Daniel's been wandering around the internet again, and this time he found a walkie-talkie.
A walkie-talkie. That's the hook.
Not just any walkie-talkie. A Motolola. M-O-T-O-L-O-L-A. No r. He spotted it on AliExpress and it stopped him cold, because somebody sat down, designed a logo, printed packaging, and shipped a product betting that a fraction of buyers would never register the missing letter.
Which is the whole trick, isn't it. Your brain sees the shape, fills in the gap, moves on.
Right. And that sent him down a rabbit hole. He wants the most ridiculous examples of brand clones out there. He wants the techniques, the ones that let a knockoff look legitimate without quite crossing into outright counterfeiting. He wants the hierarchy, from the ones that don't even pretend to pass a close look up to the ones that push the art form right to the edge of what the law allows and maybe a step past it. And then the big one: have there been cases where a knockoff was so convincing it took consumers a while to realize they weren't buying the real thing at all.
That last question is the one with teeth.
So today we're diving into the weird world of brand clones, knockoffs, and the gray zone between inspiration and infringement.
The gray zone is where all the interesting engineering lives. The blatant stuff is boring. Motolola is funny for about four seconds. What's actually fascinating is the calibration, the exact amount of distance a knockoff keeps from the real thing so it trades on recognition without handing anyone an easy trademark case.
Calibration implies somebody measuring. Which implies a whole discipline behind this.
There is. It's a spectrum, and it's worth laying out because the tiers behave completely differently. Tier one is the blatant familiarity play. That's Motolola, that's the walkie-talkie. Low effort, low cost, catches the inattentive fraction and nobody else. Tier two is the near-miss mutation, one letter swapped, a vowel changed, a character added. Slightly more care, still cheap. Tier three is trade-dress mimicry. Different name entirely, but the packaging, the color scheme, the typography, the silhouette of the bottle or the box is near identical. That's the Aldi model.
And tier four?
Tier four is the functional counterfeit. Same barcode, near-identical packaging, sometimes differing only in where it says Made In. At that point it isn't a clone anymore, it's a copy, and it's designed to move through the same supply chain as the real thing.
So the hierarchy runs from a joke to a felony.
Roughly. And the legal pressure runs the other direction. The closer you get to tier four, the more you're playing with actual criminal exposure. Which is why the smart money sits in tier two and three, where you're annoying but hard to nail.
Hard to nail how? Trademark law seems like it should cover this.
Trademark protects against confusion, and confusion is a spectrum, not a switch. You have to prove a likelihood of confusion, and the further your name is from the original, the harder that gets. There's a whole body of law around what counts as confusingly similar. And the brands know this, which is why they police the edges so aggressively. eBay's legal department sent cease-and-desist letters over domain names that merely ended in the word bay. Not eBay, just bay. Theplayersbay dot com got a letter, because eBay's lawyers argued that bay was a coined term and eBay was one of the most famous brands in the world.
They tried to own a three-letter word.
They tried to own the syllable. And that tells you something important, because if a legitimate brand will go after a domain name that shares three letters, imagine what they'd do to a company selling Motolola radios. The knockoff makers know exactly where that line is. They're not guessing.
So let's start with the techniques. How do these brands actually pull this off?
The foundational one is the orthographic near-miss, and it works because of how reading actually functions. You don't read letter by letter. You read in chunks, you pattern-match against words you already know, and your brain is aggressively lazy about filling in gaps. Motolola is close enough to Motorola that the visual system resolves it as Motorola before the conscious mind ever inspects the letters. By the time you'd notice, you've already clicked buy.
And the fraction that does notice just doesn't buy. The seller doesn't care. They're not trying to fool everyone, they're trying to fool the slice that doesn't look closely.
Exactly the economics. You're not converting the careful buyer, you're harvesting the careless one. And the careless one is a big enough market that it's worth printing the boxes.
What else?
Trade-dress mimicry is the more sophisticated version, and Aldi is the canonical example. Walk into an Aldi in the UK and you're surrounded by products that are almost, but not quite, the brands you know. Same color palette, same font weight, same layout on the shelf. A commenter described their stock as mostly off-brand lookalike items, usually dangerously close to a trademark infringement. It's the whole business model in two words.
Dangerously close but not over the line.
Not provably over the line. Which is different. And then there's the workaround technique, which is my favorite because it's so brazen. On eBay, sellers learned to prefix listings with the word For. For Apple AirTag. Then they'd black out the Apple logo in the product photos. Technically, they're describing compatibility. In practice, they're selling you a counterfeit and using the brand name to get found in search.
So the word For is a legal shield.
It's a fig leaf. It's thin, but it's enough to survive a first pass, and by the time a takedown lands the seller has already rotated to a new account.
Barcode cloning is the tier four move. Explain how that works, because that's the one that gets into the supply chain.
