Post
literally these are books that would be pulped if not bought for 10 cents on the dollar.
do rare books sometimes make it into there? yeah, if someone is careless, but that's *upstream* of what happens on sale, not *downstream*. you fundamentally are rescuing something from the fucking pulper.
I literally worked on Google Books scanops SRE, ama
the same guy who ran Google Books now has the equivalent job at Anthropic and I can promise you that all of us handled every library book with the utmost care, returning them with not even a single ripped page or ding. but destructive scanops...
Damn that's a pretty cool line to have on your resume. May I ask what the SRE component of scanops entailed please? Were you monitoring/operating the services that were ingesting, processing and storing the scans? The physical scanning machines too?
Correct. If the book scanning machines or the miniature Google datacentre located right next door to the scanning floor broke down, then people were out of work for a day, they were hourly contractors and we cared about making sure they could smoothly get their jobs done.
Strangely enough, we were genuinely one of the few SRE teams at Google who owned the whole stack from top to bottom, because we had to run our own datacentre adjacent to each scanning centre, and all of the Google infrastructure for compute, data processing, etc.
The software HAD to work and give operators real time feedback on "hey that page didn't capture clearly" or "your finger cot is visible in the captured image", and they had to then turn back the correct number of pages before it was too late and the book was already closed and put away.
Literally doing real time data processing and image recognition. We obviously weren't transcribing it to text on the spot, but QC had to be an immediate feedback loop.
Fascinating, I remember reading a story about the handful cases where fingers slipped through into the scans in some cases. I imagine never fun to have your work on blast in the media although I recall the coverage was mostly in good humor
Every scan *of a library book* was done non-destructively with a human being flipping through page by page, wearing finger cots to protect the pages from damage. Occasionally they snuck through QA but yeah they're a reminder of the care we took to preserve the books.
Oh neat, I thought they used human operators because the auto page turning machine ended up being slower and jankier than something simpler with human operators for page turning, I might have hallucinated this though. Did the auto-page-turning machine that they demoed really get used at scale too?
For some books, yeah, the auto turning would work, but robotics back then was not what it is today. A majority of the effort by time certainly was human
Makes sense, either way, pretty rad what you all were able to accomplish. Thanks for sharing!
10:41 AM · Jul 30, 2026
Genuinely one of the things I'm proudest to have worked on in my career!