Methods and sources

1

What the bulletin counts

Every pump.fun coin born after $Kimchi whose name, ticker or description says kimchi, read straight from the pump program, priced every minute, placed in a family tree by an open rule, and summed into the indicators of one economy.

2

Who is family

Match

name, ticker or description contains "kimchi" (any case, inside words too) or the Hangul gimchi (Hangul)

Window

creates since the root (Aura Final Boss $Kimchi, CSocnzcRpgSCr1naKcNTieWZUxb6rABe45j844m7pump)

Index floor

indexed at market cap >= $4000; leaves only after 2 days below $3600; the root never leaves

Why a floor

about 4,000 coins a day say kimchi; most are $3k dust with one or two holders. The floor needs about $1k of real net buying.

3

The rule of kinship

Each coin gets exactly one parent. The rules are tried first hit wins, in this order.

Names are normalised (NFKD, lowercase, possessive 's removed, non-alphanumerics to spaces). A coin can only point at an indexed coin created strictly earlier (slot, then time). The root (Aura Final Boss $Kimchi) is pinned by its mint; coins created before it are older namesakes, never part of the tree. The July 2026 fan coin is shown as a predecessor outside the tree. A clone whose original never reached the index points at the earliest indexed copy and says so. Same creator wallet on both ends is shown as a "same hand" badge and never changes the parent. There are no manual overrides.

The printout loads with the live tree.

4

Indicators and formulas

gdp

sum of market caps of indexed family coins (Jupiter mcap; Dexscreener when Jupiter is down). A coin with no fresh price for 6 h is left out of the sum (it still counts as indexed); when more than half of the coins have no fresh price, no new GDP point is written.

gdp backed

sum over indexed coins of min(market cap, BACKED_LIQ_X x pool liquidity). A thin clone with a huge cap cannot pass as value.

launches h

every pump.fun create in the last 60 minutes whose name, ticker or description says kimchi. Mostly spam waves: shown as raw activity, not growth.

new indexed h

coins that crossed the index floor for the first time in the last 60 minutes (the headline number).

seats

sum of holder counts over indexed coins. One wallet holding five family coins is five seats; unique people are not measurable.

turnover

24 h buy + sell volume of indexed coins divided by GDP.

mortality

share of indexed coins with less than DEAD_VOL_USD of 24 h volume.

top5 share

share of GDP held by the 5 largest coins.

gini

Gini of market caps over indexed coins (0 = all equal, 1 = one coin has everything).

root share

the root's market cap divided by GDP.

economy n

indexed coins with >= ECON_MIN_HOLDERS holders and >= ECON_MIN_TRADERS traders in 24 h; all coins of one creator wallet count once.

growth 24h

GDP now vs our own snapshot 24 h ago; not shown before 24 h of snapshots exist (a back-cast from per-coin price changes would mix coin ages).

5

Beta to the runner

Returns are 5-minute log-returns from pump.fun candles (close prices). A derivative return counts only in a 5-minute bucket where the derivative traded, measured from its previous traded bucket (at most 1 h back), and $Kimchi's return is taken over exactly the same span; empty buckets are never filled. beta = cov(derivative, Kimchi) / var(Kimchi); corr = Pearson correlation. A number is shown only with >= BETA_MIN_N_24H joint returns in 24 h (>= BETA_MIN_N_1H of 12 in 1 h); a correlation inside the 5 % significance bound (|r| < t / sqrt(n - 2 + t^2)) is labelled "no evidence". It describes the last window only and is not a prediction.

Minimum in 24 h

30 joint returns

Minimum in 1 h

8 of 12

Coins

economy tier only

6

Thresholds of no estimate

cap to liq

Every coin prints market cap / pool liquidity; a big number means a thin pool can move the cap a lot.

stale

A value older than 3 x its refresh period is marked stale with its age.

An indicator without data is not printed. A figure without data shows its empty frame, never invented numbers.

7

The register and the token

Holders

see every new kimchi launch the moment it is created and every newly indexed coin at once

Everyone else

newly indexed coins after 10 minutes; launches below the floor are not shown

Holder threshold

$150 of the token

How the threshold moves

USD threshold at the 6 h median Jupiter price, re-priced hourly, moves only beyond a 15 % band

Delay for everyone else

10 minutes

Holding is checked by signing a login message in the wallet. It is a message, not a transaction: nothing is sent and nothing is spent.

8

Sources

• Helius WebSocket logsSubscribe on the pump mint-authority account (every create, ~1 s)
• Helius getTransactionsForAddress (backfill and gap repair of creates)
• Helius DAS getAssetBatch (descriptions)
• Jupiter lite tokens v2 (cap, price, liquidity, holders, volume, traders; every 60 s for indexed coins)
• Dexscreener (fallback for cap/price/volume when Jupiter is down)
• pump.fun swap-api 5-minute candles (beta)
• Jupiter price v3 (holder threshold)

9

Refresh

creates

live, ~1 s; gap repair every 120 s

indexed stats

1 min

candidates stats

10 min

econ

5 min

beta

15 min

tree

rebuilt within 60 s of a listing change

hold price

60 min

10

What the bulletin does not claim

• No advice: beta is co-movement in the last window, not a forecast.
• No unique people: holder-seats count wallets per coin, a wallet in five coins is five seats.
• No hand-made tree: every parent comes from the open rule; there are no manual overrides.

11

Parameters