{"schema_version":"1.1","publication":{"name":"Hooray Machine","url":"https://hooraymachine.com","role":"Editorial summaries of linked external sources, not original research."},"access":{"mode":"public","public_discovery_enabled":true,"note":"Published stories are readable without an application API key."},"filters":{"q":"","category":null,"topic":null,"evidence":null,"from":null,"to":null,"source_from":null,"source_to":null,"page":1,"limit":20},"items":[{"id":"culture-southampton-ai-reads-port-books","topics":[{"id":"languages-culture","name":"Languages & Culture"},{"id":"science-nature","name":"Science & Nature"}],"article":{"version":1,"sections":[{"kind":"scene","text":"From Easter 1565, customs officials in England’s ports wrote down every ship that came and went: its name, home port, master, cargo and the merchants who paid duty. About 20,000 port books survive from the period, each in the hand of whichever clerk kept it, and the team scanned the leather-bound originals at the National Archives in Kew. On an earlier project, the Southampton team typed in about 53,000 voyages by hand.","sourceUrls":["https://maritimebritain.org/english_merchant_shipping/the_crows_nest/?story_id=4","https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page"]},{"kind":"clever-bit","text":"The team photographed some 100,000 pages and taught AI to read them. In its 2023 project blog, the team described how it works: historians transcribe sample pages by hand as “ground truth,” and neural networks learn from them to turn pixels into text. The AI had to learn each official’s handwriting and spellings that drifted over the decades, then pull out the ship, port, master and cargo from entries that clerks didn’t all write in the same order.","sourceUrls":["https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page","https://maritimebritain.org/english_merchant_shipping/the_crows_nest/?story_id=4","https://geographical.co.uk/news/how-ai-transformed-centuries-old-handwriting-into-vast-database-of-britains-merchant-shipping-past"]},{"kind":"result","text":"The finished database and interactive map cover hundreds of thousands of voyages and more than 3,000 ports at home and abroad, over 1,000 of them in Britain. Coal appears in more than 101,000 cargo entries, the most of any commodity. The cargo lists also turn up children’s toys from China, sheep guts, bowstrings and Venice treacle, a remedy people believed could cure almost anything. In 1696–97, European markets made up about 80 percent of the recorded value of England’s overseas trade.","sourceUrls":["https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page","https://www.bbc.co.uk/news/articles/cr93z9zq5l9jo","https://geographical.co.uk/news/how-ai-transformed-centuries-old-handwriting-into-vast-database-of-britains-merchant-shipping-past"]},{"kind":"quote","text":"Without AI extracting and organising the data, this would have taken years.","speaker":"Professor Craig Lambert, historian and lead researcher, University of Southampton","sourceUrls":["https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page","https://geographical.co.uk/news/how-ai-transformed-centuries-old-handwriting-into-vast-database-of-britains-merchant-shipping-past"]},{"kind":"hooray","text":"Two centuries of handwritten customs paperwork is now a free map anyone can explore, right down to lost places like the town of Dunwich and Ravenspurn, a once-thriving Yorkshire port that sank into the North Sea.","sourceUrls":["https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page","https://www.bbc.co.uk/news/articles/cr93z9zq5l9jo","https://maritimebritain.org/english_merchant_shipping/the_crows_nest/?story_id=4"]},{"kind":"next","text":"Professor Jason Sadler of Southampton’s GeoData Institute says the map lets the public, scholars and citizen historians dig in by period, trading hub, ship, goods and route. Lambert says the AI’s output still needed manual checking, and the team hasn’t published an error rate, so treat any single record as a lead back to the original port book.","sourceUrls":["https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page","https://geographical.co.uk/news/how-ai-transformed-centuries-old-handwriting-into-vast-database-of-britains-merchant-shipping-past"]}],"shareBlurb":"Southampton historians taught AI to read about 100,000 pages of handwritten port books from 1565 on. Now anyone can explore a free map of hundreds of thousands of voyages, from coal cargoes to Venice treacle and lost ports like Ravenspurn."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/culture-southampton-ai-reads-port-books","api_url":"https://hooraymachine.com/api/v1/stories/culture-southampton-ai-reads-port-books","title":"AI read about 100,000 pages of handwritten port books, giving everyone a free map of Britain’s sea trade from Elizabeth I to George III","summary":"University of Southampton historians trained AI to read about 100,000 pages of handwritten port books held at the National Archives, an unbroken customs record of voyages from 1565. The result is a free interactive map and database of hundreds of thousands of voyages. It shows coal topped the cargo lists and Europe made up about 80 percent of England’s overseas trade in 1696–97.","category":{"id":"everyday","name":"Delight"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Live public resource. The University of Southampton’s release (October 19, 2026) describes the AI-built database and interactive map, now online at maritimebritain.org, with named comments from lead researcher Professor Craig Lambert, co-lead Dr John McAleer and Professor Jason Sadler of the GeoData Institute. BBC News and Geographical reported it on October 20. The project’s 2023 blog explains the handwriting-recognition method.","caveat":"The figures come from the University of Southampton’s own release; BBC News and Geographical reported them without separate checks. The team hasn’t published an error rate for the AI transcription, and Lambert says the output still needed manual checking. The port books record what customs officials wrote down, so the map shows recorded trade only. Early findings, such as the 80 percent European share, are from one sample year, 1696–97."