Over 100k years ago according to serpent mound in Ohio.
Wikipedia first line:
> The Serpent Mound is believed to have been built by the Native American Adena peoples around 320 BCE, and then either added to or repaired by the Fort Ancient peoples around 1100 CE.
So maybe 2k years?
(See correction below.)
Much older.
T̶h̶e̶r̶e̶ ̶a̶r̶e̶ ̶v̶a̶r̶i̶o̶u̶s̶ ̶d̶i̶s̶p̶u̶t̶e̶d̶ ̶s̶i̶t̶e̶s̶,̶ ̶b̶u̶t̶ ̶t̶h̶e̶ ̶r̶e̶c̶e̶n̶t̶ ̶d̶i̶s̶c̶o̶v̶e̶r̶y̶ ̶o̶f̶ ̶f̶o̶o̶t̶p̶r̶i̶n̶t̶s̶ ̶i̶n̶ ̶W̶h̶i̶t̶e̶ ̶S̶a̶n̶d̶s̶ ̶e̶s̶t̶a̶b̶l̶i̶s̶h̶e̶s̶ ̶a̶ ̶l̶o̶w̶e̶r̶ ̶b̶o̶u̶n̶d̶ ̶o̶f̶ ̶b̶e̶t̶w̶e̶e̶n̶ ̶2̶1̶k̶ ̶a̶n̶d̶ ̶2̶3̶k̶.̶ (This is less settled than I'd thought.)
https://en.wikipedia.org/wiki/White_Sands_fossil_footprints
T̶h̶e̶ ̶p̶r̶e̶v̶i̶o̶u̶s̶ ̶u̶n̶c̶o̶n̶t̶r̶o̶v̶e̶r̶s̶i̶a̶l̶ ̶l̶o̶w̶e̶r̶ ̶b̶o̶u̶n̶d̶ ̶w̶a̶s̶ ̶b̶e̶t̶w̶e̶e̶n̶ ̶1̶2̶.̶8̶k̶ ̶a̶n̶d̶ ̶1̶3̶.̶5̶k̶.̶ (This hasn't been a lower bound for a while, see below.)
https://en.wikipedia.org/wiki/Clovis_point
It's my understanding this is an active area of debate. To give a tentative upper bound, the Cerutti mastodon kill - who's anthropogenic origins, while they seem compelling to me (a nonexpert), are hotly debated, have a date of 130k years ago.
https://en.wikipedia.org/wiki/Cerutti_Mastodon_site
(I suspect GP meant to reference the Cerutti Mastodon site, rather than Serpent Mound?)
(And of course, much of this is referenced in the article, I'm busted for reading the comments first! Interesting article that says it far better than I, worth reading.)
The current uncontroversial bound is roughly 15kya--as the article notes, the White Sands fossil footprints are still controversial. The Clovis First hypothesis has been considered thoroughly discredited in the academic community for about 20 years now.
The Cerutti Mastodon site is not compelling to me, because a date of 130kya would put it before any other site of homo sapiens outside of Africa. Being so far out of line with other evidence means that the evidence in favor of hominids needs to be exceptionally rock solid, which it isn't.
I would submit we should be cautious with metaobservations like, "this is much older than other sites," because we should be trying to prove ourselves wrong, whereas these sorts of metaobservations amplify confirmation bias. I'm not saying skepticism is unwarranted, just that we need to be careful with this sort of reasoning.
As a side note, I've seen people strike through stuff out on HN, and I'd like to strike through this claim to emphasize there's a correction; it's not in the undocumented features repo[1], and I've tried a few standard things like ~this~ and ~~this~~ and -this- and --this-- and have never figured it out. Can anyone tell me what the markdown is?