How this works

Every fact comes from a public source and carries its date. If we couldn’t read something, the page says so. We don’t give a score or say whether to buy.

Code

We look for a project’s GitHub in three places: links on its official site, DefiLlama, CoinGecko and CoinMarketCap. Each owner shows where we found it.

We count commits by people, not bots, over 90 days in the five most recently pushed repos, reading at most 100 per repo, which is why large projects show “+”.

No code found doesn’t mean no code. A closed team or a repo no source points to looks the same to us, so a project can show us the right one.

Usage

Fees, revenue and TVL come from DefiLlama. Fees are everything users pay, including to liquidity providers, so compare them with care across different kinds of project.

DefiLlama covers mostly DeFi and chains. Elsewhere we write “not read”.

Numbers

We read every token CoinGecko and CoinMarketCap return: 21,523 in all, 5,223 on both. The gap is the larger value divided by the smaller.

A match needs the ticker plus a price that agrees: within 50% if the name matches too, within 3% with a market cap within 1.5x if it doesn’t. Anything else stays as two rows, so a real price gap above 1.5x shows as two tokens. Matching hasn’t been reviewed by hand.

On the chain

For the top 1000 we read the token’s home contract: its supply from public RPC nodes, and owner powers and top holders from GoPlus. Supply is compared with the aggregators only for one confirmed contract on one chain; a token on several chains or a wrapper isn’t judged.

Owner powers are facts about a contract (can mint, can freeze, can edit balances), not a verdict on its team. Unlocks for the next 30 days come from DefiLlama’s vesting schedules.

Paper and real market cap

Paper cap is every token at today’s price; market cap is the circulating part. Real cap discounts the market cap when order books are thinner than usual for a token that size: we sum 2% sell depth across exchanges (CoinGecko tickers, outlier venues set aside) and 0.5% of honest DEX liquidity, and compare it with a curve fitted on the top 200 projects.

DEX liquidity counts only pools quoted in a real asset, at twice the quote side: a pool against a made-up token can claim any size. Exit days are how long selling the whole cap would take at 2% a day.

Volume gets the same test, venue by venue: a day of honest trading passes through a book a bounded number of times its 2% depth (the norm is the median on large venues). What a venue reports past three times the norm is not counted as carried by its book, and venues past ten times are named. On those we also read recent trades and set them against Binance’s (or Gate’s) for the same token: real trade values follow Benford’s law, people trade round sizes, and real markets have a few large trades. A venue failing two of the three is named as not trading like people.

Where a coin lives

The home contract, the deployments CoinGecko lists, separately listed wrappers, and every token DEXes trade under the same ticker. One no list names is unknown; one whose owner can take or freeze tokens is a dangerous copy. A chain’s own bridge counts as the token carried over.

For a token on several chains we read its supply on each, and on Ethereum what the canonical Arbitrum, Optimism, Base and Polygon bridges hold of it: a bridged copy should be fully backed by tokens locked there. Copies minted on their own chain (Binance-Peg, burn-and-mint bridges) are counted, but their backing can’t be checked.

Leverage and liquidations

Every hour we read open interest, volume and funding for perpetual futures on Binance, Bybit, OKX and Hyperliquid, and match each to a token by ticker and price. CME and other venues are not counted.

Liquidations come from Bybit’s feed (every one), Binance’s (the largest per symbol each second) and OKX’s latest orders: a lower bound. Hyperliquid publishes every position’s liquidation price, so its map of where positions would close is exact, for the largest accounts and busiest traders we read.

Signals

A signal is a change between two readings: an unlock within a week, an owner gaining a power or handing the contract on, large holders moving between wallets, exchanges, DEX pools and locks, pool liquidity falling 30% or more, a dangerous copy with real money in its pools, a mint on the home chain its vesting and emission schedule doesn’t cover, a bridged copy losing its backing, supply growing on another chain with no mint at home.

Each is sized in days of 2% sell depth and as a share of a day’s trading, and is news only past a quarter of a day’s trading or a day of depth. Holder moves must still show on the next read. Exchange, custodian and market-maker addresses come from the exchanges’ own reserve lists, Dune and eth-labels. Only changes read after signals were switched on count.

What a project can do

Prove it runs the site its token lists and point to its repo, supply page and unlocks. These appear labelled as stated by the project, next to what we found, and feed no order, shelf or signal.

It can’t set a number, and money doesn’t change a position.

What this prototype doesn’t do yet

Claim requests are accepted by hand and the deployer signature isn’t verified. Data licences haven’t been reviewed. Sources are re-read on a schedule, and each page shows when.