Research build. Archetype and MCM tags below are rule-based narrative flavor, not a validated behavioral model. See Methodology.
Chronalcrypt™ Labs — Final Edition

The Vault Diary

Point it at a public wallet. It reads normal ETH transaction history and renders a first-person account of what the ledger looks like it's doing — narrated, tagged, and held to an honesty layer that says exactly how much to trust the tags.

CHRONALCRYPT™ the engine
CHRONALCRYPTEX™ the brains
The Vault Diary the voice
Demo mode — use synthetic sample data, no wallet needed
Advanced: connect your own proxy ▾

Without this, real-wallet lookups aren't possible in this static build — the app has no server of its own and doesn't ship an API key. Deploy proxy/etherscan-proxy-worker.js (included alongside this file), edit its ALLOWED_ORIGINS to this app's exact deployed URL before shipping, then paste the worker's *.workers.dev URL here — the app only accepts HTTPS *.workers.dev endpoints by default. Demo mode works with no setup.

Based on normal ETH transactions only (Etherscan txlist). Token transfers and internal transactions aren't included.

Share on X

Methodology & honesty layer

What this build actually does

This Final Edition computes a 12-feature aggregate profile from a wallet's normal ETH transaction history only (frequency, volume, UTC timing entropy, counterparty spread, dormancy, and related ledger statistics), then scores it against five hand-authored archetype rubrics using a fixed weighting — not a trained clustering model. The "Mechanics of Consciousness Mapping" (MCM) mode is a second, simpler rule layer over the same features. Both are narrative devices, not diagnostic instruments. The displayed "rubric margin" is the gap between the top two rubric scores — it is not a statistical confidence or probability.

What earlier validation actually found

Before this edition, the project tested a more ambitious ML pipeline (per-transaction GMM/HDBSCAN clustering into archetypes) against a 600-wallet synthetic benchmark. The honest results:

~19–32%
label recovery (chance ≈ 20%)
κ ≈ 0.02–0.27
temporal stability
κ ≈ 0.99*
wallet-level aggregate redesign (*synthetic only)

The per-transaction clustering did not clear the bar for even a descriptive-segmentation claim. A later redesign using wallet-level aggregate features scored far better in synthetic testing (κ≈0.99, silhouette≈0.76), but has not yet been validated against real, labeled wallet data — so it isn't shipped as a live model here. This edition uses the transparent rubric approach instead, precisely so it doesn't borrow credibility from an unvalidated backend.

Data scope & handling
  • Analysis covers normal ETH transactions only (Etherscan's txlist endpoint). Internal transactions, ERC-20/NFT transfers, and contract-level activity are not included — this is a slice of the ledger, not the whole thing.
  • "UTC tx-hour entropy" and "weekend activity" are computed on UTC timestamps. They describe patterns in the raw ledger clock, not a person's local schedule or timezone.
  • In demo mode, all transaction data is generated locally from a synthetic seed and clearly labeled as such throughout the results.
  • In live mode, requests go through your own Etherscan proxy (Etherscan V2 unified API, chain-ID based) so no API key is ever exposed in this page. The proxy locks CORS to your app's exact origin, rate-limits requests, and paginates results — it's a gatekeeper, not just a key hider. The app itself only accepts HTTPS *.workers.dev proxy URLs by default.
  • This app doesn't intentionally persist wallet data in its own database or browser storage, and reloading clears the page's own state. It can't make promises about your hosting provider's, your proxy's, Cloudflare's, or Etherscan's own operational logs — those are outside this app's control.
Naming, for the record

CHRONALCRYPT™ is the engine tier — the raw ledger-fetching and feature-extraction layer. CHRONALCRYPTEX™ ("The Brains") is the interpretation layer that turns features into archetype/MCM tags. The Vault Diary is the voice — the layer that renders those tags as first-person prose. The umbrella brand across all of it is CHRONALCRYPT™ LABS.