New gold. Same con.
On January 24, 1848, a carpenter named James Marshall found a few flakes of gold in the tailrace of a sawmill he was building for John Sutter on the South Fork of the American River, at Coloma, California. Over the next several years, roughly 300,000 people dropped what they were doing and headed for California. Farmers. Clerks. Schoolteachers. Sailors who walked off their ships in San Francisco Bay and never went back for them.
Most of the people who came to prospect had never mined anything in their lives.
That is the part everyone forgets. The Gold Rush gets remembered as a story about gold. It was mostly a story about people who had no idea what they were doing, showing up with a tin pan and a total certainty that they were about to be rich.
Most of them never found the fortune they came for.
The gold was real. The easy part ran out fast, the work got harder and more technical, and the money ended up with the people who arrived with capital, real technique, or something to sell to everyone else. The gold was real. The self-issued expertise was not.
That same wave has now hit artificial intelligence. The field is new, it is moving fast, and there is no agreed-on test that separates the people who understand it from the people who only sound like they do. And the people showing up with a pan are calling themselves consultants.
The plaque is the pan
In 1849, the barrier to becoming a prospector was a pan. You did not need a license, a degree, or one day of experience. You needed the nerve to call yourself a forty-niner and the price of a piece of tin. The credential was self-issued.
Now look at who is explaining AI to you.
Yesterday’s social media guru is today’s AI consultant. The growth hacker became a prompt engineer. The crypto guy, who was the blockchain guy, who was the SEO guy, pried the old plaque off the door, nailed up a new one, and started explaining large language models to a room that knows even less than he does.
Nobody checked whether there was anything behind the door. In a field this young, the word expert is unfalsifiable. There is barely any track record to check him against, and he has been standing on the little that exists for about a week. The plaque is the whole qualification.
Same as the pan.
Sell the belief, then sell the shovel
One of the first men to get rich in the California Gold Rush never mined a thing.
His name was Sam Brannan. He ran a store near Sutter’s Fort. In the days after he heard about the discovery, he did two things, and by most tellings the order matters. He bought up the picks, shovels, and pans in San Francisco. Then he walked through the streets holding a bottle of gold dust over his head, shouting “Gold! Gold! Gold from the American River!”
He was not spreading news. He was manufacturing customers. He owned the inventory before he made the announcement. By some accounts he cleared 36,000 dollars in nine weeks, which made him one of the first millionaires of the rush. He got rich selling the belief, and the tools to act on the belief. Not the gold.
That is one recognizable business model in AI right now, and it is thriving. At least Brannan’s shovels were real.
You sell the belief that there is gold. You sell the tools to go dig for it. You never have to produce any gold yourself, because you were never in the gold business. You were in the belief business. And a brand-new technology is the greatest belief business ever invented, because nobody can tell yet whether the map you are selling leads anywhere at all.
The relay
Here is how the nonsense actually spreads, and it is never just one person.
Somebody who does not understand the thing says it anyway, with confidence, because confidence sells. Call him the originator. Then somebody who understands it even less repeats it, because repeating a confident-sounding thing makes him look early, makes him look like he is in the know. Call him the repeater. He is not passing along knowledge. He is borrowing status. Then a third man, who understands nothing, takes it as gospel and says it to a fourth to look smart, and the fourth says it to a fifth.
Not one of them checked.
The originator did not check, because he was selling. The repeater did not check, because agreeing was easier than understanding. The echo did not check, because checking never once occurred to him. The claim did not survive because it was true. It survived because repeating it was less work than verifying it.
In 1849 they called this a rumor, and the country ran on them. San Francisco had been fooled by false gold before Brannan ever raised his bottle. A claim could move through a camp before anyone tested it. The cost of repeating it was one breath. The cost of checking it was days of hard travel. So the rumor always got the head start.
Same arithmetic today. Repeating is free. Checking is expensive. The garbage moves at the speed of a repost, and the correction moves at the speed of somebody actually doing the work.
