$GLASSCA drops here at launchpump.fun
Method 1.2 · generated from the scanner's own constants

The method.

Everything Glass decides, and the exact number it decides with. If a number on this page changes, the method version changes with it and old maps are never mixed with new ones.

Link rules
8
Cohort signals
6
Score inputs
6
Labels typed by hand
0
01

The scan

Five reads, in order. Each step streams to the page while it runs, so you watch the map being built instead of waiting on a spinner.

  1. 01TokengetAsset

    Name, symbol, decimals, supply.

  2. 02HoldersgetTokenAccounts

    Every token account, summed per owner. Up to 12,000 accounts; the top 100 are mapped.

  3. 03Fundingfunded-by

    The first SOL each wallet ever received. Cached forever: it cannot change.

  4. 04First buytoken account history

    The oldest transactions of each holder's own token account: slot, payer, tip, size, source. Then the SOL that paid for it.

  5. 05Transferswallet transfers + mint history

    Every holder's last 100 transfers, and everyone who got the token in its first slots. Then the rules in src/lib/graph.ts.

Cost of one fresh scan22,240 Helius credits
  • Funding sources100 × funded-by at 100 credits10,000
  • First buys100 token accounts × getTransactionsForAddress at 101,000
  • Reloads100 × the last SOL in before the buy at 101,000
  • Transfers100 × last 100 transfers at 10010,000
  • Names1 × batch-identity per 100 wallets100
  • Token and holdersgetAsset, deployer and up to 12 holder pages at 10140

A rescan of a known token costs a fraction of this: funding and first buys are stored permanently, only new holders are read.

03

Cohorts

Bundlers spread SOL through fresh wallets so no transfer ever joins them. What they cannot hide is timing and habit. A cohort is 3 or more wallets whose first buys land within 2 slots of each other, or 2 or more that bought in the block the token was created in: those buys were placed together with the launch.

First buys by slot · example launch
2-slot window312,004,108312,004,113312,004,118312,004,123
  1. 7 first buysin slots 312,004,117 and 118
  2. One fee payer9Wzk…k3Lq paid for 6 of the 7
  3. Same size1.98 to 2.04 SOL, a 3% spread
  4. Same Jito tip0.001 SOL on every buy
  5. Cohort, high confidenceno transfer joins them, the timing does
SignalCounts when
One fee payerOne wallet paid the fee for 2+ of the buys. On its own this makes a cohort high confidence.
Same sizeSOL spent spreads less than 12% of the median.
Same Jito tipHalf or more paid an identical tip to a Jito tip account.
Fresh walletsHalf or more were funded less than 6h before their first buy.
ReloadedHalf or more received SOL less than 6h before their first buy, however old the wallet.
Launch blockThe buys landed within 2 slots of the token's creation. Counts as two signals.
high

A shared fee payer, or two signals.

medium

One signal.

low

Timing alone. Shown on the map, never scored.

Launch bundle

Separately, the token's own history is read from its creation slot: every wallet that received it in that block or the 2 after it, except the deployer. Those buys were placed with the launch. The panel shows how much of the supply they got and how much they still hold, and what they still hold is scored. A bundle that sold or moved its tokens still shows as bought.

04

Fresh wallets

A wallet is fresh when its first SOL arrived less than 6 hours before its first buy of the token. Real traders have history; a bundle's wallets are usually made for the launch.

Fresh
funded 2h 10m before buying
Not fresh
funded 41 days before buying
first SOL in first buy 6h window
05

Score

The score starts at 100 and loses points for each signal, measured as a share of total supply. The caps add up to 100. Move the inputs to see the exact arithmetic the scanner runs.

Score
45/100
Concentrated

penalty = min(max, max(0, share − free) × weight). Free share: top 10 wallets 20%. Everything else counts from the first wallet.

InputFreeWeightMax
Top 10 wallets20%×0.510
Linked wallets0%×130
Bought together0%×125
Still held by the launch bundle0%×1.515
Deployer and its links0%×215
Fresh wallets0%×0.255
Heavy0+
Concentrated40+
Watch60+
Clear80+

The score describes a holder list. It does not say a token is good or bad, and it does not accuse anyone.

06

Labels

Known entity

Names from Helius' identity data: exchanges, protocols, market makers. Exchange wallets never join a cluster.

Rule

Deployer, Burned, Program account. Derived from the chain, never typed by hand.

Project

Shown with who set it. A label changes the name on a bubble, never the lines around it.

07

Limits

What Glass does not see yet, stated plainly so nobody reads more into a map than it holds.

  • Only the top 100 holders are drawn and scored.
  • Funding is the first SOL a wallet received. A wallet funded through an exchange and then used by a bundler does not link by funding; it can still land in a cohort.
  • Two-hop funding (A funds B, B funds C) is not followed yet. Reloads through an exchange link only as a batch, so a bundler who withdraws at random times and amounts can still pass.
  • A cohort is evidence of coordinated timing, not proof of one owner. Bots chasing the same launch can look alike.
  • Finished maps are cached for 90s. Rescan reads everything again.
08

Source

Glass is open source under the MIT license. The rules on this page live in src/lib/graph.ts, the thresholds in src/config/site.ts, every chain read in src/lib/helius.ts, and the tests that pin each rule in test/graph.test.ts.