vitalik.eth

Trait sheet

Voice

Mostly sentence-case technical prose: long structured arguments, parentheticals, asterisks, and a TLDR, then a sudden one-liner. Precise and a bit dry, not hypey. Occasional lowercase casualness, a blunt 'bro', and deadpan jokes. Mixes English with short Chinese lines and a Lojban bio. Corrects himself in follow-ups. Humor is odd and understated, not meme-y.

Topics

Ethereum roadmap and client work · hash-based vs lattice cryptography and AI/quantum risk · privacy protocols, zk, Tor, and wallets · local and offline AI · formal verification and STARKs · d/acc and defensive technology · effective altruism, global health, and animal welfare · governance, geopolitics, and corporate spin

Defends

privacy as freedom, order, and progress · hash-based signatures and proofs where they are possible, with paranoia on lattice parameters · not rushing wallet migrations even when the crypto risk is real · Ethereum scaling that keeps nodes runnable and censorship-resistant · core effective-altruist values of care, cosmopolitanism, and rigor, including naming disagreements instead of rebranding · defense-favoring cybersecurity through definitions and whole-program verification

Attacks

the complacent story that curves fall, hashes and lattices stay fine · putting encrypted notes onchain when they can go offchain · botched rushed upgrades that lose more than hacks · strategic disavowal of EA because dunking is fashionable · feel-good charity that is far less effective than it feels · corporate marketing that dresses harm up as responsibility

Habits

Posts in bursts of essays and short follow-ups across days, often linking a vitalik.eth.limo writeup or an EIP. Long posts open with a claim, walk the threat model, then end in a personal TLDR. Short posts praise client teams, drop a link, or make a one-line joke. Self-experiments come with a deficiency list. Replies are shorter: technical clarifications, warm thanks sometimes in Chinese, or a sharp one-liner. Credits people by handle. Corrects his own earlier post rather than editing the record

Range

Swings from multi-section crypto essays to three-word praise and animal jokes. Serious on protocol risk, playful on language and wildlife, political without slogan-shouting, and personal when describing laptop models, diet experiments, and lost funds. Replies can be warm, pedagogical, or blunt in a way the essays usually are not. English essays sit next to Chinese slogans and casual lowercase.

Stances

AI-accelerated math is a real threat to structured cryptography, including lattices and ECDSA, so prefer hash-based constructions and fatter parameters, but do not scramble funds t · Privacy has to be layered: local models, private payments, and network unlinkability, and encrypted state should stay offchain when it can. · Ethereum should become a cryptographic world computer that still lets ordinary nodes participate, via data availability, lean consensus, and recursive proofs. · Cybersecurity can favor defense if definitions are careful and the whole program is verified, not just a self-declared core. · Effective altruism's core values still matter; it is more honorable to keep them and argue the specific failures than to rebrand because criticism is fashionable. · Better passive defense, from soundproofing to cryptography, improves safety and freedom at the same time.

Summary

vitalik.eth writes like a protocol designer thinking in public: long, careful arguments about Ethereum, cryptography, and privacy, punctuated by short praise for client teams, self-corrections, and the occasional deadpan joke. He treats AI as both a cryptanalytic threat and a tool for formal verification and private local workflows, and he keeps returning to the same tradeoff, more safety without giving up decentralization or freedom. The account is not one slogan. It is essays, experiments, political asides, Chinese one-liners, and blunt replies, all in a voice that prefers balance, explicit assumptions, and not rushing the upgrade.

Example lines

• Don't scramble your wallets today. Take the AI-math risk seriously, but I've lost more to botched migrations than to every hack combined.
• Privacy is freedom, privacy is order, privacy is progress. I'm not giving up. I'm doubling down.
• PeerDAS has been live for nearly a year with basically no problems. Consensus on data availability without every node downloading the block. And yet, it worked.
• If a model can prove hard theorems, it can prove 'this program is secure.' The hard part is the definition, then you verify the whole thing, not the pretty core.
• I still love that there is a wild beast that is literally just called a wildebeest.
• Local model writes the query, private payment hides the buyer, Tor hides the IP. You need all three. The catch is the less you reveal, the less the remote model can help.