Pressfront / Trace / Where did the GameStop mention spike start?
Traced story · Updated September 2026
Not every spike is a rumour. Some are an event, and the useful question is which one you are looking at. We traced a week of GameStop chatter and the tracer landed on the trigger: an earnings announcement, 56 hours ahead of the amplifier that carried it.
The earliest relevant public post we could verify is @ThePPseedsShow on X, 8 September 2026, 15:23 UTC measured, announcing blowout earnings. The biggest amplifier posted 56 hours later derived: @unusual_whales, reach 5.36M measured. The path ran x then hackernews then reddit derived.
Confidence is 59% derived. Two things hold it there: only one platform saw the early window, and 56 of 105 in-window mentions were dropped as off-topic (the 74 posts dated outside the window are set aside and not held against the pull), because a one-word company name pulls in a great deal of unrelated text.
The screen, as the tracer printed it
GameStop, window 2026-09-08..2026-09-15https://x.com/ThePPseedsShow/status/2097344978867429583Verbatim from the tracer: “earliest we could VERIFY, not proven origin. deletions & screenshot-laundering are invisible to us.”
The private-account rule. We never log in, follow anyone, request access to a locked account, or use scraped private data. A private account is invisible to us and stays that way. So what you see below is the earliest public appearance we could verify, never a proven origin. If a story started in a group chat, a DM or a locked profile, the first public post is the node we can see, and we say so rather than dressing it up as the source.
Run one yourself: Trace a story. How the sweep and the confidence number are built: the methodology card.
This is the case people underestimate. A mention spike with a real corporate event underneath it and a mention spike with an anonymous claim underneath it produce the same shape on any dashboard that only counts posts. The response you would want is completely different.
The only way to tell them apart is to look at the earliest node and read what it says. Here the earliest relevant node is an earnings announcement. There is no rumour to get ahead of. The correct action is a communications decision, not a crisis one.
59 out of 100 is a modest number for a result that is, as far as we can tell, correct, and it is worth reading the number rather than the verdict.
Two penalties fired. Only one platform saw the early window, so nothing corroborates the earliest node. And relevance health was 0.47, because a one-word company name drags in unrelated text at volume (56 of 105 in-window mentions were dropped as off-topic (the 74 posts dated outside the window are set aside and not held against the pull)). Both are true statements about the evidence, not about the answer. A trace that printed 90% here would be claiming a quality of evidence it does not have.
On a fourth story in the same batch, a bank layoff rumour, the earliest node we held was a post that cited Reddit as its own source: “Employees on Reddit are reporting another round of restructuring and layoffs”. We did not hold that Reddit post, and the paid Reddit collector we had at the time could not search by date.
The tracer flagged it as a relay, named Reddit as the likely true origin platform, counted a citation hole, and cut confidence to 42%. It did not invent a source. That is the behaviour we care about most, because the expensive failure in this category is not a miss, it is a confident wrong answer with a real person’s name attached to it.
On the week we traced, the earliest relevant public post we could verify is an earnings announcement posted by @ThePPseedsShow on X at 8 September 2026, 15:23 UTC, and the biggest amplifier we measured, reach 5.36M, followed 56.4 hours later. So the spike has an event under it rather than a rumour. Confidence in that reading is 59%.
It can show you which one you are looking at, by putting the earliest node and its text on the screen next to the volume curve. An earnings release and an anonymous claim make the same shape on a mention chart and require completely different responses. Reading the earliest node is how you tell them apart, and that is the whole reason the screen leads with it instead of with a score.
Because a short, common query pulls in text that has nothing to do with the story. On this trace 56 of 105 in-window mentions were dropped as off-topic (the 74 posts dated outside the window are set aside and not held against the pull). Dropping them is correct, and the share dropped is reported rather than hidden, because a corpus that was mostly noise is a weaker basis for an origin claim than a clean one, and the confidence number carries that.
No, it is usually an accurate one. Confidence measures the quality of the evidence, not our enthusiasm for the answer. On the bank-layoff case in the same batch, confidence came back at 42% because the earliest post we held cited Reddit and we did not have the Reddit post. That number is the tracer telling you precisely where to be sceptical, which is more useful than a round number with nothing behind it.
Free trace, no account for the first one. You get the same screen you just read.
Honest fine print. Trace reports the earliest public appearance we could verify inside the window we searched, which is not the same thing as a proven origin. We only read public posts and public metrics, so deleted posts, screenshots laundered out of private accounts, and anything that started in a DM or a locked group are invisible to us, and every result names the sources that came back empty. Confidence is a number we publish together with the gaps behind it, never on its own. Trace describes what has already happened and how fast it is moving right now; it does not predict what will go viral, and no honest tool does.