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page"},"sources":[{"title":"AI reads centuries-old handwriting to reveal Britain’s vast historic shipping trade","url":"https://www.southampton.ac.uk/news/2026/10/ai-reads-centuriesold-handwriting-to-reveal-britains-vast-historic-shipping-trade.page","published_date":"2026-10-19","source_type":"Primary: University of Southampton news release with named researchers","primary":true},{"title":"How AI revealed 400 years of seafaring history","url":"https://www.bbc.co.uk/news/articles/cr93z9zq5l9jo","published_date":"2026-10-20","source_type":"BBC News report (Ariana Botero, South of England)"},{"title":"How AI transformed centuries-old handwriting into vast database of Britain’s merchant shipping past","url":"https://geographical.co.uk/news/how-ai-transformed-centuries-old-handwriting-into-vast-database-of-britains-merchant-shipping-past","published_date":"2026-10-20","source_type":"Geographical magazine news report (Victoria Heath)"},{"title":"Unlocking the Potential of ‘Big Data’ in Historical Research","url":"https://maritimebritain.org/english_merchant_shipping/the_crows_nest/?story_id=4","published_date":"2023-02-03","source_type":"Project blog (The Crow’s Nest, Maritime Britain) explaining the handwriting-recognition method"}],"dates":{"edition_date":"2026-10-30","reviewed_at":"2026-10-30T12:00:03Z","first_curated_at":"2026-10-30T12:00:03.584Z","modified_at":"2026-10-30T12:00:03.584Z"},"image":{"url":"https://hooraymachine.com/images/1d3716d1d15ad6eaf783a40342aff462e0537e64868fa40bb39f56e165e369c4.webp","alt":"Clay illustration of an open handmade clay ledger book with blank cream pages on a tabletop, with a small yellow clay magnifying glass resting over it, the only smiling face in the scene. Little clay sailing ships rise from the pages and sail along orange clay routes across a cream clay sea to tiny forest-green clay harbours.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"safety-deepmind-ai-model-called-isaias-landfall","topics":[{"id":"safety-disaster","name":"Safety & Disaster Response"},{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"In early October, Hurricane Isaias was crossing the Gulf of Mexico, and the forecast models disagreed about where it would come ashore. The traditional physics-based models leaned west, toward southeastern Louisiana or coastal Mississippi, with a slower storm. On October 9, Isaias made landfall near Destin, on the Florida Panhandle, as a Category 2 hurricane.","sourceUrls":["https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias","https://www.npr.org/2026/10/19/nx-s1-5996666/ai-models-are-changing-how-forecasters-predict-the-paths-of-hurricanes","https://michaelrlowry.substack.com/p/new-post-isaias-analyses-show-not"]},{"kind":"clever-bit","text":"Google DeepMind’s model skips the physics equations that traditional models solve. It found patterns in thousands of past storms and the atmospheric data around them, then takes the current picture and projects the storm forward. Forecasters at the National Hurricane Center (NHC) never lean on a single model, so it sits alongside the physics models. The NHC started folding it into official forecasts last year, as part of a blend of the best performers.","sourceUrls":["https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias","https://www.npr.org/2026/10/19/nx-s1-5996666/ai-models-are-changing-how-forecasters-predict-the-paths-of-hurricanes","https://www.nhc.noaa.gov/verification/pdfs/Verification_2025.pdf"]},{"kind":"result","text":"From its first runs on Isaias, the DeepMind model pointed at the western Florida Panhandle. A preliminary verification by HurriLab, published by hurricane specialist Michael Lowry, found it was closest to the real landfall spot at Destin and timed landfall within a few hours, three days out. The physics-based European and American ensembles put landfall too far west and a full day or more too late. Over the 2025 season, the NHC’s own verification report found the DeepMind ensemble slightly beat its official Atlantic track forecasts from 12 to 72 hours.","sourceUrls":["https://michaelrlowry.substack.com/p/new-post-isaias-analyses-show-not","https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias","https://www.nhc.noaa.gov/verification/pdfs/Verification_2025.pdf"]},{"kind":"quote","text":"Google DeepMind, in particular, has really proven itself the last few years as the gold standard model.","speaker":"Michael Lowry, hurricane specialist, WPLG-TV Miami","sourceUrls":["https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias"]},{"kind":"hooray","text":"A model that learned from old storms pointed at the right stretch of coast from day one, and hurricane forecasters now have it in the blend they use to warn people when to get ready and when to leave.","sourceUrls":["https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias","https://michaelrlowry.substack.com/p/new-post-isaias-analyses-show-not"]},{"kind":"next","text":"NHC Director Michael Brennan’s team gives Google DeepMind regular feedback to improve the model. AI models can be updated in weeks or months, while traditional ones get updated once a year at best, and Lowry says the gains in a single year would have taken a decade or more. The NHC sticks with a consensus of the best models, because, as Brennan puts it, every storm has different forecast challenges.","sourceUrls":["https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias"]}],"shareBlurb":"An AI weather model from Google DeepMind called Hurricane Isaias’s Florida Panhandle landfall while physics models aimed at Louisiana, and timed it within a few hours three days out. The National Hurricane Center now blends it into its forecasts."