One Redhead, Four Kids,
and Way Too Many Coins

I didn’t set out to make software. I’m a photographer (coins are one of many subjects) and I deal coins, which means I already spend my days doing what this program does: lighting a surface until it gives up its secrets, deciding whether a line was struck into the metal or scratched onto it, and then saying so out loud to somebody who’s about to spend money on my answer. No pressure.

The coin tools I could find came in two flavors. Some cost more than the coins I was pointing them at. The rest handed me confident answers that weren’t true, and that flavor bugs me the most. A program that calmly says no doubling detected about a coin with obvious doubling hasn’t failed politely. It’s told you something false in a nice voice, and if you’re a dealer, you’ll eventually repeat it to a customer.

So I wrote my own. Then I kept going. Then I kept going some more. If there’s even a small chance a feature helps uncover what’s hiding on a coin, and the physics says it’s possible, it goes in. You can see where this is headed.

The Science Part (My Favorite Part)

Here’s the thing that changed everything for me. Brightness on a coin is shape multiplied by lighting, and one photograph can’t divide the two. A shadow beside a letter and a genuinely doubled letter make the same repeated edge. Nine attempts to read doubling out of a single frame failed on this bench, and they weren’t badly written. They were asking a question the evidence doesn’t contain.

So the program asks different questions. Move the lamp between shots and each point on the coin brightens and dims in a pattern that gives away its tilt; that’s photometric stereo, and it hands back the actual height of the metal. Turn the focus knob and every point is sharpest at one lens position, which is a height too, measured a completely different way. Put a polarizing filter on the lens and turn it, and Fresnel’s 1823 physics says the polarization of the reflected light carries the tilt of the surface. Three methods with different weak spots. When they agree, you can believe them.

And once you have real height, the old arguments turn into geometry. Hub doubling leaves two rounded crests at about the same height, because the die had two impressions to fill. Machine doubling leaves one crest and a low, flat shelf of shoved metal, because nothing was struck twice. A die crack stands up from the field; a scratch sinks into it. That’s the stuff I can’t stop nerding out about.

Why It Works the Way It Does

Most of what’s quirky about this program comes from one fact: I have to live with its answers.

Every tool that guesses says so. The tools that measure and the tools that infer live on separate tabs, and the inferring ones wear a caution color. When the doubling screen finds a second image but can’t pin down the gap, it tells you what it found and skips the distance, because a made-up number dressed as a measurement is worse than no number. That part isn’t a limitation to me. It’s the whole point.

Your photographs are not watermarked. They’re your photos. I’d be furious if somebody stamped a logo across mine, and I’m a redhead, so furious is a whole event.

I never see your coins. A copy does check in with my server, once, to start the trial or activate a key. I’m not going to pretend otherwise. That check carries the key, the emailed code and a fingerprint of the machine, and that’s it. No photos, no collection, no measurements, no record of what you were looking at. No telemetry, no analytics. The only other thing that ever leaves your computer is a bug report, and only because you typed it and hit send.

There’s one price, paid once, because that’s what I’d buy. $199.99, and the license never expires. Updates come with it for the life of the product, it goes on three computers, and nothing bills you again, ever. Owning what you paid for shouldn’t be a premium option.

It won’t hold your work hostage. The key is checked once and then the program runs offline for good, so there’s no server of mine to hiccup on you. If the trial ends and you don’t buy, it stops working except to enter a key or export your photos and database. Coin photos are often a one-shot deal (the coin gets sold, slabbed or rolled back up), and locking you out of pictures you can’t retake would be a rotten thing to do to a person.

Dealer and Collector Are Not the Same Person

I’m both, and they want different things. The dealer in me needs to know what a coin’s worth and wants it photographed, graded, labeled and listed by Thursday. The collector in me wants to know what it is: which die, which variety, which stage of that die’s life, and what happened at the mint the day it was struck.

That’s why the collection half of this program is as big as the microscope half. A tool that only examines coins leaves you with a folder of gorgeous photos and no idea which coin they belong to. So every photo is filed to a coin record that also holds the grade, the cost, the current value, the provenance and the actual shelf it sits on, and it’ll print the 2x2 label with a barcode so the shelf and the software agree.

The Part Nobody Asks About

I like staring at code. That’s the honest explanation for the size of this thing. The doubling detector took five rewrites before it would tell a real 1955 doubled die from an ordinary cent, and then a sixth pass to figure out that its measurement was off even though its verdict was right. That’s not a business plan. That’s somebody having a great time. (Also ADHD. Hyperfocus is a heck of a drug.)

Four kids is part of the explanation too, in a way anyone with kids will get. I have zero patience for software that takes four minutes to open, makes you sign in every time, or interrupts you to announce a feature. The program starts, the camera goes live, and the coin’s on screen.

Where It’s Going

Forward, steadily, and out in the open. Features get added because a real coin needed them, usually one sitting under my own lens that the program couldn’t help with yet. If there’s something you want it to do, tell me. That’s honestly how the list gets written.