$GPU

The index has not attributed an ETH-quoted pool to this token, so there is no pool to read a tape from. That is a statement about our index, not about whether this token trades.

• · V4 depth is UNMEASURED, not zero: V4's PoolManager is a singleton holding every pool's tokens in one contract balance, so per-pool reserves cannot be read as balances. We say so rather than guess.
• · Volume is the 24h ETH leg the chainwide index attributed to each pool — a floor over what its per-tick budget could walk, never a lifetime total. Share divides by $GPU's 24h ETH across all 1 of its pools, not just the rows listed.

last 24h 0 ETH added over 0 events · 0 ETH left the pools over 0 movements (V3 collects and V2 burns — the two ways tokens actually leave a pool contract). Floors over the pools and blocks we walked, not a census.

No liquidity event has been indexed for $GPU inside the blocks we walked. Read the coverage line above before treating that as calm: it is the absence of an observation over a bounded window, not evidence that nothing moved.

• Every row is a log we decoded, stamped with the block it came from. This panel states what moved, when, and who signed for it — it does not label a removal, and it never will. Taking liquidity out is an ordinary operation; the reader draws the conclusion from the numbers and the times.
• V3 splits a removal in two: closed is the burn, which only credits what the position is owed, and withdrawn is the collect, where the tokens actually leave the pool. A collect carries accrued fees as well as principal and the event does not separate them, so we do not either. V2 does not split: a V2 removed transferred in the same call, with no collect to wait for, which is why the summary counts it as ETH that left rather than as principal credited.
• Percentages are deliberately absent. “80% of the liquidity left” needs the depth before the removal, and we hold depth only as a live read at one block — no history to divide by. The absolute amounts above are what we actually measured.

• · Every row is the account that SIGNED the swap transaction — never the router the trade was routed through. (Measured while building this: every swap-event “sender” in a live window was a contract; a board of those would be a list of routers.)
• · Totals cover ONLY the indexed window: 1 swaps attributed for $GPU between 2026-08-05 22:33Z and 2026-08-05 22:33Z — at least 12% of the 8 swaps our volume index counted over the same hours. Attribution is budgeted per tick and takes the highest-value blocks first, so every figure here is a floor over a stated period, never a lifetime total.
• · ETH in and ETH out are shown separately because net alone cannot tell “never sold” from “round-tripped”.
• · An account signing on others' behalf (a keeper, a relayer) appears as one signer. The “tokens” column is shown so bot- and infrastructure-shaped accounts are visible rather than silently ranked as people.

• · These strings were written into the token contract by whoever deployed it, and some of what is above came from our index rather than from this request — each part says which. Anyone can deploy a contract and write anything into it. A link here is not evidence that the account, site or person it points to has anything to do with $GPU — we do not check, we do not ask them, and we never follow the link to see where it lands.
• · Read them yourself: the contract on the explorer ↗

No holder concentration is shown for $GPU: no Transfer events appear in the window we walked, and the token emitted no logs at all in the blocks we probed to check — so it looks quiet rather than clipped. No wallet could be nominated for a balance read.

· The read itself worked — this is what it found, not an outage. Candidate discovery covers the last 120,000 blocks (about 3.3 hours) of chain history, so a token whose transfers all predate that window is invisible to it.