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/safety-deepmind-ai-model-called-isaias-landfall","api_url":"https://hooraymachine.com/api/v1/stories/safety-deepmind-ai-model-called-isaias-landfall","title":"Google DeepMind’s AI model called Hurricane Isaias’s Florida Panhandle landfall while physics-based models pointed toward Louisiana","summary":"Hurricane Isaias hit Destin, Florida, as a Category 2 storm on October 9. Google DeepMind’s AI model, which the National Hurricane Center has used in its official forecasts since 2025, pointed to the western Florida Panhandle from the start, while conventional models first sent the storm toward southeastern Louisiana. A preliminary HurriLab analysis found it timed landfall within a few hours, three days out.","category":{"id":"public-life","name":"Public Life"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Real-world forecast checked against a real landfall. NPR (October 26, 2026) reports the model’s Isaias call and the National Hurricane Center’s use of it, with named comments from NHC Director Michael Brennan and hurricane experts Michael Lowry and Brian McNoldy. A preliminary HurriLab verification, published by Lowry on October 16, compares every major model on Isaias’s track and landfall. The NHC’s own 2025 verification report found the Google DeepMind ensemble mean slightly beat the official Atlantic track forecast at 12 to 72 hours.","caveat":"On Isaias’s strength, NHC Director Michael Brennan says the NHC’s forecast had 20% to 40% more skill than the DeepMind model, and the model didn’t track Hurricane Simon in the eastern Pacific well. The HurriLab numbers are preliminary, and HurriLab is a weather tech company. One storm proves little on its own; the season-long NHC figures are the steadier evidence. Forecasters call the model a black box, and the NHC still issues a human-made forecast from a blend of models. Google didn’t answer NPR’s questions, and NPR notes Google is one of its financial supporters."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias"},"sources":[{"title":"How AI is transforming hurricane forecasts","url":"https://www.npr.org/2026/10/26/nx-s1-6012820/ai-hurricane-forecast-google-deepmind-isaias","published_date":"2026-10-26","source_type":"Primary: NPR news report (Jessica Meszaros, WUSF) with named NHC and expert sources","primary":true},{"title":"AI models are changing how forecasters predict the paths of hurricanes","url":"https://www.npr.org/2026/10/19/nx-s1-5996666/ai-models-are-changing-how-forecasters-predict-the-paths-of-hurricanes","published_date":"2026-10-19","source_type":"NPR Morning Edition transcript (NHC Director Michael Brennan on Isaias intensity)"},{"title":"New Post-Isaias Analyses Show Not All AI Models are Created Equal","url":"https://michaelrlowry.substack.com/p/new-post-isaias-analyses-show-not","published_date":"2026-10-16","source_type":"Independent expert analysis (Michael Lowry, Eye on the Tropics) with preliminary HurriLab verification"},{"title":"National Hurricane Center Forecast Verification Report: 2025 Hurricane Season","url":"https://www.nhc.noaa.gov/verification/pdfs/Verification_2025.pdf","published_date":"2026-03-30","source_type":"Official NOAA National Hurricane Center verification report"}],"dates":{"edition_date":"2026-10-29","reviewed_at":"2026-10-29T11:52:46Z","first_curated_at":"2026-10-29T11:52:46.478Z","modified_at":"2026-10-29T11:52:46.478Z"},"image":{"url":"https://hooraymachine.com/images/67a97e7757b3ccb5a486a761fd8b0cdd5f336d4864ab9691484861cb04395d9d.webp","alt":"Clay illustration of a small handmade clay coastline with a curving forest-green shore, a few faceless clay beach houses and palm trees, and a swirling cream-and-orange clay hurricane over a cream clay sea. On the shore stands a little forest-green clay weather station with a yellow beacon, the only smiling face in the scene, shining a yellow clay beam that points straight to the spot where the storm is heading, while a faded clay arrow points off to the left.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"public-uk-child-maintenance-ai-flags-missed-payments","topics":[{"id":"government-public-life","name":"Government & Public Life"}],"article":{"version":1,"sections":[{"kind":"scene","text":"When separated parents use the Child Maintenance Service’s Collect and Pay service, the government takes the money from one parent and passes it to the other. Some months the money stops. The UK’s Department for Work and Pensions (DWP) estimates that around 8,000 child maintenance cases a month see an unplanned breakdown in payments. Once a case slips into arrears, getting the money moving again takes far more work than keeping it on track.","sourceUrls":["https://www.nao.org.uk/reports/dwps-approach-to-innovation/","https://www.gov.uk/algorithmic-transparency-records/dwp-child-maintenance-service-cms-predictive-analytics-compliance"]},{"kind":"clever-bit","text":"At the start of every month, a machine-learning model reads the payment history, changes of circumstance and recent letters and calls on about 100,000 cases that are currently paying, then scores how likely each one is to miss an upcoming payment. Each score comes with a short explanation of what is driving the risk. More than 40 caseworkers use the high, medium and low flags to decide whose case to open first, check the file and contact the parent before a payment slips.","sourceUrls":["https://www.gov.uk/algorithmic-transparency-records/dwp-child-maintenance-service-cms-predictive-analytics-compliance","https://www.nao.org.uk/reports/dwps-approach-to-innovation/"]},{"kind":"stat","value":"715","text":"more child maintenance cases a month kept out of arrears, by DWP’s own estimate","sourceUrls":["https://www.nao.org.uk/press-releases/dwp-is-innovating-to-improve-services-for-customers-and-value-for-taxpayers-say-nao/","https://www.nao.org.uk/reports/dwps-approach-to-innovation/","https://www.globalgovernmentfinance.com/nao-backs-dwp-innovation-efforts-urges-more-consistent-lesson-sharing/"]},{"kind":"result","text":"The model became part of everyday work in March 2024, run by a new Early Intervention Team. DWP told the National Audit Office (NAO) that the early contact it prompts stops an extra 715 cases a month from moving into arrears and saves staff time equal to 49 full-time employees. In testing, the final proof-of-concept model was right 70% of the time, up from 65% for the first version in May 2022. Building it cost £625,000, and it costs about £460,000 a year to run.","sourceUrls":["https://www.nao.org.uk/reports/dwps-approach-to-innovation/","https://www.nao.org.uk/press-releases/dwp-is-innovating-to-improve-services-for-customers-and-value-for-taxpayers-say-nao/"]},{"kind":"hooray","text":"In 715 more cases a month, by DWP’s estimate, the maintenance a child is owed keeps arriving, and a caseworker sorts out the wobble early instead of chasing a debt later. The NAO, Parliament’s independent spending watchdog, picked the model out as one of six DWP innovations with lessons for the rest of government.","sourceUrls":["https://www.nao.org.uk/reports/dwps-approach-to-innovation/","https://www.nao.org.uk/press-releases/dwp-is-innovating-to-improve-services-for-customers-and-value-for-taxpayers-say-nao/"]},{"kind":"next","text":"DWP plans a full evaluation once more data has gone into the model in late 2027. The Child Maintenance Service is also looking at other ways machine learning could help it run the service.","sourceUrls":["https://www.nao.org.uk/reports/dwps-approach-to-innovation/"]}],"shareBlurb":"A machine-learning model scores about 100,000 UK child maintenance cases each month for the risk of a missed payment. Caseworkers call the riskiest first, and DWP estimates 715 more cases a month stay out of arrears."