Insider Radar

Each verdict is frozen when the token is first published, then its price is followed for 7 days. Scores appear once 10 tokens of a verdict have finished.

Once a day a copy of the track record and the radar data goes into proofs/ with a list of their SHA-256 hashes, timestamped with OpenTimestamps in the Bitcoin blockchain. That shows each day's verdicts existed before the prices that followed, without trusting this site or its git history. To check a day: download its folder, run sha256sum -c manifest.txt, then ots verify manifest.txt.ots if you run a Bitcoin node. Without one, ots --no-bitcoin verify manifest.txt.ots prints a block number and Merkle root to compare with any block explorer, or drop the two manifest files on opentimestamps.org. A new stamp takes a few hours to reach a Bitcoin block.

Comparisons fixed in advance

Each High risk verdict is checked a second time, apart from the tracer: every link re-read on the chain, the token's age, and how active each linking wallet is. Reviews don't change verdicts. radar-review.json

Paste a wallet to see which traced tokens it funded. Wallets are remembered since launch.

Seen behind more than one token

For the 20 largest holders of a token, Sift finds who paid for each holder's first transaction (two hops back) and who sent it its first tokens. Holders that lead to the same private wallet form a linked group. Exchanges, pools and programs link nobody.

Verdicts

Weaker evidence

• A link through a private wallet with over 3,000 transactions but a small balance (possibly an exchange withdrawal wallet or a bot) is not enough for a "High risk" verdict.
• A wallet seen behind top holders of several tokens is flagged: it can be a shared service, or one deployer behind several tokens.

Track record rules

• A token enters the first hour it is published with a price on GeckoTerminal; its verdict is frozen then. Its price is sampled hourly for 7 days.
• Collapsed means the lowest hourly price fell to 20% of the entry price or less, or the token stopped trading (no price for its last 24 hours).
• From 15 October the ledger also records the first hour a token is 50% and 80% below its entry price. It is extra information and changes no score; tokens already below a line by then have an unknown time.
• Verdicts are not assigned at random, so read differences between them with care. An entry whose prices could not be checked for over 6 hours in a row is not scored.
• The radar covers memecoins. On 23 October stablecoins, wrapped coins, staking tokens, tokenized stocks and large protocol tokens (listed by exact address) were taken out of its scope. Their linked holders are usually the issuer's own wallets. The 8 such entries already in the track record are kept in the file, marked out of scope, and left out of the scores.
• Until 21 October, the price requests ran into the price service's rate limit, and many entries went over 6 hours without a price. Those entries stay unscored. Since then the oldest unchecked entries are asked first and refused requests are retried. The 6-hour rule itself has not changed.

Mint and freeze authority

Read from the token's mint account every hour. A wallet holding the mint authority can create new tokens; one holding the freeze authority can freeze holders' tokens. Bridged and tokenized assets keep these with a program, which is normal. "Revoked" includes authorities set to the System Program, which nobody can use.

LP burn

For each token's main pool (the one with the most liquidity), the share of LP tokens burned: the pool's record of LP ever minted against the LP mint's current supply, read every hour. This works for Raydium AMM and PumpSwap pools. Concentrated pools have no LP token, and other pool types are marked as not checked. Lockers are not read, so LP that is locked rather than burned counts as not burned.

Still held

For High and Watch tokens, the 20 largest accounts are read again every hour and matched to the linked wallets. A wallet that has dropped out of the top 20 isn't counted, so the figure is a minimum.

Token age

The date of the oldest trading pool found for the token. On a token that has traded for more than 30 days, a linked group is often the project's own wallets (treasury, vesting, market making) rather than insiders from the launch. Those tokens keep their verdict, because the grading rules don't change after the fact, but the page points it out.

Self-audit

Each hour a sample of published links is re-checked on a separate Solana node with separate code. A link the node could not answer for is never counted as passed. You can run the same check yourself on any token: download verify.py (Python 3.11+, no key or packages) and run python3 verify.py <mint>. It checks every published link of that token against the public Solana node, or your own with SOLANA_RPC_URL.

Limits

• Covers the tokens Sift's scanner watches, requested tokens, and up to four trending tokens an hour.
• Only the 20 largest accounts are sampled, and wallets with over 1,000 transactions are not traced back. A clean result is not a safety guarantee.
• Scanner traces are shown for 24 hours (2 if incomplete), requested ones for 7 days.
• A linked group suggests one actor holds several wallets. It says nothing about what they will do. Not financial advice.
• Sift publishes on-chain relationships anyone can check. It does not identify people, prove who owns a wallet, or predict prices.

Free to use, updated hourly: radar.json, radar-ledger.json, radar-wallets.json, radar-audit.json. Alerts for new High and Watch verdicts by RSS.