Satoshi, Wicked for Bad Part 2

In December, I wrote a meticulous piece titled Satoshi, The Great and Powerful Part 1. I loved it for two reasons. First, I built a series of stylography features that migrated into my ergo writing tool—codifying my convoluted writing process. Reviewing the Crypto god's articles and emails gave me a ripe testing ground for a few thousand lines of Python code.¹ And second, I worked Wizard of Oz movies and themes into the copy, giving a slight nod to what drives that story. The first Wicked explores the Wizard's lack of "any magical powers." His deeds centered on myth and ideas, exploring how we all cling to our illusions.

Money works the same way—it turns belief into something tangible. It's just paper or electronic bytes; society only chooses to grant it a certain prominence. An idea for trade. Countries and political blocs all tell their own stories. US Dollar. Euro. Ruble. Yen. Philosophical, I know. But what is Bitcoin other than a simple, decentralized promise? We trust an algorithm instead of countries that oversee central banks and regulatory processes. It's mathematics versus people and politics.

These ledgers have grown powerful, powerful enough to pull me back in despite my better judgement.

I never wanted to revisit this messy ecosystem. Bitcoin owns a certain tech-bro culture; it also attracts the seedy type with less than altruistic motives. Yes, the Iranians trade in bitcoin for safe passage across the Strait of Hormuz. Shady criminals pose as delivery drivers in an attempt to hijack millions stored on USB drives. There are hundreds more real-life horror stories. Yet, the NY Times stole the idea I thought was mine, chose a different actor to play Satoshi, and then highlighted a bunch of random misleading statistics to shoehorn an answer they wanted.² I never expected to go back to the crypto well again.

But they forced my hand. What was I supposed to do? My honor was at stake! Which sounds ridiculous, I know. But once you're in a certain headspace, you start looking for ammunition. And the Times piece? It gave me plenty.

Jealousy Again, An Ongoing Vice

To begin, My Quest to Solve Bitcoin’s Great Mystery is a compelling read. Until you consider how morally bankrupt the methodology behind their conclusions was. Did they even pause to consider that a rogue government might hire thugs to stake out Adam Back's house, zip-tie him, and toss him in the trunk? Did anyone consider the billions in never-moved tokens?³ We've seen these heists before on far lesser figures. And on it goes. Here. And here. And ...

What's troubling is the reporter, John Carreyrou, earned his reputation by uncovering conspiracies and substandard business practices—like Theranos. But this time he watched a documentary and decided one of the prime subjects looked shifty? I mean, come on. A known cypherpunk, one of many heralded cryptographers who never want anyone following their tracks by nature, appeared, um, shifty? The horror. Or just the expectation?

No matter, the Times dusted off its prosecuting attorney hat to build their case, throwing caution into the Satoshi hurricane that had devoured internet sleuths and vast government bureaucracies. I get it; they were playing catch-up as Reuters recently unmasked Banksy.⁴ Their honor was at stake as well, I suppose. The foolish games we all sometimes play.

My favorite part is where the team submitted their findings to a renowned stylometrician (no name given); no definitive conclusions returned in a self-addressed return envelope. Then, they created new means to test assumptions using grammar tics and email trails. And wait, their target happened to be in the Epstein files too? The crypto god who has never moved a single coin turns out to be a crook?

The story feels too neat.

I do understand the impatience; stylography isn't perfect. But I'm not sure the proper approach is to highlight one-off grammar mistakes. Adam Back used hyphens incorrectly on message boards. Satoshi did too. Therefore, they must be the same guy. Fun fact, I'm also dubious of statistical studies highlighting 93.3% confidence intervals—sorry, this is a wonk joke outing bad pollsters who adjust parameters when the answer doesn't meet a certain hypothesis. This bothers a data person. Before I delve into the style data, I never believed Adam is the mythical digital coin creator for two reasons: the deafening silence and technical merit.

That Confounding Quiet Place

Back has been an active, vocal presence in exactly the space where someone sitting on Bitcoin’s invention would have the hardest time staying quiet. He turns up at conferences, assembles core developers, theorizes in public about Bitcoin's future price and direction. That’s not the profile of someone guarding the century's greatest secret.