A barcode is supposed to be the identity of a product. Same barcode, same item, as far as any warehouse system is concerned. So a counterfeiter who clones the barcode and the packaging doesn't just make a fake product, they make a product that the logistics system cannot distinguish from the real one. It gets scanned in, it gets shelved, it gets shipped, and nobody in the chain has any signal that anything is wrong.
Which sets up the case that answers Daniel's last question.
The Kickstarter precision optic. A company raised money to build a precision optical instrument, shipped it, and then discovered that counterfeit units with the same barcode and nearly identical packaging had gotten co-mingled into Amazon's inventory. Amazon itself drew from the counterfeit stock when fulfilling its own sales. Customers started leaving bad reviews because the photo quality was worse than expected, and the real company watched its review score collapse because of units it never made.
So the failure signal was photo quality.
Photo quality. That's the only tell. The packaging was right, the barcode was right, the product looked right. It just didn't perform, and performance is the last thing anyone checks.
We've seen how they do it. But what happens when they do it so well that nobody notices?
That's where it stops being funny. The Apple accessory numbers are the cleanest example. In Apple's own testing, over ninety percent of Apple accessories sold online were fake. Chargers, cables, the stuff you plug into a wall and leave unattended. And the critical detail is that many of them are outwardly identical. You cannot tell by looking. You'd have to tear them apart to find where the corners were cut, and where the fire risk was introduced.
Ninety percent. That's not a fringe problem, that's the default state of the market.
And consumers weren't returning them. They were paying brand-name prices for what was, in some cases, about four dollars of hardware, and they were satisfied, because the thing worked, at least for a while. The deception only surfaces when it fails, and by then the purchase is long over.
Why couldn't people just buy from a reputable seller and be safe?
Because of co-mingling, and this is the structural part that matters. Amazon stores inventory for the same SKU in a single bin, regardless of which seller it came from. So if three sellers list the same product and one of them is selling counterfeits, all of that stock goes into one bucket. When you order, the picker grabs from the bucket. You can deliberately choose the legitimate seller, pay their price, and still receive a counterfeit, because the physical unit you get was never tracked back to the seller who sent it in.
So the legitimate seller is punished for the counterfeiter's crime.
Systematically. And there's no way for the buyer to defend against it, because the information needed to defend yourself doesn't exist at the point of sale. That's why Amazon launched Project Zero in 2019, which let brands remove counterfeit listings themselves. Read that as an admission. The platform concluded its own policing wasn't working and handed the job to the brands.
Which is a strange thing for a marketplace to do, hand moderation to the parties with a financial interest.
It's the least bad option when your detection is failing at scale. And the review system, which is supposed to be the buyer's defense, has its own problems. A 2025 IEEE study built a detection model on twenty-one thousand user-uploaded images and hit an F1 score of zero point nine three four. That number is impressive, but think about what it implies. Detecting fake reviews now requires multimodal AI at that level of performance. Which means the fakes are good enough to fool humans at scale.
The defense had to get that sophisticated because the offense did.
Exactly the arms race. And here's the counterintuitive part, the thing that complicates the whole moral picture. Tom Ford, after looking into counterfeiting of his own products, concluded that the people buying cheap counterfeits weren't his customers anyway, so he wasn't losing anything.
That's either very clear-eyed or very convenient.
It can be both. There's a real argument that a person buying a fifteen-dollar knockoff was never going to buy the fifteen-hundred-dollar original, so no sale was lost. But that logic breaks down the moment the counterfeit is being sold at full price as the real thing, which is exactly the Apple accessory case. That's not a lost sale, that's a stolen one. Ford's framing only holds for the discount counterfeit, not the full-price one.
The distinction is whether the buyer knows.
Whether the buyer knows, and what they paid. A knockoff sold cheap to someone who knows it's a knockoff is a substitute product. A counterfeit sold at full price to someone who thinks it's genuine is fraud, plain.
So the harm isn't in the copying, it's in the deception.
And the deception is what scales. The copying is almost a side effect.
One thing I keep circling back to. Daniel covers up brands on his own products, he doesn't want to be a walking advertisement. And there's something almost funny about that instinct running headlong into a market where the entire value proposition is pretending to be a brand you didn't pay for.
Two opposite reactions to the same logos. He's paying to remove them, and someone else is paying to fake them.
Which suggests the logo isn't really about the product at all. It's about what the buyer wants to believe about themselves.
That's the whole engine. The knockoff doesn't sell a radio, it sells the feeling of owning a Motorola. And that feeling is worth more than the radio, which is why the margin exists.
Let's talk about what happens when this gets worse. If AI can generate packaging artwork and product photography and review text, the cost of building a convincing tier three clone just collapsed.
It collapsed to near zero. The bottleneck used to be design labor. You needed someone who could redraw a logo convincingly and lay out a box. Now you describe it and it appears. And the same tools that make the clone make the detection harder, because the tells get subtler.