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/public-uk-child-maintenance-ai-flags-missed-payments","api_url":"https://hooraymachine.com/api/v1/stories/public-uk-child-maintenance-ai-flags-missed-payments","title":"715 more UK child maintenance cases a month stay out of arrears after machine learning flags payments at risk, DWP estimates","summary":"The UK’s Department for Work and Pensions estimates that 715 more child maintenance cases a month now stay out of arrears. Each month a machine-learning model scores about 100,000 up-to-date cases for the risk of a missed payment, and caseworkers contact the parents most at risk first. The National Audit Office reported the figures on October 21.","category":{"id":"public-life","name":"Public Life"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Government-reported results in an independent audit report: the National Audit Office’s October 21, 2026 report on DWP’s approach to innovation sets out the model’s timeline, costs, test accuracy and DWP’s estimates of its effect. The UK government’s algorithmic transparency record describes how the model works and how caseworkers use it. Global Government Finance (October 21, 2026) reported the same figures.","caveat":"The 715 cases and the 49 staff equivalent are DWP’s own estimates, based on a test-and-learn approach and staff feedback. The NAO reported them without measuring them itself, and DWP plans a full evaluation only after it adds more data to the model in late 2027. The 65% and 70% accuracy figures come from testing the proof-of-concept models. The model ranks cases; caseworkers check each case file and decide what to do."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.nao.org.uk/reports/dwps-approach-to-innovation/"},"sources":[{"title":"DWP’s approach to innovation","url":"https://www.nao.org.uk/reports/dwps-approach-to-innovation/","published_date":"2026-10-21","source_type":"Primary: National Audit Office report (UK public spending watchdog), HC 616","primary":true},{"title":"DWP is innovating to improve services for customers and value for taxpayers, say NAO","url":"https://www.nao.org.uk/press-releases/dwp-is-innovating-to-improve-services-for-customers-and-value-for-taxpayers-say-nao/","published_date":"2026-10-21","source_type":"National Audit Office press release"},{"title":"DWP: Child Maintenance Service (CMS) - Predictive Analytics Compliance","url":"https://www.gov.uk/algorithmic-transparency-records/dwp-child-maintenance-service-cms-predictive-analytics-compliance","published_date":"2025-10-30","source_type":"UK government algorithmic transparency record (how the model works), updated July 13, 2026"},{"title":"NAO backs DWP innovation efforts but urges ‘more consistent’ lesson-sharing","url":"https://www.globalgovernmentfinance.com/nao-backs-dwp-innovation-efforts-urges-more-consistent-lesson-sharing/","published_date":"2026-10-21","source_type":"Independent trade news coverage (Global Government Finance, Ian Hall)"}],"dates":{"edition_date":"2026-10-28","reviewed_at":"2026-10-28T11:59:29Z","first_curated_at":"2026-10-28T11:59:29.729Z","modified_at":"2026-10-29T11:52:46.729Z"},"image":{"url":"https://hooraymachine.com/images/6a789f6f10ee82737ec0f001ef9025165c3d3c3b4aadc665d1a82f85bc73e373.webp","alt":"Clay illustration of a little forest-green clay scanner with a glowing lens, the only smiling face in the scene, standing beside a stack of clay envelopes while a small clay hand catches one envelope before it falls; across the room an orange clay piggy bank sits on a wooden table next to a toy train as a coin drops into it.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"science-cape-town-ai-spots-toads-in-tunnel-photos","topics":[{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Peninsula Road cuts straight through Cape Town’s False Bay Nature Reserve, right across a breeding ground of the endangered Western Leopard Toad, a species found only in the Western Cape. Every rainy winter the toads cross to breed, and cars kill them. So the nonprofit Nature Connect built six tunnels under the road, two in 2021 and four in 2023, and put cameras at the ends. Toads are cold-blooded and don’t trip the heat sensors that usually trigger camera traps, so these cameras take a photo every 15 seconds, all night. That adds up to a mountain of pictures of empty tunnel floor.","sourceUrls":["https://natureconnect.earth/2026/07/30/toad-tunnels-ai-endangered-species/","https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/"]},{"kind":"clever-bit","text":"Joshua Salako, a master’s student at the African Institute for Mathematical Sciences (AIMS) South Africa, trained a toad detector, AmphiLens, that starts from just 20 hand-labelled photos. It sharpens the contrast of dark night frames, tries four detection models, and picks which photos people should label next so it learns fastest. Llewellyn Williams of Nature Connect built it into the team’s staff portal, where a person confirms every photo it flags. A second tool compares each toad’s lifelong pattern of dark, yellow-edged patches to find the same animal at both ends of a tunnel.","sourceUrls":["https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/","https://github.com/joshsalako/amphilens"]},{"kind":"stat","value":"1,176,380","text":"tunnel photos this season, checked by AI; 384 of them show a toad","sourceUrls":["https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/"]},{"kind":"result","text":"Those 384 photos add up to 130 toad visits once repeat shots taken minutes apart are counted once. The busiest night, 17 August, brought 16. Since October the system has checked 2,160,620 photos from the 2023 to 2026 seasons. On 21 August a toad the team calls T2026-7 was photographed at one end of tunnel 4 at 00:38 and at the other end eight minutes later, and 35 of the 36 points the computer matched on its back line up.","sourceUrls":["https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/","https://www.capetownetc.com/conservation/ai-spots-endangered-western-leopard-toad/"]},{"kind":"hooray","text":"A