Industries / Government, education and community / Government / Libraries
Lending
Getting items to people and back: checkout and return, fines, loans between libraries, outreach vans, self-service machines and digital lending.
7
Milestones achieved, from 1905 to 2024.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1905
Books go out by wagon
The Washington County Free Library in Hagerstown, Maryland, stopped waiting for readers to come to town and sent a horse-drawn wagon of books to farms and villages. The wagon carried 2,500 books, and within five years the library had set up 66 "book stations" in stores and post offices. ALA calls it "the first bookmobile".
Credit
Mary Lemist Titcomb, the librarian; Joshua Thomas, the library janitor, who drove it
1946
Authors paid for library loans
Denmark set up the first public lending right (PLR), by WIPO's account: the state pays authors because libraries lend their books free. Norway followed in 1947, Sweden in 1954 and the UK in 1979, and WIPO counted 33 countries with PLR in 2018. Some pay by copies held, as Denmark does, and some per loan, as the UK does; the UK scheme first counted e-book and e-audiobook loans in 2018–19.
LessonLending can stay free to borrowers while authors are still paid, if a state fund covers the cost.
Credit
Denmark's 1946 law; the UK's 1979 Act, run by the British Library
1979
Interlibrary loan goes online
Interlibrary loan is one library borrowing a book from another for its own reader; ALA passed the first national code for it in 1917. On 1 April 1979 OCLC switched on an online system in which staff ordered from a terminal and an unfilled request passed automatically to the next of up to five libraries holding the item. OCLC processed 565,680 such transactions in 1979 and nearly 7 million in its 2018 financial year.
LessonRequesting got faster, but finding, copying and posting the item stayed manual, and OCLC's cost per transaction rose from 92 cents in 1979 to $1.20 in 1981.
Credit
ALA (1917 code); OCLC
1998
Radio tags speed up checkout
Singapore's Bukit Batok Community Library opened in 1998 with RFID tags in its books. RFID (radio-frequency identification) uses small radio chips that a reader detects without a direct line of sight, so borrowers could check out many books at once by themselves. City University of Hong Kong authors, writing independently, confirm that Singapore's National Library Board brought RFID into its libraries in 1998; the "world's first" claim comes from Singapore's government.
LessonEarly choices set the default: high-frequency tags were the only type available in 1998, and they became the standard for library systems that came after.
Credit
National Library Board, Singapore
2003
Libraries lend e-books online
Cleveland Public Library set a public launch for March 2003 of a service that let cardholders download e-books to read offline, built and run by OverDrive, and other large systems joined within the year. OverDrive calls Cleveland the first public library to launch a digital lending library. By its own count, libraries on its platform made 820.5 million digital loans worldwide in 2025.
Credit
Cleveland Public Library; OverDrive
2019
Fines end, borrowing rises
Chicago Public Library ended overdue fines on 1 October 2019 and wiped out old balances, giving up $800,000 to $900,000 a year. By its own count, long-overdue returns rose from about 900 to about 1,650 books a month. An independent study of Illinois libraries from 2007 to 2019, comparing libraries that dropped fines with similar ones that had not yet done so, found visits up 20% to 27% and circulation up 25% to 40% within three years, with no measurable harm to budgets.
LessonThe first headline figure, a "240%" rise in returns in November 2019, covered only a few weeks; the library's five-month figures work out to about 83%.
Credit
Chicago Public Library (Commissioner Andrea Telli); Erica Ryan, University of Maryland doctoral dissertation, 2025
2024
Court rules against scan-and-lend
Controlled digital lending means a library scans a print book it owns and lends the scan to one reader at a time in place of the paper copy. The Internet Archive lent this way, and from 24 March to 16 June 2020 it dropped the one-reader limit, letting up to 10,000 people borrow a book at once. Four publishers sued over 127 books, and on 4 September 2024 the Second Circuit appeals court held that lending scans of books already sold as e-books is not fair use, the legal defence that allows some copying without permission. The Archive let its deadline to appeal to the Supreme Court, 3 December 2024, pass.
LessonThe ruling covers books "commercially available" as e-books, which leaves out-of-print works an open question.
Credit
Internet Archive (defendant); Hachette, HarperCollins, Wiley and Penguin Random House (plaintiffs); Judges Menashi, Robinson and Kahn
Watch for
Humanoid robots shelve library books
At the 2nd World Humanoid Robot Games in Beijing, robots had 30 minutes inside a working library to collect returns, shelve books and correct misplaced ones, carrying five books at a time on a cart. No results, success rates or error counts were published.
