Who really mines the memecoins?

Merged-mined coins do not have their own miners in the usual sense. Their blocks are found by a handful of Litecoin (and, for Namecoin, Bitcoin) parent pools, because a parent block's coinbase can commit to the child chain at no extra hashing cost. Here is the measured split. Nobody else publishes it, because it requires reading the parent coinbases yourself from your own full nodes, one aux block at a time. We did, for about eight days. Across the nine live chains below, F2Pool found 64.4% of blocks and AntPool 10.4%.

Findings

• F2Pool and AntPool dominate. Over all nine live chains, F2Pool found 49,563 of 76,977 blocks (64.4%) and AntPool 8,040 (10.4%): together 74.8%.
• Who carries a chain decides who can win it. On JKC, CRC, SHIC and TRMP, AntPool and ViaBTC barely appear (AntPool: 17 of CRC's 10,411 blocks, none on JKC, SHIC or TRMP; ViaBTC: none on those four) and F2Pool alone found 84.7% to 94.3%. TRMP is the extreme: 9,999 of 10,609 blocks from one pool tag family.
• PEP, LKY, DINGO and NMC are shared by three or four large pools. F2Pool, ViaBTC and AntPool found 76.8% of PEP and 87.1% of DINGO; F2Pool plus AntPool found 77.4% of LKY; F2Pool, AntPool and ViaBTC found 76.5% of NMC (parent chain Bitcoin). Kupool is the fourth name on PEP (7.5%), LKY (10.9%) and DINGO (9.0%).
• DOGM is one operator story. Parent tags on DOGM are the most fragmented of any chain: no pool above 17.5% (okminer), then yiimp 14.5%, and 13.4% unidentified. yiimp is a solo/merged-mining frontend, not a single miner. Yet 299 of the last 300 DOGM coinbases pay one address (measured on our node, 2026-10-09). Many parent tags, one payee: the tag says who found the parent block, the coinbase says who got paid, and on DOGM those are different people. We report the measurement and do not name the payee.
• Unidentified shares, stated plainly: DOGM 13.4% (284 of 2,120); CRC 0.2% (21); JKC 0.2% (18); LKY 0.1% (10); PEP 3 blocks; DINGO, SHIC, TRMP and NMC none. The parked BELLS sample is 26.9% (902 of 3,347).
• Chain-ID quirks. DINGO accepts two AuxPoW chain IDs. DOGM rotates its chain ID yearly: 0x62 became 0x6C at height 6,148,000, found by bisecting on our own node.
• An odd tag on JKC and LKY. "Made in USA," (verbatim, with the trailing comma) found 8.0% of JKC and 3.9% of LKY, and 2.8% of PEP. We do not know who writes it; we only count it.

Method

Pipeline, in full: own full nodes (one per chain, synced from the P2P network only, see /nodes) → for each new block, the AuxPoW header → the parent block's coinbase → the printable text in the coinbase scriptSig, stored as the pool tag next to the parent block hash and chain. One row per aux block: chain, height, hash, parent chain, parent block hash, parent pool tag, parent coinbase, time first seen.

Normalization, display only. /F2Pool/a to e (and tags with trailing bytes) show as F2Pool; every Mined by AntPool variant, including +, x and padding, as AntPool; /ViaBTC/ and its sub-account forms as ViaBTC; /LP/ as Litecoinpool; okminer, as okminer; yiimp as yiimp; Secpool variants as Secpool; "Made in USA," kept verbatim as its own tag. Matching is by substring, so a tag with leading junk bytes still lands on its pool. The raw tag is preserved unchanged in the archive. Three raw rows, so you can check us:

What "unidentified" means. Either no parent coinbase was recorded for that aux block (NULL tag: 273 rows on DOGM, 21 on CRC, 902 on the parked BELLS sample), or the coinbase carries the merged-mining marker (fabe6d6d) but no readable pool text, so the extractor returned only the marker and a byte of the commitment (DOGM 11, JKC 18, LKY 10, PEP 3). It does not mean solo mining, and it does not mean a small pool. Other is every tagged block not matched to one of the named pools above (long-tail pools, odd tags, unparsable bytes).

Observations can exceed distinct heights by a few rows when a reorg re-records a height (largest gap: SHIC, 11). Shares are computed on observations. BELLS: 2026-10-02 11:16 → 2026-10-03 14:09 UTC, then the node was parked; it is shown for reference and excluded from every total above. CRC starts 2026-10-02 19:17 UTC.

Limits

• Eight days is short. Pool shares drift. DOGM (2,120 blocks) and NMC (1,085) are small samples; treat their fourth and fifth digits as noise.
• Tags are self-reported. A pool writes whatever it likes into its coinbase. One pool can use several tags (F2Pool a to e, which we merge), several pools can share a frontend (yiimp), and a tag can be copied. We count what is written.
• Carrier versus finder. A pool that carries a merged-mined coin pays its miners however it chooses. The tag names who found the parent block, which is where the aux block came from. It is not a statement about who was paid. DOGM is the visible case where the two diverge.
• Normalization is ours. The rules above are substring matches written by us; a different grouping would move a few points between named pools and other.

Not a recommendation about any coin. This is what the coinbases say.

FAQ

Who mines the merged-mined memecoins?

In the measured window, mostly the same few large pools that mine Litecoin itself. F2Pool found 64.4% of the 76,977 blocks observed across nine merged-mined chains, AntPool 10.4% and ViaBTC 7.9%. The split differs a lot by chain, see the per-chain bars on this page.

How do you know which pool found a merged-mined block?

Every merged-mined block carries an AuxPoW proof that points at a parent block, and the parent block has a coinbase transaction. Our own full nodes read that coinbase, and we record the pool tag written in its scriptSig. The tag is self-reported text, so it names the pool that found the parent block, as that pool chose to label it.

Does a pool tag mean that pool paid its miners for the coin?

Not necessarily. The tag names who found the parent block. A pool that carries a merged-mined coin decides how that coin is paid out, and this page does not measure payout. DOGM is the one chain where we also looked at the payout side, and it points at a single payee.

Is this a live feed?

No. It is a dated study of about eight days, 2026-10-02 to 2026-10-10 UTC, with every number baked into the page. A live version is coming. The raw tags are preserved in our archive, so every figure can be recomputed.