small conservation team can now show its tunnels at work, one toad at a time, without anyone reading a million photos by eye. Nature Connect says just one toad has died on the tunnelled stretch of road in three years.","sourceUrls":["https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/","https://natureconnect.earth/2026/07/30/toad-tunnels-ai-endangered-species/"]},{"kind":"next","text":"The team is still confirming this season’s matches and plans to share how many separate toads made the crossing once the checks are done. Salako has published AmphiLens openly, with its code, trained models and an install guide, so any camera-trap project can use it.","sourceUrls":["https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/","https://github.com/joshsalako/amphilens"]}],"shareBlurb":"Cape Town’s toad tunnels took 1,176,380 photos this season. An AI model found the 384 that show an endangered Western Leopard Toad, and one toad was caught crossing under the road in eight minutes."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/science-cape-town-ai-spots-toads-in-tunnel-photos","api_url":"https://hooraymachine.com/api/v1/stories/science-cape-town-ai-spots-toads-in-tunnel-photos","title":"AI checked 1,176,380 tunnel photos this season and found the 384 showing endangered Cape Town toads using their road underpass","summary":"Cameras in six toad tunnels under Peninsula Road in Cape Town’s False Bay Nature Reserve snap a photo every 15 seconds, all night. This season that came to 1,176,380 photos. An AI model built by AIMS South Africa student Joshua Salako picked out the 384 that show an endangered Western Leopard Toad, and the Nature Connect team confirms every one.","category":{"id":"science-nature","name":"Science & Nature"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Nature Connect, the Cape Town conservation nonprofit that runs the tunnels, reported the 2026 season figures on October 26, 2026, and Cape Town Etc (Angelica Rhoda, October 27, 2026) reported the same figures. The detector comes from Joshua Salako’s master’s research at AIMS South Africa with Lorène Jeantet (AIMS) and Kim Gordon (Nature Connect), and its code and trained models are public.","caveat":"All the counts come from Nature Connect, and Cape Town Etc repeats them; nobody outside the project has audited them. The 130 figure counts toad visits rather than separate toads, and the team is still checking possible matches by hand. A fast toad can cross between photos taken 15 seconds apart, so some crossings go unrecorded. The figure of one road death in three years on the tunnelled stretch also comes from Nature Connect’s own monitoring."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/"},"sources":[{"title":"AI is changing how we track the endangered Western Leopard Toad: what a million tunnel photographs are telling us","url":"https://natureconnect.earth/2026/10/26/joshua-salakos-toad-detector-at-work-the-2026-western-leopard-toad-season/","published_date":"2026-10-26","source_type":"Primary: Nature Connect (Cape Town conservation nonprofit that runs the toad tunnels) 2026 season report","primary":true},{"title":"Look: AI spots endangered Western Leopard Toads among a million pictures","url":"https://www.capetownetc.com/conservation/ai-spots-endangered-western-leopard-toad/","published_date":"2026-10-27","source_type":"Independent local news report (Cape Town Etc, Angelica Rhoda)"},{"title":"AI Confirms Toad Tunnels Are Saving an Endangered Species","url":"https://natureconnect.earth/2026/07/30/toad-tunnels-ai-endangered-species/","published_date":"2026-07-30","source_type":"Background: Nature Connect project history and road-death monitoring"},{"title":"AmphiLens: open-source camera-trap detector for amphibians","url":"https://github.com/joshsalako/amphilens","published_date":"2026-10-23","source_type":"Code repository by Joshua Salako (README, install guide, citation; last updated October 23, 2026)"}],"dates":{"edition_date":"2026-10-27","reviewed_at":"2026-10-27T11:55:33Z","first_curated_at":"2026-10-27T11:55:33.739Z","modified_at":"2026-10-27T12:21:28.690Z"},"image":{"url":"https://hooraymachine.com/images/5956e17ef140036e37af217969699db0c34e4e586ee3b55f59c3f866363ff246.webp","alt":"Clay illustration of a small spotted yellow-and-brown toad crawling through a cream clay tunnel under a dark green road at night, while a little forest-green clay camera at the tunnel mouth, the only smiling face in the scene, shines a soft yellow beam on the toad; reeds and a pond sit behind.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"human-mcmaster-dairy-ai-camera-lameness","topics":[{"id":"food-farming","name":"Food & Farming"}],"article":{"version":1,"sections":[{"kind":"scene","text":"George McMaster’s farming career began with milking eight cows in a byre, and his own eyes were the health check. The family herd near Islandmagee in County Antrim now numbers 300. Lameness, often from painful foot lesions, is a major welfare problem in dairy farming, and the early signs are easy to miss in cows you see every single day.","sourceUrls":["https://www.bbc.com/news/articles/cw80zk9k51j1o","https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2023.1111057/full"]},{"kind":"clever-bit","text":"Let a camera watch every cow walk, every day. The CattleEye system in the McMasters’ milking parlour gives each cow her own ID, records a short video of her daily, analyses how she moves and flags changes, often before they are visible to the human eye. George’s son Jack, who is taking over the farm, signed up when a local developer approached him. George now checks the herd on his phone.","sourceUrls":["https://www.bbc.com/news/articles/cw80zk9k51j1o"]},{"kind":"result","text":"Jack says the camera flags cows the family would never have thought of as lame, and much earlier, so they can check them before they lose condition and get sick. The system has also been tested by vets. In a 2023 University of Liverpool pilot study on three dairy farms, its mobility scores largely agreed with two experienced vets, and it caught 52% of cows that turned out to have painful foot lesions, against 29% for one of the vets.","sourceUrls":["https://www.bbc.com/news/articles/cw80zk9k51j1o","https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2023.1111057/full"]},{"kind":"quote","text":"There’s cows there we would never have thought of as lame when we were just watching them every