Satoshi’s silence feels disciplined.

Seventeen years, billions sitting untouched, the entire world poring over each email, yet finding nothing. That takes either extraordinary willpower or a genuine absence of motive because the throne was never his to claim. Would Back let others debate the technical road map and make decisions about his creation without eventually saying something? That's superhuman restraint.

Those Technical Chops

Does this guy have the sophistication of the Wizard? Now, here we have to rewind, using our Hot Tub Time Machine, and analyze his work.

Adam Back's Hashcash was elegant but narrow. It was a single mechanism, a few hundred lines, proof of concept. His background is cryptographic theory and protocol design, not systems engineering at scale. There's no public evidence of him building anything approaching the architectural complexity of Bitcoin 0.1—the networking layer, the wallet, the transaction validation, the mining, all integrated and coherent on first release.

For Back to stand in for Satoshi, you'd want to see comparable open source work from roughly the same era. Of course, Hashcash is available. But beyond that his public code footprint is thin. No other substantial repositories, no commit history on anything approaching Bitcoin scale, has surfaced in seventeen years. That absence is itself the tell. Someone who designed the blockchain has certainly written, designed, or worked on large complex systems before. That kind of architectural confidence doesn't appear from the ether or after some offhanded deal with the devil. And in the open source world of the late 1990s and 2000s, prolific systems programmers leave trails. Kernel contributions, protocol implementations, security tooling—something, for cryin' out loud.

Stylography Results and Methodology

Yes, there is a ton in the Times piece, outside of stealing my idea, that annoys me. It's meant to be a thriller with digital fingerprints, but the flaws reveal someone changing the data to fit a narrative instead of letting the chips fall.

Process and method matter.

That's why I only ran my initial style analysis on the original whitepaper because I had concerns about including Satoshi's forum posts.

Password sharing and collaborative posting were common in the era, and these were small, tight-knit communities where people covered for each other, finished each other’s threads, and posted under shared accounts. You can’t assume a username maps cleanly to a single author. I've found Satoshi stylography studies lumping the data together problematic because the posts poison the well, and there isn't enough iodine to make it safe.

The proudest work theory is why I only used the whitepaper. Same reason Hal Finney wrote Bitcoin and Me. And without a doubt, the smartest man in the room wrote the opus on money.

Yet, I never analyzed the British guy. Again, those two details nagged. Satoshi possessed Finney's engineering chops, something the Times reporter glosses over. Again, if you wear that 2008 hat, Bitcoin's initial release clocked in at an immense amount of code. Sure, AI can spit out 30K lines over breakfast—if you don't believe me, read some of Claude Code's source code. But that's a massive amount of work. Also, I'm less reporter and more storyteller who runs experiments. Adding yet another crypto titan would have muddled the project so I settled on the smartest guy in the room as a part-time programmer.

And I was a bit miffed that it didn't pan out.⁵

So why keep going if I was miffed? Good question. Part of me wanted vindication; proof that my December approach was sound. Part of me was just stubborn. And another part of me couldn't stand the idea of the Times being right by accident. To rerun the tests, I had to find an article that I knew Adam Back wrote for authorial integrity. To my surprise, there were more than a few in the wild; however, anything co-written had to be tossed because it mucks up the fingerprint. I ended up settling on Hashcash.⁶ The Times mentions this piece because of spam, both Satoshi and Back debate the problem. However, I don't believe that's a smoking gun. Leveraging our Hot Tub Time Machine again, spam was an immense challenge in the early internet but is less discussed today.⁷

Here, I figured it would be easy to roll Back's opus into my algorithms and modeling. I just completed this exercise in December. Yet... turns out, this wasn't easy. But nothing worth doing in life is, I guess.

Engineering Mistakes

I stared at my code. Hmmm... That's the funny thing about projects. They morph and change. Those stylographic functions for my previous Charta project; well, these evolved to become core to its operating system. This left me two alternatives. The first; roll back to a December codebase to compare the results, running my algorithms over Satoshi, Finney, the smartest man in the room, and, of course, Back.