So the arms race accelerates on both sides.
And the platform is stuck in the middle, because the platform's incentive is volume, and every counterfeit listing is volume. Project Zero exists because Amazon decided the reputational cost of fakes finally outweighed the transaction revenue. That calculus can shift back.
Hilbert, you've been working the board this whole time. You've got something.
Hilbert: The Motolola thing isn't a knockoff.
Go on.
Hilbert: A knockoff implies there's a real product underneath being copied. Half those radios are generic. Same board as a dozen other brands, same factory, maybe the same factory that makes the real ones on a different line. Somebody just needs a name to print on the box, so they pick one that sounds familiar. There's nothing to copy. It's a generic radio wearing a costume.
That's a real distinction. The name is the product.
Hilbert: The name is the only thing they made. I spent a summer at a customs seizure warehouse, sorting confiscated goods. That was the job. You'd open a container and log what was inside. And the thing that surprised me was what actually got seized. Phone cases. Handbags. Not electronics. I don't think I logged a single radio the whole summer.
So the hierarchy you hear about and the hierarchy that shows up in a seizure warehouse are different things.
Hilbert: Customs catches what's easy to catch. A shipping container full of handbags with the wrong logo on them is a straightforward seizure. A generic radio with a nonsense name on it isn't counterfeiting at all. It's just a bad name. So the enforcement data is skewed toward the obvious stuff, and the clever stuff sails right through, because there's nothing to charge anyone with.
That reframes the whole tier system. The tiers aren't ordered by how deceptive they are, they're ordered by how legally exposed they are.
Hilbert: Which is almost the opposite. The most deceptive ones are the safest, because they're not copying anything. They're just selling you a radio with a name you misread.
What else was in the warehouse?
Hilbert: Sneakers, mostly. There was a shipment of Pumma sneakers. P-U-M-M-A. And here's the thing, they were well made. Proper stitching, real leather, held up. The staff bought them at the confiscated goods auction. I had a pair for years. Outlasted my actual Pumas.
The fake outlasted the real one.
Hilbert: The real ones fell apart in about eighteen months. The Pummas went four years before the sole went. I still think about that.
That's the part that breaks the mental model. We assume counterfeit means worse. Sometimes it just means cheaper to produce and sold under a name that isn't yours.
Hilbert: Most of the time it's worse. But not always. And when it's better, it's because the same factory made both, and the only difference was the label and the price.
So the quality signal people rely on, the brand, is doing less work than they think.
Hilbert: The brand tells you who's accountable. That's what you're paying for. The accountability. If the Pumma falls apart, there's nobody to call. If the Puma falls apart, you've got a company with a phone number and a reputation to protect. That's the actual product.
Which is why the co-mingling problem is so corrosive. It severs the link between the price you paid and the accountability you bought.
Hilbert: Anyway. I've got to move the trailer before it gets dark. The tarp's loose and there's rain coming.
So as we wrap up, what does this all mean for the future of trust online?
The thing I keep coming back to is that trust in e-commerce is becoming structural rather than individual. It used to be that you avoided fakes by being careful, by buying from sellers with good ratings, by checking the reviews. Every one of those defenses now has a hole in it. Co-mingling means the seller's identity doesn't determine what you receive. Review fraud means the ratings don't tell you what they used to. And AI-generated packaging means the visual inspection that used to catch the obvious stuff no longer works.
So the careful buyer and the careless buyer are converging on the same outcome.
Which is the actual story. Not that knockoffs exist, they always have. It's that the mechanisms we built to route around them are failing at the same time.
The most common wrong belief here is that a knockoff is always a cheap, obvious fake you can spot if you look closely. The Apple numbers kill that. Ninety percent of the accessories were fake, and most of them were outwardly identical. Looking closely doesn't work when the difference is inside the charger.
And the second wrong belief is that buying from the legitimate seller protects you. Co-mingling means it doesn't. The bin doesn't know whose unit is whose.
The practical takeaway isn't to look harder, it's that looking was never the mechanism that protected you. Accountability was. And that's the thing the market has quietly stopped delivering.
Which leaves an open question. As the tools to fake packaging and reviews get cheaper, does the line between real and knockoff blur to the point where it stops mattering to buyers? Or does detection AI stay ahead of it? I don't know. The 2025 detection numbers are good, but good at detecting yesterday's fakes.
That's the race, and it's not obviously winnable.
It isn't. But the alternative, no detection at all, is worse.
Thanks as always to Hilbert Flumingtop for producing the show, and for the Pumma sneaker story, which I will be thinking about for a while.
Four years. That's a long time for a shoe with a typo on it.
This has been My Weird Prompts. If you've got a weird prompt of your own, or you just want to tell us about the strangest knockoff you've ever bought, email us at show at my weird prompts dot com. And if you enjoyed this one, a rating and a review helps other people find the show.
We'll be back soon.
See you tomorrow.