day, but whenever the camera’s watching they’re picking them up so much earlier.","speaker":"Jack McMaster, dairy farmer, McMaster family farm, Islandmagee, County Antrim","sourceUrls":["https://www.bbc.com/news/articles/cw80zk9k51j1o"]},{"kind":"hooray","text":"Sore cows on the McMaster farm get looked at sooner, before a limp turns into a sick cow. Jack’s reasoning is simple: healthier cows give more milk and are happier, and happy cows make happy farmers. For George, a third-generation farmer, it is “absolutely incredible to move from all hand-done to sitting on the phone, watching what’s happening outside.”","sourceUrls":["https://www.bbc.com/news/articles/cw80zk9k51j1o"]},{"kind":"next","text":"Jack is taking over the farm with the camera already part of the daily routine. Other Northern Ireland farmers in the same BBC report are moving the same way, from robotic milking in County Tyrone to apps that keep every cow’s records in one place. Gareth McDonald, who built one of those apps, says the AI should advise farmers, not make the decisions for them.","sourceUrls":["https://www.bbc.com/news/articles/cw80zk9k51j1o"]}],"shareBlurb":"George McMaster started out milking eight cows in a byre. Now an AI camera films his family’s 300 cows every day and flags changes in how they walk, and his son Jack says it catches lame cows they’d never have spotted."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/human-mcmaster-dairy-ai-camera-lameness","api_url":"https://hooraymachine.com/api/v1/stories/human-mcmaster-dairy-ai-camera-lameness","title":"An AI camera in a County Antrim milking parlour is helping the McMaster family spot lame cows in their 300-strong herd much earlier","summary":"George McMaster started farming by milking eight cows in a byre and checking them with his own eyes. Now an AI camera in the milking parlour near Islandmagee films each of the family’s 300 cows every day and flags changes in how they move. His son Jack says it picks out lame cows the family would have missed.","category":{"id":"science-nature","name":"Science & Nature"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"In daily use on the McMaster family’s dairy farm near Islandmagee, County Antrim, as reported by BBC News NI (Louise Cullen, October 26, 2026). The earlier detection is the family’s own account. The same camera system, CattleEye, was tested in a peer-reviewed 2023 pilot study by University of Liverpool vets in Frontiers in Veterinary Science (6,040 mobility scores on three dairy farms), which CattleEye funded.","caveat":"The earlier catches are the McMasters’ own account, and the BBC report gives no figures for their herd. The validation study was small, three farms, and was funded by CattleEye, though the company had no part in collecting or analysing the data. In that study the camera caught about half of the cows that had painful foot lesions, so it misses some, and the farmers still check every cow it flags."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.bbc.com/news/articles/cw80zk9k51j1o"},"sources":[{"title":"From agriculture to AI-griculture: How tech is changing farming","url":"https://www.bbc.com/news/articles/cw80zk9k51j1o","published_date":"2026-10-26","source_type":"Primary: BBC News NI report (Louise Cullen, agriculture and environment correspondent) with on-record interviews with George and Jack McMaster","primary":true},{"title":"Initial validation of an intelligent video surveillance system for automatic detection of dairy cattle lameness","url":"https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2023.1111057/full","published_date":"2023-06-13","source_type":"Peer-reviewed pilot study (Frontiers in Veterinary Science), Anagnostopoulos et al., University of Liverpool; funded by CattleEye"}],"dates":{"edition_date":"2026-10-26","reviewed_at":"2026-10-26T11:53:13Z","first_curated_at":"2026-10-26T11:53:13.777Z","modified_at":"2026-10-26T11:53:13.777Z"},"image":{"url":"https://hooraymachine.com/images/b47c61f4f8a04147cd1f5f1453d75c0426f7143096e42ae2cad6a02d2c202e2a.webp","alt":"Clay illustration of three black-and-white dairy cows walking through a milking parlour lane while a smiling green camera on a wooden rail shines a yellow beam on the middle cow, with an orange ring around its front hoof and green fields and a stone farmhouse outside.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"delight-ai-decodes-elephant-rumble-choruses","topics":[{"id":"everyday","name":"Delight"},{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Elephants keep in touch with rumbles, deep calls that partly dip below what human ears can hear. About 60% of those rumbles overlap with other elephants’ calls in choruses, especially during big moments like greeting ceremonies. Most studies have left the overlapping calls out, because pulling apart several low voices at once is the same puzzle as following one conversation at a crowded party.","sourceUrls":["https://www.nature.com/articles/s41598-026-68525-3","https://www.smithsonianmag.com/smart-news/african-elephants-often-produce-choruses-of-low-pitched-rumbles-now-scientists-are-using-ai-to-decode-the-joint-vocalizations-180989620/"]},{"kind":"clever-bit","text":"Treat the whole chorus as one message. The team used 902 recordings of adult female elephants made in Amboseli National Park between 1986 and 2022 by observers who noted in real time what each family was doing. The calls fell into six situations, including coos to calves, contact calls between separated elephants, let’s-go calls before the group moves and greeting rumbles at reunions. Machine-learning models learned the sound patterns of each situation, such as which low frequencies carry the most power, then tried to name the situation for recordings they had not trained on.","sourceUrls":["https://www.nature.com/articles/s41598-026-68525-3","https://www.scientificamerican.com/article/scientists-are-decoding-elephant-rumble-calls-including-greetings-and-coos/"]},{"kind":"result","text":"The models picked the right situation better than chance, for single rumbles and for overlapping choruses. A model trained only on single rumbles could also read choruses, which suggests each elephant in a chorus uses the same kind of call it would make alone. In cadenced, contact and greeting exchanges, the callers’ rumbles also became more acoustically similar as the exchange went on. The study was published October 1 in Scientific