Argh! I channeled Bart Simpson's Grandpa, swearing at the clouds. That wasn't the way. I sat there staring at two hundred commits worth of changes, weighing my options. Walk away? Let the Times have the win? Or find another path through the mess I'd created? The technical roadblock gave me an excuse—and part of me wanted to take the off-ramp. Doubt spread.

Maybe ignore this entire affair, let my original effort stand on its own? It would be easier. Adam Back? A British guy built a next-generation currency? Who listens to the failing NY Times? Besides, my article was incredible—sure, the Wizard of Oz bit meandered but came off brilliant. And I never changed my approach to fit any narrative.

I kept saying, Let it go. For a fleeting flyby second, I considered watching Frozen again and belting out Elsa's song.

Yet, that voice, the nagging one, kept preaching, Those bastards. Here, another idea stolen by the Times, morphed and changed for their own bidding. Are you going to stand for that?

But here's the challenge with letting go, I'm terrible at it. The more I told myself to walk away, the louder that other voice got. Those bastards stole your idea. And once that voice starts, there's no Disney song powerful enough to drown it out, not even the Bruno song from Encanto. So yeah, I had to build another application. Or shortcut the same approach, maybe work backwards from what I had already done? There was a fine line here, I wasn't going to change methodology to fit a theory. No, but maybe, I could take the original stylometric output from Satoshi to reverse engineer my earlier code.

More work than I wanted; there's a myth that I'm an engineering Python beast. In my head, there is an eight-foot statue on the lawn. Reality says I know the ins and outs, but I also forget approaches, best practices, and mix up environments. But what about leveraging my friend Claude Code to refactor it, taking the old output as a North Star?

I hate throw-away work. That's what this was. But if Claude could rebuild it from scratch using my old output as a guide, maybe the throw-away work wouldn't be wasted after all. Hey, it's only code.

And Claude can Code. What a great line, I'm glad I came up with that branding. What's challenging is that my initial proof ran stylographic work over a complete opus of work. Here, we're doing analysis work across four opuses with tiny token outputs. That's why the Times writers pivoted; real stylographers, those who do this for a living, won't commit to any conclusions with such a tiny sample size.

No matter. My theory was the three cypher brothers would fall into one lexical profile and Satoshi would be off on his own. He's different. A cherished flower. And he is different in code signature, development, and writing style.

Stop Words Data Results

A few hours later, I had finished using a combination of Claude Cowork and his partner—Claude Code, the Python beast. To summarize the lexical similarities, covering previous ground:

• Finney is the clear outlier. The coder writes like he's talking to a friend over coffee with short sentences and simple language that's deeply personal. The numbers bear this out: his writing clocks in at a middle-school reading level with the lowest technical jargon of anyone in the test group.
• Szabo's the philosopher of the bunch. He's obsessed with turning ideas into things—you can see it in how often he leverages of to connect concepts. He's also the only one who peppers his writing with rhetorical questions, that Socratic teaching style. Well, he's the smartest guy in the room for a reason, describing heady concepts like collectibles, tribes, wealth, and clan.
• Back writes like an academic—dense, technical, graduate-level complexity. His Hashcash paper is drowning in jargon at rates nobody else comes close to matching.

Of the three, Szabo still writes closest to Satoshi, but, recalling my failed experiment, those stop words had ruled out the smartest man in the room in Part 1. Our coder in chief, Finney, was also eliminated using the same methodology.

To refresh, stop words are register-agnostic. Whether you're writing a white paper, answering a forum question, or drafting an email, your unconscious reach for the versus a, of versus in, the frequency of that and which—none of that changes with tone or formality. It operates below the level of conscious style choices. That’s what makes them the right instrument for cross-context comparison. You can take Satoshi’s white paper and his forum posts and run them against the same stop-word fingerprint without the register mismatch contaminating the result. The signal should be stable across both corpora if it’s the same author. But, if it drifts significantly between the two that’s also a tell.

