this is the recovered map drawing itself back in, tile by tile, the way it actually came back. every shaded county is a real rate from real kids who got tested. the grey is just honest: those states are not in the recovered rollup, so they sit blank instead of pretending to be zero. Source: Reuters, "Off the Charts" / "Unsafe at Any Level," M.B. Pell and Joshua Schneyer, 2016-2017. The underlying childhood blood-lead data was obtained by Reuters from the U.S. Centers for Disease Control and Prevention and state health departments via public-records requests.
somebody spent years finding out where american kids were being lead-poisoned, neighborhood by neighborhood, and published the map. then it quietly went dark. nobody fought to take it down. a website moved to a new system, the article got unlisted, and the data just kept sitting there, running, with no door left to it. the whole thing came back in about seventy seconds.
back in december 2016, two reuters reporters, M.B. Pell and Joshua Schneyer, published a pair of stories called "off the charts" and "unsafe at any level." the heart of it was a map you could zoom into: the share of tested children under six who came back with elevated blood lead, drawn not by state, not even by city, but all the way down to the census tract and the ZIP. close enough to point at a street.
the part that stuck with people, the line everyone repeated afterward, was that thousands of american neighborhoods had childhood lead rates at or above the worst of Flint. and they did not mean that as a figure of speech. they meant a real percentage, in a real tract, with a real count of tested kids sitting underneath it.
that is not the kind of thing that should ever have to be found a second time. it is government testing data. it came off real children. and the thing it measures does not heal. lead does not wash back out of a kid. whatever it took, it keeps.
this was never really a tech story. it is a map of where children already lost something nobody gets back.
the article and its map got retired in a site migration. the old special-report and graphics-era pages stopped being listed. you can still reach the shell of the story through the wayback machine, but when you open the archived page, the map loads empty. just the words, sitting over a blank space where the country used to be.
so it looked deleted. it was not deleted.
the data was never taken down. the stack of map tiles that fed the whole thing is still sitting on reuters' own servers right now, no link pointing at it, still answering when you ask, last touched september 2025:
https://www.reuters.com/investigates/graphics/lead-water/v-3/lead-tiles-v3/{z}/{x}/{y}.pbf
nobody pulled the data. somebody just cut the path to it and walked away. the evidence kept running in a room with the door bricked over, lights still on, nobody coming.
the map is not one file. it is a tile map. the data loads as thousands of little protobuf squares, and the browser only goes and fetches a square when a person pans or zooms over it. nothing loads until somebody looks.
archive crawlers do not pan maps. they do not grab a tile, drag west, and grab the next one. so during archiving, the squares were never asked for, which means they were never saved. the article text is in there. the methodology is in there. the styling is in there. the one layer that actually mattered, the kids, the rates, the places, is the one part that never made it.
that is the quiet trick of how this happens. nobody has to delete anything. you just take away the link, and let a machine keep failing to ask for the part only a human ever triggered. given enough time it reads, to everyone, as gone.
it turned out to be less of a fight than it should have been, which is sort of the whole point.
it is either the whole dataset, checked against itself, or it does not go up. half a map of poisoned kids is worse than none.
| what | how much |
|---|---|
| total areas recovered | 30,187 |
| census tracts (all valid 11-digit GEOID) | 18,127 |
| ZIP areas (ZCTA) | 12,060 |
| areas with real testing data | 28,564 |
| areas where testing was withheld | 1,623 |
| areas at 2× Flint's ~peak or higher | 8,280 |
| states present | 29 |
| time it took to pull all of it | ~70s |
"2× Flint's peak" means at least 6% of tested kids under six came back elevated, roughly double the worst tract-level rate measured during the Flint crisis. eighty-two hundred neighborhoods cleared that bar.
the five worst tracts that came back
share of tested children under six with elevated blood lead. the count of kids tested is shown under each bar, because the number matters as much as the rate.
still running, just unlinked
this is the tile stack that fed the map, still alive. every square is a piece of data a crawler never asked for, because no person ever panned to it. press play to walk down the zoom levels the same way the recovery did.
z0 . the whole country in one square
Reuters, "Off the Charts" / "Unsafe at Any Level," M.B. Pell and Joshua Schneyer, Reuters, 2016-2017. the underlying childhood blood-lead data was obtained by Reuters from the U.S. Centers for Disease Control and Prevention and state health departments via public-records requests.
this rebuild reproduces reuters' own published reporting from reuters' own data that is still live. it does not re-report it and it does not re-attribute it. the credit stays on its face, here and anywhere this map ever shows up. they did the work. it should keep their name on it.
this was a read of an open, public endpoint. no paywall, no token, no login, nothing asking who you were. it was read slowly and politely, and the attribution was kept attached the whole way through. no access control got bypassed because there is no access control on it to bypass.
same as the baby-food heavy-metals recovery: nothing taken that was not already public, nothing stripped of where it came from, nothing dressed up as ours. it is reuters' reporting about government records about kids who got hurt. it just should not be one website migration away from gone.
the internet never forgets.
even if it got lead poisoning.
pulled back 2026-05-18. reproduced from Reuters' own still-live source, Reuters' credit kept on its face, and rebuilt here so the map of where it happened to children cannot quietly go missing again.
