Industries / Government, education and community / Government / Parks and recreation
Recreation programs
Running the classes, sports, camps, play sessions and events that public recreation departments offer, and measuring what they do for the people who take part.
5
Milestones achieved, from 1886 to 2018.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1886
A sand pile for children
Dr. Marie Zakrzewska, born in Berlin, had seen children playing on sand piles in Berlin's parks, and in 1886 she brought the idea to Boston as its first "sand garden". Its success led to more sand piles the next year in the yards of two Boston chapels, and sand gardens and kindergartens spread quickly to other American cities.
Credit
Dr. Marie Zakrzewska
1906
A national playground group
The Playground Association of America was founded in 1906 to promote public playgrounds to communities across the country. On 14 August 1965 five organizations, among them the National Recreation Association, the American Recreation Society, the National Conference on State Parks and the American Institute of Park Executives, merged to form the National Recreation and Park Association (NRPA).
Credit
the Playground Association of America; NRPA's founding organizations
1974
Streets closed for play
Bogotá's Ciclovía, started in 1974 and run by the city's sports and recreation institute, closes main roads to motor traffic so people can cycle, walk and skate. In 2009 it covered 97 km and drew 600,000 to 1.4 million users each time it ran. A 2011 analysis found its health benefit from physical activity was worth 3.23 to 4.26 times its cost, which was about US$6.0 per person a year; the ratios for Medellín, Guadalajara and San Francisco were 1.83, 1.02 to 1.23, and 2.32.
Credit
Bogotá's Instituto Distrital de Recreación y Deporte; analysis by Felipe Montes, Olga L. Sarmiento and colleagues (Journal of Urban Health)
2004
A free weekly 5 km
In October 2004 Paul Sinton-Hewitt organized a timed run in Bushy Park, London, for 13 club runners, with five more helping. By June 2026 parkrun had staged its millionth free, timed, volunteer-run event, counting 5 km runs and 2 km junior runs. Reuters reported more than 12 million registered runners, 2,800 parkruns in 23 countries and about a quarter of a million people taking part each week.
Credit
Paul Sinton-Hewitt; parkrun's volunteers
2018
Park prescriptions tested
In the SHINE trial, doctors at a US clinic serving low-income families counseled 78 parents about nature and gave each a map of local parks, a journal and a pedometer. About two-thirds, chosen at random, were also offered group park outings with free transport and food. Stress fell by 1.71 points on a standard stress scale across the whole group, with no difference between the two arms, and each extra park visit a week went with lower stress.
LessonOrganized outings added nothing measurable to a doctor's advice and a map; what tracked lower stress was how often parents went to a park.
Credit
Nooshin Razani and colleagues (PLOS ONE, February 2018)
Watch for
An AI front desk
Raleigh, North Carolina, tested Daisy, an AI avatar on a kiosk in the lobby of the Barwell Road Community Center, which answers questions about city parks, programs and opening hours. It draws on RaleighNC.gov, the city's May–August 2026 Leisure Ledger and the Parks Annual Report, spoke English and Spanish with four more languages planned, and could not see login-only records such as those in RecTrac, the city's recreation software. The city warned that "Because Daisy is still learning, it may not get everything right," reports that the pilot was completed in July 2026, and says the results will shape a customer experience center in the new City Hall, due in 2027.
Achieved whenthe city or an outside tester publishes how often Daisy's answers were right, or Daisy goes into routine service.
Claimed by
City of Raleigh, May 2026
Open
01
What city programs cost
A public dataset of the programs listed in 100 city recreation guides, such as swim lessons, youth sports, camps and senior fitness classes, with ages, number and length of sessions, resident and non-resident price, and place. An AI document reader extracts them from the guides' PDFs and web pages and people check a sample. Then compare the price of an hour of instruction across cities.
The briefCredit · unclaimed02
Residents' questions, with answers
A public test of 300 questions that residents ask a recreation department, for each of three cities, such as whether the outdoor pool has a lifeguard on Sundays or whether a four-year-old can join a class, each with an answer linked to the city's own page. Then score AI assistants against it, including general models given the city's guide.
The briefCredit · unclaimed
Open 01 · the brief
What city programs cost
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public dataset of the programs listed in 100 city recreation guides, such as swim lessons, youth sports, camps and senior fitness classes, with ages, number and length of sessions, resident and non-resident price, and place. An AI document reader extracts them from the guides' PDFs and web pages and people check a sample. Then compare the price of an hour of instruction across cities.
Why it is still open
Cities make these comparisons by hand when they set fees; in 2025 Lawrence, Kansas, charted what other Kansas and US cities charge for memberships and day passes. A search for a cross-city dataset of program prices found only individual cities' fee schedules.
Built with what exists
AI models that read PDFs, such as Claude, Gemini and GPT (see the Note "Seven more flagged"); Python with pandas; a shared spreadsheet; a hand check of a random sample.
First step
Download five cities' current guides, extract every swim lesson listing, and check them all by hand.
Done looks like
100 cities and at least 10,000 program listings, each linked to its source page, with a hand check of 500 listings finding at least 98% of prices and age ranges correct.
Open 02 · the brief
Residents' questions, with answers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of 300 questions that residents ask a recreation department, for each of three cities, such as whether the outdoor pool has a lifeguard on Sundays or whether a four-year-old can join a class, each with an answer linked to the city's own page. Then score AI assistants against it, including general models given the city's guide.
Why it is still open
Raleigh's Daisy pilot ended in July 2026 with no published accuracy figures, and the city warned that it "may not get everything right." I found no public test set of questions about recreation programs or park rules.
Built with what exists
each city's published guide, fee schedule and facility pages; AI models such as Claude, Gemini and GPT given those documents; a scoring rubric of right, partly right, wrong and "said it did not know"; bilingual reviewers for the Spanish version (see the Note "Answering the question behind the question").
First step
Write 30 questions from one city's current guide with answers linked to its pages, and score one model given the guide.
Done looks like
300 questions for each of three cities in English and Spanish, answers checked by two people, at least four assistants scored, and a table of results by type of question.