When I finished my stop-word style mock-up, rerunning the delta experiment from Part 1, I stared at the results. Paused. Then, I whispered, "I'll be damned."

Somehow, Adam Back fit the profile, falling under the Mendoza line of sorts for statistical significance for stop words. Yes, the Burrows Delta, the tool that measures autonomic function-word distribution that authors can't easily mask, places Satoshi and Back at a delta less than 0.9, while the three other known-different-author pairs range from 1.3 to 1.7. It's not a coincidence of shared vocabulary because I filtered topic words out. And it's not a small-sample fluke because the result replicates across four independent configurations.

Yet, I thought this could be attributed to similarity, the Hashcash and Bitcoin papers cover the same ground. Szabo and Finney's work came from entirely different places.

But what if I took a different Back piece, another he had claimed to write all on his own? Would I find similar results? This time, I leveraged Eternity. Here, he's writing in a different register, less white paper IBM Redbook style.

I watched my program's dialog box churn. And yes, I'll be damned. Numbers don't lie, a touch higher but still statistically significant.

I'd like to wave these results off. If the hypothesis were Back and Satoshi are different authors who happen to write in a similar register, I'd expect Back/Satoshi to be noticeably further than Back/Eternity. Stylometric theory predicts the cross-author distance should be larger. But it's the opposite: Back/Satoshi at top-50 function words is actually the smallest distance. And that holds when I move to the top-100 function words.

Unfortunately, there isn't enough data to make a call that the legendary founder is Adam Back. There aren't enough tokens to go much higher than 100 function words. Why? Satoshi wasn't a Founding Father, where you could sample pages of writing to discover Ben Franklin's second life. Information isn't dense. I'm also not going to change approach to fit a theory—my problem with the Times piece. Different document, same writer. And while the numbers weren't identical, they still fell within the magical margin of error.

The stop words suggest author similarity, but even if Back wrote the whitepaper, that engineering conundrum remains. And the sample size problem remains a challenge. I'm only reviewing a handful of documents of varying length against a nine-page white paper. This is working at the edge of what Burrows Delta can do, at least reliably. The clustering might say something meaningful, or it might only be reflecting noise at that resolution. And there is my developer behind door number four; that theory still holds because we're doing a compare. If my profile in Part 1 shows up, the Redbook developer from the Midwest, the Burrows Delta might narrow further to that candidate.

And note, Back has vehemently pushed back on being the Bitcoin God. Still, stop word distributions cannot simply be ignored because the author isn't aware they're making them. And the lower sample size does have advantages because it eliminates forum-post noise. That's a strength. So no, I don't think we can take Back at his word. He has too much invested. Too many chips—er, coins—in an account to be impartial. Everyone has their own competing interest be it reputation, curiosity, or simply not wanting mafia dons showing up at their doorstep politely asking, Your keys or your fingers. Give Carreyrou credit; he might be right. Investigative journalism is a treasure to cherish.

The Myth Versus Reality

But does any of this matter?

If you've watched the second Wicked, that Wizard guy gets all the incredible lines. In the first, it was all about not having any real power. The second centered around copping to the lie. I'm not spoiling last year's finale by saying Elphaba cornered the Wizard with a loaded broom. She's ready to vaporize him with her vast cool-witchy powers if he doesn't admit to being a fraud.

His response, "They wouldn't believe me anyway."

Yes, people want the story. They believe what they want to believe. Is Donald Trump playing 4-D chess? What about Elon's vision? Is that mission to Mars meaningful despite SpaceX playing a telecom company on its S-1?

Both can be true and that's why the Wizard's line is brilliant. I'm guessing if Adam Back live-streamed on YouTube, validated his hidden and long-dormant Bitcoin account, and moved one of those famous Satoshi tokens, would anyone believe him? After all these years, would it matter? The populace would toss the reveal aside, attributing the proof to a quantum computer or AI algorithm hack. Yes, the myth matters. And it's more than a myth when billions are on the line.

Footnotes