Vitalik Buterin got caught. The Ethereum co-founder confirmed that AI-assisted analysis correctly identified his anonymous contribution to an Ethereum proposal — ending a two-week public challenge he’d thrown down.
It started on June 22. Buterin asked whether AI tools could pierce online anonymity. He revealed he’d published a document of “medium importance” to Ethereum at some point in the past decade under a different name. The challenge: find it.
Franklyn Wang, CEO of Co-Invest, took the bait. His winning submission identified an anonymous rewrite of EIP-7503 — not by analyzing word choice, but by studying how it explained mathematical and technical concepts.
Here’s the kicker: Buterin had written the anonymous version in Chinese, machine-translated it to English using Qwen 2.5, and manually corrected the translation to disguise his prose. It didn’t matter. Wang’s analysis picked up on Buterin’s intellectual habits — the way he explains math and algorithms.
Co-Invest ranked Buterin as the most likely author with about 20% confidence. That was roughly 10 times higher than the next candidate out of 27 documents analyzed.
This has bigger implications. Some of crypto’s most prominent figures — including Satoshi Nakamoto — relied on pseudonyms to stay hidden. If AI can reliably match authors by reasoning patterns alone, anonymous technical contributions across open-source blockchain communities could get a lot harder to pull off.
Researchers from ETH Zurich and Anthropic published a paper in February claiming large language models have made online deanonymization practical at scale. The study found AI could identify pseudonymous users by extracting identity-related information from text and reasoning over likely candidates — outperforming traditional methods.
Buterin’s own summary says it best: the stylistic hints the AI picked up were intellectual habits, which his prose-level obfuscation strategy didn’t even touch.
