Every label, the chains covered, where the data comes from — and, at the end, the three things this tracker cannot tell you.
Two sources: Telegram call channels and X. Both are read through GMGN's trackers rather than from Telegram and X directly, which is worth knowing because it sets a floor on how fast anything here can possibly be. See the limitations below.
Posts are polled every few seconds and stored permanently. GMGN's own feed is a rolling window — it forgets. The archive here does not, which is the whole reason you can paste a contract address from three days ago and still get its order back.
Every chain the trackers carry, not Solana only. An address is stored against a chain, never on its own — the same 0x… string exists on a dozen EVM chains and they are different tokens.
When a post carries an 0x… address and nothing indicates which chain it belongs to, it is filed as EVM? rather than assigned to the most likely one. A wrong chain is indistinguishable from a right one once stored, so it is left unresolved instead.
| Label | Meaning |
|---|---|
| FIRST | This post is the earliest recorded mention of that address. One per token. |
| FWD @name | A Telegram forward or an X repost, and who it came from. Counts as a mention, never as a first caller. |
| IN POST | The address was in the author's own text. |
| FROM QUOTE | The address was in a post they quoted, not in their own words. |
| FROM REPLY | The address was in the post they replied to. |
| FROM LINK | The address came from a link — a pump.fun or explorer URL — rather than from prose. |
| EVM? | An 0x address with no evidence of which EVM chain it belongs to. Stored unresolved rather than guessed. |
| Spread over | Time between the first mention and the most recent one. |
| Only one so far | Exactly one mention. There is no second post to measure a gap against. |
| Not seen yet | Every mention on record is a forward, so the original call has never been observed. |
| ×3 | Mention count for that token, in the left rail. |
The vertical axis is time since the first caller. The account at the top posted at 0s; the number on the right of every other row is how far behind it was. Distance down the axis is elapsed time, so a tight cluster means the token moved in seconds and a long drop means it sat there untouched.
The ruler is logarithmic, not linear — ticks at 0s, 10s, 30s, 1m, 5m, 15m. Most calls bunch into the first seconds and then trail for hours. On a linear axis the first six names would collapse onto one line and the gaps that matter most would be the ones you could not see.
The small bracket to the left of the axis. When two rows would land on top of each other and become unreadable, the lower one is nudged down. That bracket marks where the row really belongs: its top edge is the true position on the ruler, its bottom edge is where the row got drawn. Rows that carry a bracket are closer together in time than they look — the delta printed on each row is always exact.
A token with a single mention shows that row without a ruler at all. There is nothing to measure a gap against, and drawing an axis for one point would imply a comparison that does not exist.
A watchlist belongs to a wallet address, not to a browser or an account you register. Connect a wallet, sign a message to prove you hold it, and the list follows that address on any device.
The signature is checked on the server before anything is stored, and the address is taken from that verified session — never from what the browser claims. Connecting a wallet is not the same as signing in, and only the second one grants anything.
Signing costs no gas and authorises no transaction. It proves you control the address and nothing else. There is no read access to your balances or your holdings.
Three limits worth knowing before you draw a conclusion from a number on this screen.
Posts are read from GMGN's trackers, not from Telegram and X directly. That means our timestamps can never be earlier than theirs, and polling adds a few more seconds on top. Order between accounts stays accurate — everyone is delayed by the same pipeline — but the absolute clock is not the moment the post existed. If a few seconds decide your trade, this is not the tool that wins it. What it gives you is the ordering afterwards.
A post that says $ABC without an address is not recorded as a mention of that contract, because the author never posted one. The consequence is that people genuinely talking about a token do not appear in its cascade at all, and both Mentions and Combined reach read lower than reality. This is a deliberate trade: the alternative lets someone who typed a ticker outrank the account that published the address, which would corrupt the one number this whole product exists to report.
If every mention we have seen for a token is a forward or a repost, there is no observed original call — so there is no zero point, and the time axis has nothing to measure from. Those tokens show their mentions in order, without a ruler and without gaps, and say Not seen yet where the first caller would be. It is a missing answer shown as missing, rather than the earliest forward promoted into a number that looks certain.
The mechanics behind all of this are on how it works. Kestrel reports what it observed and says so when it observed nothing.