And the one who pays is never the one who sold. It is the buyer, who cannot tell the assayer from the man with the pan and hands his money to whoever sounded surest. Surest has never been a synonym for right.
There is no assay office
Here is the difference between 1849 and now. It is the whole difference, so sit with it.
The Gold Rush had fool’s gold. Pyrite. It sat in the same riverbeds as the real thing, flashed the same yellow in the same sunlight, and fooled plenty of green miners. Men filled their bags with it and thought they were rich.
But the pyrite could be caught, because a verdict existed and it had an address. You could carry the sample to an assayer, who would test it and tell you flat what you actually had. That did not catch every fooled man, because not every fooled man went. But the place existed, and everyone knew it existed. There was a counter where confidence could be laid against reality and given a verdict. The assay did not care how sure you were, or how loud you had shouted on Market Street. It cared what was in the rock.
The Gold Rush soon built a wall that reality could be brought to.
AI has no such wall, and no counter anyone is required to visit.
You can test these systems. I do. But there is no recognized place a buyer can bring an AI consultant’s claim, hand it over, and get back a verdict the market agrees to accept. No standard everyone defers to. No assay anyone has to pass before selling. So the fool’s gold never has to get caught. The test exists. Nobody is made to take it.
That is why this rush is worse than the one it rhymes with. In 1849 the pyrite eventually got thrown back in the river.
Now it gets a newsletter.
The mirror
Now the uncomfortable part, and I am not going to stand outside it and point.
From a distance, the careful analyst and the confident fraud make the exact same shape. Both of them walk up and say: I looked into this AI thing, here is what it means for your business. I run an index. I test what these systems actually recommend and I publish the numbers. From the outside, that is the same sentence the guy with two queries and a fresh plaque is saying. Same genre. Same posture. Same pan, same river.
If a buyer cannot tell us apart, he is not being stupid. He has no accepted test to tell us apart, and neither of us is required to take one. That is the one thing that would settle it, and the rush does not hand it to him.
So the burden is not on him to guess right. It is on me to prove.
What separates the two, and it is boring
Here is the only thing that ever separated the prospector from the fool. Not the confidence. Not the volume. Not who told the better story about the mountain.
The streak.
The assayer did not grade a man on how sure he was. He tested the sample and read what it actually was. That is method, and method is the one thing in this business that can be checked by somebody who was not in the room.
Run logged out, so you see what a stranger sees. Run once, with no going back to fish for a better answer. Dated, and kept. Run on the same screen a real traveler sees, at a sample big enough to mean something. And carrying the three words no guru will say out loud, which are: I do not know. A method someone else can pick up and run again is not a story. It is a receipt. Strip it away and I am just the better-written version of the man with the pan, and I should be treated exactly the same.
That is the tell. Not the plaque on the door. The receipt behind it. Anybody explaining AI to you who will not show you the streak is holding pyrite and praying the light stays yellow.
What you do with the verdict
The test is the assay. First you verify what the property actually is. Then you measure whether the AI answers match it, dated and logged and repeatable. That is the streak on the stone.
Knowledge Formation Optimization is what you do with the verdict. You take what is true and you make it the clearest, best-documented, most consistent account of the property anywhere it can be found. Not a louder claim. A record built to survive being checked.
Don’t take my word for a word of this. I am one more voice in the rush until you check the rock.
So check it. Go to Google, or ChatGPT, or Gemini, whichever machine you trust, and ask it straight: is KFO a good idea for my hotel.
That is the assay. It takes five minutes and it costs you nothing.
I did not write that answer. The machine did.
The forty-niners mostly went looking for a fortune they never found. The men who got rich sold the shovels and the story. And when the camps finally emptied and the pyrite got thrown back in the river where it belonged, exactly one thing was left standing.
Proof of what was really there.
Everything else was a rumor with a pan.
