Grok-Assisted Stylometry Finds Little Support for Musk-Satoshi Theory

What the GitHub analysis set out to test

A new open-source project on GitHub has examined a longstanding claim that Elon Musk created Bitcoin under the name Satoshi Nakamoto. The researcher combined Grok, the chatbot developed by Musk’s company xAI, with stylometric techniques: statistical comparisons of habits such as spelling, punctuation and common-word use. The project tests whether Musk’s writing resembles Satoshi’s; it does not reveal a confirmed identity or establish who authored Bitcoin’s original texts.

The Musk theory has circulated for years, and the source notes that a former SpaceX intern made a claim about it in 2017. Musk has repeatedly denied being Bitcoin’s creator. In a 2017 post, he said a friend had sent him some bitcoin but that he did not know where it was. The new comparison therefore examines a public theory against writing samples, rather than reporting a new claim by Musk or evidence that he made one.

Stylometry can compare patterns that are less obvious than subject matter, including preferred spellings, punctuation and the frequency of function words. Here, the researcher tested three features frequently cited in arguments about Musk and Satoshi: spacing after periods, use of British spellings and similarity in common-word patterns. The project’s findings challenge those specific stylistic arguments, but the method compares available texts and cannot independently determine the person behind the Satoshi name.

Three writing habits point away from Musk

The sharpest contrast in the reported results concerns spacing after full stops. Satoshi used two spaces after periods in 95.3% of the instances measured, while Musk’s raw X posts contained no examples of that habit. On its face, the difference weighs against a match on this feature. The report also cautions that smartphone software can automatically insert double spaces, making the habit an imperfect authorship marker and limiting how confidently it can be interpreted.

Spelling patterns also differed substantially in the comparison. Satoshi’s writing was mixed, with British forms accounting for 32.7% of cases, while the analysis found British spelling in only 0.65% of Musk’s writing. It also found that Musk’s earliest testable texts, dating to 2006, already used American spelling. That result does not support an explanation in which his early writing showed a British style that later disappeared, though the study’s early sample is small.

A third test compared function-word patterns, focusing on common words rather than distinctive technical vocabulary. Among 66 writers, Musk ranked 30th in similarity to Satoshi, which the project described as no closer than a typical control writer. The repository also counted Satoshi using “e-mail” 17 times and “email” three times; Musk used “email” 65 times and never used the hyphenated version in the examined material. Together, these measures show differences, not proof of identity.

Comparisons offer context, not identification

The repository’s author reports that the method correctly attributed Satoshi text samples in 27 of 30 tests. That result gives the comparison some weight within the project, but it should not be mistaken for a definitive forensic validation. The reported accuracy concerns those sample tests; the source does not present it as a guarantee that the method can identify an unknown author under every condition. The distinction matters when assessing how much to conclude from the Musk comparison.

Other names in the analysis provide a limited point of comparison. Cypherpunk Hal Finney scored closer to Satoshi’s writing style than Musk did, while Nick Szabo scored further away. Bitcoin advocate Brad Mills was also tested using X posts and ranked 52nd among the 66 writers. These relative positions show how the tested writers compared within this dataset; they do not establish that Finney, Szabo or Mills was Satoshi, nor do they settle any competing identity theory.

The source identifies a genre gap as a central limitation: Satoshi’s available writing consisted of emails and forum posts, whereas Musk mainly writes tweets. Differences between those formats can affect punctuation, spelling and word choices, so not every measured contrast necessarily reflects an author’s stable habits. The comparison is therefore between different kinds of material as well as different writers. That makes the rankings useful as evidence about the samples, but less conclusive as a direct test of authorship.

Why the result remains inconclusive

Sample size is another constraint. The analysis found only eight examples of Musk’s spelling before 2013, a small basis for assessing whether his earlier writing differed from his later posts. The source also warns that phone formatting can weaken the double-space test because devices may automatically convert two spaces after a sentence. These qualifications do not erase the reported differences, but they limit the reach of each individual clue and make broad claims about a consistent writing fingerprint uncertain.

The repository’s README characterizes the work as a hobby-scale project rather than forensic linguistics. That description sets an important boundary around its conclusions: the analysis can test whether selected habits align, but it cannot prove or disprove who Satoshi Nakamoto is. Its reported comparisons make the Musk theory less persuasive on the specific writing features examined, while leaving open the possibility that sample selection, genre and formatting affect the results.

For readers, the practical takeaway is that an AI-assisted writing comparison adds a structured check to an old theory, not a resolution of Bitcoin’s authorship mystery. Musk’s spacing, spelling and function-word results were not close to Satoshi’s in this study, and other writers also produced varying scores. Yet the limited early sample, different writing formats and non-forensic scope mean the evidence remains qualified. The source’s conclusion is appropriately narrow: Satoshi’s identity remains unknown.

Source: BeInCrypto