Reports.","sourceUrls":["https://www.nature.com/articles/s41598-026-68525-3"]},{"kind":"quote","text":"It is pretty exciting that we’re able to train a model to look at an entire chorus and have some interpretation of exactly what’s going on.","speaker":"Sara Keen, senior research scientist, Earth Species Project","sourceUrls":["https://www.scientificamerican.com/article/scientists-are-decoding-elephant-rumble-calls-including-greetings-and-coos/"]},{"kind":"hooray","text":"Researchers can now study the overlapping majority of elephant conversation instead of setting it aside. The paper says future work can include chorus recordings to get bigger samples, which matters for animals where it is hard to tell who is calling, such as whales and elephants. The study also adds evidence that an elephant chorus works as a group signal.","sourceUrls":["https://www.nature.com/articles/s41598-026-68525-3"]},{"kind":"next","text":"The team wants playback experiments, where researchers play recorded rumbles to elephants and watch how they react, to test whether rumbles growing more alike helps a family agree on a plan. Keen hopes to try similar methods on other animals that call in choruses, such as orcas and crows. The acoustic features and analysis code are public.","sourceUrls":["https://www.nature.com/articles/s41598-026-68525-3","https://www.scientificamerican.com/article/scientists-are-decoding-elephant-rumble-calls-including-greetings-and-coos/"]}],"shareBlurb":"About 60% of elephant rumbles overlap in choruses, and most studies skip them. Researchers in Kenya ran machine learning on 902 recordings and found choruses carry clues about what the herd is doing, and that rumbles grow more alike when a family coordinates."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/delight-ai-decodes-elephant-rumble-choruses","api_url":"https://hooraymachine.com/api/v1/stories/delight-ai-decodes-elephant-rumble-choruses","title":"Machine learning listened to wild elephant choruses in Kenya and found the overlapping rumbles carry clues about what the herd is doing","summary":"Researchers at Earth Species Project, ElephantVoices and the Amboseli Trust for Elephants ran machine learning on 902 rumble recordings from wild elephants in Kenya’s Amboseli National Park. The models named what the elephants were doing from overlapping choruses better than chance, and during reunions and group plans, the callers’ rumbles grew more alike as the exchange went on.","category":{"id":"everyday","name":"Delight"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed, open-access paper in Scientific Reports (published October 1, 2026) by Mahon, Keen, Poole and colleagues at Earth Species Project, ElephantVoices and the Amboseli Trust for Elephants, built on 902 recordings of wild Amboseli elephants and tested against shuffled baselines. Independently reported by Scientific American (October 1, 2026) and Smithsonian (October 8, 2026), with comment from Cambridge zoologist Arik Kershenbaum, who was not involved.","caveat":"The models beat chance but were still wrong more often than right. For chorus recordings they named the situation correctly roughly a third of the time. All the recordings come from adult females in one population, the situation labels rely on experienced field observers, and the team has not yet tested how elephants respond when rumbles grow more alike."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.scientificamerican.com/article/scientists-are-decoding-elephant-rumble-calls-including-greetings-and-coos/"},"sources":[{"title":"Social context information and acoustic convergence in elephant vocal choruses","url":"https://www.nature.com/articles/s41598-026-68525-3","published_date":"2026-10-01","source_type":"Primary: peer-reviewed open-access paper (Scientific Reports), Mahon et al., Earth Species Project, ElephantVoices and Amboseli Trust for Elephants","primary":true},{"title":"Scientists are decoding elephant rumble calls, including ‘greetings’ and ‘coos’","url":"https://www.scientificamerican.com/article/scientists-are-decoding-elephant-rumble-calls-including-greetings-and-coos/","published_date":"2026-10-01","source_type":"Independent news report (Scientific American, Jackie Flynn Mogensen)"},{"title":"African Elephants Often Produce Choruses of Low-Pitched Rumbles. Now, Scientists Are Using A.I. to Decode the Joint Vocalizations","url":"https://www.smithsonianmag.com/smart-news/african-elephants-often-produce-choruses-of-low-pitched-rumbles-now-scientists-are-using-ai-to-decode-the-joint-vocalizations-180989620/","published_date":"2026-10-08","source_type":"Independent news report (Smithsonian, Sarah Kuta)"}],"dates":{"edition_date":"2026-10-25","reviewed_at":"2026-10-25T11:55:22Z","first_curated_at":"2026-10-25T11:55:23.259Z","modified_at":"2026-10-25T11:55:23.259Z"},"image":{"url":"https://hooraymachine.com/images/a01b22b1ff379b2a375512da450a1be332527798bcfbf143109904d81822a31c.webp","alt":"Clay illustration of a family of four grey elephants on a savanna, with braided yellow and orange sound ribbons flowing from their trunks to a smiling green recorder on a tripod, an acacia tree and a snowy mountain behind.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"delight-ai-counts-singing-indri-lemurs","topics":[{"id":"everyday","name":"Delight"},{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Indri fill the rainforests of eastern Madagascar with loud, musical duets. These black-and-white animals are the world’s largest lemurs and the only lemurs that sing, using their songs to claim territory from rival groups. They are also critically endangered, and every attempt to keep them in captivity has failed, so saving them means protecting the wild ones. That starts with knowing how many are left, which is hard to work out in steep, tangled forest.","sourceUrls":["https://www.nationalgeographic.com/animals/article/indri-acoustic-monitoring"]},{"kind":"clever-bit","text":"Let the lemurs announce themselves. A University of Turin team strapped 26 recorders to trees across the Maromizaha New Protected Area and had them capture 10 minutes of every half hour for about 13 months. A neural network trained on indri songs combed the morning audio and flagged 5,395 songs. The team then borrowed a counting method first used for animals in the ocean: combine how many songs the recorders catch, how often a group sings and how likely a song is to reach a recorder, and out comes an estimate of how many