Artificial intelligence supposedly knows everything about us. But right now, the most advanced language models on the planet are failing a surprisingly simple test.
Thirteen days ago, Ethereum co-founder Vitalik Buterin issued a public challenge. He dared the internet to use AI text analysis tools to find a single, moderately important anonymous document he authored somewhere between 2020 and 2026. As of Sunday, July 5, 2026, the global tech community is completely stumped.
The experiment specifically tests stylometry. That is the science of identifying an author based on their unique linguistic quirks, syntax, and vocabulary choices.
Buterin is practically the ultimate boss fight for this kind of test. He has millions of publicly attributed words across blogs, whitepapers, and forum posts for an AI to train on. Yet the machine learning scripts scanning the internet keep coming up empty.
He finally broke his silence over the weekend. Buterin posted an update confirming the 13-day failure streak and offered a breadcrumb. The Ethereum creator advised participants to slightly broaden their document parameters, noting that current search efforts entirely omit specific categories of writing that should definitely be included in the scan.
The Threat to Open Source Identity
Why does an anonymous document matter? The entire cryptocurrency ecosystem was built on the philosophical foundation of absolute privacy.
Decentralized networks rely heavily on pseudonymous developers and researchers who build critical financial infrastructure without attaching their legal names. Bitcoin itself was created by the anonymous entity Satoshi Nakamoto. If an open-source AI tool can suddenly de-anonymize crypto contributors by simply reading their code notes and essays, that foundational myth crumbles.
What the AI Failure Means for Digital Privacy
The stakes here go far beyond a billionaire’s weekend puzzle.
Stylometry has been used in academia for decades to uncover historical authors. Generative AI supercharged that capability. If these tools eventually crack Buterin’s challenge, it hands government regulators and law enforcement an unprecedented weapon. Agencies would no longer need to wait for platform cooperation, subpoenas, or voluntary disclosures to identify whistleblowers or anonymous developers.
They could simply deploy AI models to run reliable authorship attribution in seconds.
But the current 13-day failure actually offers a glimmer of hope for privacy advocates. It suggests that even with enormous datasets and cutting-edge neural networks, human writing can still hide in plain sight. At least for now.