Industries / Construction and property / Pest control
Prevention and sealing
Keeping pests out by closing the ways into buildings, taking away the food, water and shelter that let them settle, and running programs across whole buildings, cities or regions that do this before anyone reaches for poison.
5
Milestones achieved, from 1908 to 2022.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1908
San Francisco builds out plague
Plague, carried by rat fleas, broke out in San Francisco's Chinatown in 1900. Rupert Blue of the US Public Health and Marine-Hospital Service ended that first outbreak by killing all the rats he could and cutting people off from the rest: landlords were pushed to lay concrete under houses and in back areas, light shafts and basements, and to tear down buildings that could not be fixed. When plague came back in 1907, with 55 cases by September, the city asked President Theodore Roosevelt for help, and the Surgeon General put Blue in command again. His teams examined about 154,000 rats, and the city passed ordinances requiring concrete floors in markets, stables and chicken coops. The second outbreak had 160 human cases, of whom 77 died. The city's 1909 account, written by Frank Morton Todd for the Citizens' Health Committee, says the plague had been "built out".
LessonThe city's own record credits changing the buildings, not trapping and poison alone, with ending the disease.
Credit
Rupert Blue and the US Public Health and Marine-Hospital Service; San Francisco's Citizens' Health Committee
1950
Alberta keeps rats out
Norway rats spread northwest across Saskatchewan from the 1920s at about 24 km a year and were first found in Alberta, on a farm near Alsask, in the summer of 1950. Alberta declared them a pest under its 1942 Agricultural Pests Act, which already obliged every person and municipality to destroy declared pests, and set up a Rat Control Zone 600 km long and 29 km wide along the Saskatchewan border. Inspectors visit 2,000 to 4,000 premises in the zone each year. Known infestations rose from one in 1950 to 573 in 1955, then fell after 1959 to between one and five a year, by the province's count, and Alberta remains essentially free of rats. Its 1951 pamphlet already told farmers to remove harborage and food and to rat-proof buildings.
LessonA law that made everyone responsible was in place before the rats arrived, yet it still took almost ten years of training and public education to bring the problem in hand.
Credit
William Lobay, the Alberta Department of Agriculture, and the municipalities in the zone
2009
Screens keep mosquitoes out
In 2006 and 2007 a randomized trial in and near Farafenni, The Gambia, assigned 500 homes to one of three groups: untreated screens on windows and doors with the eaves closed, a screened ceiling, or nothing. Homes without screening caught 37.5 malaria mosquitoes per light trap a night, against 15.2 with full screening and 19.1 with screened ceilings. Anemia was found in 19% of children in unscreened homes and 12% in both screened groups, though malaria parasite rates did not differ (The Lancet, 2009). After a year, 71% of screened doors and 85% of ceilings had some damage, averaging fewer than five holes a door and seven a ceiling.
LessonScreens with no insecticide cut indoor mosquitoes by more than half in fully screened homes, but they wear and need upkeep.
Credit
Matthew Kirby, Steve Lindsay and colleagues, funded by the UK Medical Research Council
2009
One visit beats routine spraying
Integrated pest management (IPM) combines cleaning, building repairs and the least toxic pesticides used only where needed. Researchers compared it with scheduled spraying in 13 public housing buildings in Brooklyn and Manhattan, covering 280 apartments. After a single IPM visit, apartments had fewer cockroaches at three months, lower cockroach allergen in kitchens and beds at six months, and less pesticide use, and residents rated building services higher (Environmental Health Perspectives, 2009).
LessonA single, easily repeated visit did better than regular spraying, where earlier IPM studies had relied on repeated visits and heavy cleaning.
Credit
Daniel Kass, Wendy McKelvey and colleagues, in New York City public housing
2022
Poison alone fails in Salvador
In Pau da Lima, a low-income area of Salvador, Brazil, the city's zoonosis center ran a rat poison campaign in 2015 that visited 939 households. The share of infested households was 70% before and 72% and 65% three and six months after, with no change in trap catches (PLOS ONE, 2022). In the same area, the city covered open sewers and built paved squares, roads and sidewalks between early 2014 and mid-2015, alongside poison in two of three valleys. The share of tracking plates (dusted plates that record footprints) showing rodent marks fell from 70% to 28% and had not returned to its earlier level by 2017; access to sewers was the condition most closely tied to infestation (Scientific Reports, 2022).
LessonPoison by itself left infestation where it was; fixing sewers and paving changed it for years.
