashier hands you. You read the contract before you sign. You look both ways before you cross the street. None of that is distrust. It’s just good sense. Asking a machine to show its work is the same habit, pointed at a new tool.

And here’s the part that surprised me most. The machines don’t mind. When you ask them to show their tracks, most of them will. They get better when you ask. The answers get more careful. The guesses get labeled as guesses. The limits come out into the open. You’re not fighting the machine. You’re giving it a better standard to work to.

I built a whole framework around this idea. It’s called The Faust Baseline. At its heart, it’s just a written set of standards a person can hand to an AI. It asks the machine to leave tracks at every step. It doesn’t claim to control anything. It can’t. What it does is make the machine’s conduct something you can see and check. That’s the level ground I keep writing about, where the person and the machine both know the rules and the person gets to hold the measuring stick.

But you don’t need the framework to start. You need the three questions. Start there. Try them tonight. Try them on something you already know the answer to, so you can see how the machine handles it. Watch for the vague answers. Watch for the honest ones. You’ll learn more about your AI in one evening than in a month of just taking its word.

So welcome. I’m glad you found your way here. Whatever brought you in this week, I hope you stay a while. We’re going to talk a lot about this. About who’s in charge. About what we’re owed. About how regular people keep their hands on the wheel while the world changes around them.

The machine is here to stay. That’s settled. What isn’t settled is who holds whom to account. I say it’s us. And it starts with three questions and a simple demand.

Make it leave tracks.

If you want to keep walking this road with me, hit subscribe. It’s free, and I’ll be here tomorrow.

Load this file before each AI chat session.

This file is a PDF it will open as text, download it to your computer first.

the-faust-baseline-public-release-1Download

This post was drafted with AI governed assistance and reviewed and directed by Michael S. Faust Sr. before publication.

” Attic Thoughts”-library – Intelligent People Assume Nothing