Achieved whena library reports a robot putting books back on open shelves in routine service, with its error rate checked against a staff hand count.
Claimed by
organizers of the World Humanoid Robot Games, August 2026
Libraries buying e-books outright
Library e-books are normally licensed on terms, not owned. Briet, a new service, asks publishers to sell e-books to libraries for "universal, perpetual access", and its backers say "thousands of e-books are already available" from independent publishers. Lenny, open-source software from the Internet Archive's Open Library team for lending owned digital books, is called by its makers an "early stage prototype". A Kickstarter campaign by Flaming Hydra, closing on 27 November 2026 with a goal of $30,000, plans to sell e-books to libraries for lending through Lenny.
Achieved whena named library system outside these groups reports patrons borrowing purchased e-books through one of these systems, with loan counts over a stated period.
Claimed by
the Brick House cooperative and Flaming Hydra (Briet), August 2025; Open Library (Lenny), August 2025; Flaming Hydra (Kickstarter), November 2026
Author payments from every loan
UK PLR payments used to be worked out from loans at a sample of library services, scaled up to a national estimate. For 2025–26 the British Library Board proposes 13.63 pence per loan, worked out from "national loans data collected from all statutory public libraries in the UK".
Achieved whenthe government publishes the final 2025–26 rate and the payment report confirms it came from full national data.
Claimed by
the British Library Board, in a government consultation open from 21 October to 15 November 2026
Open
01
Phone shelf-reading, scored by hand
A tool that takes phone photos or a slow video pan of a shelf, reads the call-number label on each spine (the shelf address printed on the label), and lists every book out of order. It is scored in public against a librarian's hand check of the same shelves, a routine task called shelf-reading.
The briefCredit · unclaimed02
Fine-free effects beyond Illinois
A public table of the dates public libraries dropped overdue fines, for every library district in at least three more US states, with a source for each date, and a rerun of Ryan's comparison of visits and circulation for those states.
The briefCredit · unclaimed
Open 01 · the brief
Phone shelf-reading, scored by hand
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A tool that takes phone photos or a slow video pan of a shelf, reads the call-number label on each spine (the shelf address printed on the label), and lists every book out of order. It is scored in public against a librarian's hand check of the same shelves, a routine task called shelf-reading.
Why it is still open
Robot shelf-readers need an RFID tag in every book. A*STAR's AuRoSS robot reached "up to 99 per cent" accuracy in its makers' 2016 trials, and Miami University's ShelvAR app (2011) needed special coded tags on each spine. Camera-only tools now exist as small projects, such as ShelfCheck and ShelfFinder, but none publishes measured accuracy against a hand count, and ShelfFinder says its shelf-label reading "haven't been tried yet".
Built with what exists
a phone camera; open-source text readers such as PaddleOCR or Tesseract, or a vision-language model such as Claude, Gemini or GPT, to read labels; a short sorting function for Dewey or LC call numbers; a local model to keep photos on a laptop (see the Note "A first stable number for local AI"); a spreadsheet for the hand count.
First step
Get one branch's permission to photograph 5 shelves outside opening hours, hand-read them for ground truth, and run a first pass. Do not photograph people or the holds shelf, whose slips show borrowers' names.
Done looks like
across every book on 50 shelves, a published table of labels read correctly and of misshelf flags against the librarian's hand check, giving the share of real misshelves caught and the share of flags that were real, with photos and ground truth posted so others can re-score them.
Open 02 · the brief
Fine-free effects beyond Illinois
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of the dates public libraries dropped overdue fines, for every library district in at least three more US states, with a source for each date, and a rerun of Ryan's comparison of visits and circulation for those states.
Why it is still open
Ryan (2025) writes that hers "is the first to evaluate the causal effects of fine-free policies", and that research is thin partly because of "the absence of a centralized data source" on when libraries dropped fines. She built that data for Illinois only, 528 of 615 districts, and stopped at 2019 because almost every Illinois library dropped fines in 2020.
Built with what exists
the IMLS Public Libraries Survey files, yearly counts per library with no borrower records; library websites, board minutes and the Wayback Machine for adoption dates; R or Python with a standard method for staggered before-and-after comparisons, such as Callaway and Sant'Anna's.
First step
Download the survey files for one state, list its districts, and email their directors the question Ryan asked; 52% replied to her. Work only with library-level counts, never borrowing records.
Done looks like
adoption dates, or "never dropped fines", for at least 90% of districts in three states, each with a source; a second person re-checks a random 50; effect estimates with confidence intervals and the code posted.