Credit
Federico Costa, Hussein Khalil and colleagues (first authors A. C. Pertile and A. M. Awoniyi); the City of Salvador, which did the works
Watch for
Sealed bins cut city rats
New York City's Department of Sanitation says 311 reports of rat sightings fell in each of 12 months compared with the same month a year earlier, from 25,230 (December 2023 to November 2024) to 20,025 (December 2024 to November 2025), a 20.6% drop. It ties the fall to the rule that buildings with 1 to 9 homes put their trash in sealed bins from November 2024. Calls to 311 count complaints, not rats. An outside case study published in July 2026 estimates that the bin rules cut rodent complaints by about 12 per community district a month and found inspection failures falling under some of its methods but not others; it also reports that its main assumption fails a statistical test and calls its inspection result directional support, not confirmation.
Achieved whena peer-reviewed or other independent study using the city's rat inspection results, not only 311 calls, finds fewer rats where bins were required than where they were not.
Claimed by
NYC Department of Sanitation, December 2025
Open
01
Measure entry gaps by phone
A phone app that photographs a gap under a door or around a pipe, vent or cable with a printed reference card in view, outlines the opening and reports its narrowest width in millimeters. It flags any gap a mouse could pass and turns a walk-around into a list of gaps, each with a photo, location and width, for the owner and the pest controller.
The briefCredit · unclaimed02
Spot rat-prone lots in photos
A model that looks at street-level photos of New York City lots and predicts which ones a city inspector will fail for rat activity, by finding what feeds and shelters rats: bagged trash on the curb, bare soil, overgrown yards, cracked foundations. Score it against the city's inspection results on districts it has not seen, and compare it with a model that uses only past inspection records.
The briefCredit · unclaimed
Open 01 · the brief
Measure entry gaps by phone
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone app that photographs a gap under a door or around a pipe, vent or cable with a printed reference card in view, outlines the opening and reports its narrowest width in millimeters. It flags any gap a mouse could pass and turns a walk-around into a list of gaps, each with a photo, location and width, for the owner and the pest controller.
Why it is still open
The US Centers for Disease Control and Prevention says a mouse can fit through a hole 6 mm (1/4 inch) across, and US food plants must take "effective measures" to keep pests out (21 CFR 117.35). Inspectors still judge gaps by eye or with a hand gauge. FieldProxy says its vision agent finds entry points in inspection photos, but it publishes no accuracy figures and does not say it measures widths. Searches found no published test of gap widths measured by phone against calipers.
Built with what exists
A phone camera; a printed ArUco marker card for scale, read with OpenCV; Meta's Segment Anything Model 2 to outline the opening; a phone with a depth sensor, such as an iPhone Pro's lidar, as a second method; digital calipers and feeler gauges for the true widths; Label Studio for labels.
First step
Photograph 50 gaps under doors and around pipes in your own building with a printed marker card in view, and measure each with calipers.
Done looks like
300 gaps at 30 or more buildings, each measured with calipers and by the app, with a published median width error (target: under 1 mm), the share of gaps 6 mm or wider that the app flags (target: 95% or more), and all photos and measurements released.
Open 02 · the brief
Spot rat-prone lots in photos
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A model that looks at street-level photos of New York City lots and predicts which ones a city inspector will fail for rat activity, by finding what feeds and shelters rats: bagged trash on the curb, bare soil, overgrown yards, cracked foundations. Score it against the city's inspection results on districts it has not seen, and compare it with a model that uses only past inspection records.
Why it is still open
New York's health department publishes its rodent inspections as open data: 3.16 million rows, about 406,000 of them "Failed for Rat Activity" and 132,000 more failed for rat activity and another reason (counts from the city's data service, November 2026). A 2024 Rio de Janeiro study used street view and satellite images to find mosquito breeding sites and matched them with egg-trap counts, but searches of papers and GitHub found no model that predicts rat inspection results from street images; the projects found model the inspection records or 311 calls.
Built with what exists
NYC Open Data's Rodent Inspection dataset (p937-wjvj); street-level images from Mapillary, which are openly licensed; an open object detector such as Ultralytics YOLO11 or Grounding DINO to count trash bags, bins, bare soil and plants; scikit-learn for the records-only baseline.
First step
Download one borough's 2024–2025 inspection results and match 500 inspected lots to Mapillary images taken within a year of the inspection.
Done looks like
A model scored on held-out community districts, with its area under the ROC curve (a ranking score where 0.5 is chance and 1.0 is perfect) reported beside the records-only baseline; a published map; code and the list of images released.