groups live nearby.","sourceUrls":["https://conbio.onlinelibrary.wiley.com/doi/10.1111/csp2.70428","https://www.nationalgeographic.com/animals/article/indri-acoustic-monitoring"]},{"kind":"result","text":"Maromizaha made a fair test because researchers already knew the answer. Scientists have followed habituated indri groups there for 15 years, and the real density is 6.17 groups per square kilometer. The recordings put it at 5.55, with the true figure inside the estimate’s range. The peer-reviewed paper, published October 9 in Conservation Science and Practice, calls that strong agreement.","sourceUrls":["https://conbio.onlinelibrary.wiley.com/doi/10.1111/csp2.70428"]},{"kind":"quote","text":"Every song we detect is a data point that could guide where conservation efforts need to go next.","speaker":"Valeria Ferrario, postdoctoral fellow, University of Turin","sourceUrls":["https://www.nationalgeographic.com/animals/article/indri-acoustic-monitoring"]},{"kind":"hooray","text":"Conservation teams can count a critically endangered lemur by listening. There are no loudspeakers playing calls that can change indri behavior, and far less time hiking survey lines in bad weather. The authors say the setup needs only a network of recorders, about one per square kilometer, which keeps costs within reach for groups with small budgets.","sourceUrls":["https://conbio.onlinelibrary.wiley.com/doi/10.1111/csp2.70428"]},{"kind":"next","text":"The team hopes to use the method across the indri’s whole range in Madagascar. National Geographic reports that the University of Turin and partners also support GERP, the local group that manages Maromizaha and helps residents grow native trees to replant cleared forest.","sourceUrls":["https://conbio.onlinelibrary.wiley.com/doi/10.1111/csp2.70428","https://www.nationalgeographic.com/animals/article/indri-acoustic-monitoring"]}],"shareBlurb":"Indri, the world’s largest lemurs, sing. University of Turin researchers let 26 forest recorders and an AI model listen, then estimated 5.55 indri groups per square kilometer in a Madagascar forest where the real figure was 6.17. A gentler way to count a critically endangered species."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/delight-ai-counts-singing-indri-lemurs","api_url":"https://hooraymachine.com/api/v1/stories/delight-ai-counts-singing-indri-lemurs","title":"An AI listening network estimated how many singing indri lemur families live in a Madagascar forest, and it landed close to the real number","summary":"University of Turin researchers hung 26 recorders in Madagascar’s Maromizaha forest and used an AI model to pick out the songs of the indri, the world’s largest lemur. Their peer-reviewed estimate of 5.55 indri groups per square kilometer came close to the known 6.17, a promising way to count a critically endangered species without disturbing it.","category":{"id":"everyday","name":"Delight"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed, open-access paper in Conservation Science and Practice (published October 9, 2026) by Ferrario and colleagues at the University of Turin, validated against a forest where the real indri group density was already known from 15 years of fieldwork. Independently reported by National Geographic on October 22, 2026. Tested in one forest so far.","caveat":"The estimate’s range is wide, from about 2.8 to 10.7 groups per square kilometer. About 38% of the AI’s song detections were false and had to be corrected for, and the method counts family groups, not individual lemurs. So far it has been tested in one forest where the answer was already known, and new sites need some field recordings of their own to work out how far songs carry."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.nationalgeographic.com/animals/article/indri-acoustic-monitoring"},"sources":[{"title":"Passive acoustic monitoring for the estimation of density of the singing lemur (Indri indri)","url":"https://conbio.onlinelibrary.wiley.com/doi/10.1111/csp2.70428","published_date":"2026-10-09","source_type":"Primary: peer-reviewed open-access paper (Conservation Science and Practice), Ferrario et al., University of Turin","primary":true},{"title":"How to Save the World’s Largest Lemurs? Listen to Their Songs.","url":"https://www.nationalgeographic.com/animals/article/indri-acoustic-monitoring","published_date":"2026-10-22","source_type":"Independent news report (National Geographic, Sruthi Gurudev)"}],"dates":{"edition_date":"2026-10-24","reviewed_at":"2026-10-24T11:49:42Z","first_curated_at":"2026-10-24T11:49:43.265Z","modified_at":"2026-10-24T11:49:43.265Z"},"image":{"url":"https://hooraymachine.com/images/f3978be3c18b00c86233652a6be6a563547a021780fe9a4a0d4e9edd29616461.webp","alt":"Clay stop-motion scene: a lush clay rainforest of chunky green leaves and vines under a misty cream sky; two black-and-white clay indri lemurs cling to a tall tree trunk on the right with heads tipped up, eyes closed and mouths open in song, while wavy yellow clay sound ribbons float across to a small square forest-green clay audio recorder with a round black microphone, sitting on a stump tied with an orange band; the recorder is the only smiling face, with two bead eyes and a gentle smile; conceptual illustration.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"science-umass-ai-counts-bees-by-buzz","topics":[{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Counting bees has an awkward catch. The usual way to check how bees are doing is to set out small, colorful cups of soapy water, which trap and kill the insects so experts can identify them later. Getting an accurate count can kill tens of thousands of bees. That’s a rough trade when populations of many pollinators, bumble bees included, have been crashing for decades.","sourceUrls":["https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations"]},{"kind":"clever-bit","text":"Instead of trapping bees, the team listens to them. Bees make distinctive sounds when they fly, and bumble bees make a separate buzz when they shake pollen loose from a flower. Laura Figueroa, an ecologist at UMass Amherst, and colleagues at Cornell trained two AI sound models on 594 field recordings made over five years at 43 sites in the northeastern U.S. One model picks out bee activity in general. The other tells honey bee flight from bumble bee flight and bumble bee pollen-shaking.","sourceUrls":["https://resjournals.onlinelibrary.wiley.com/doi/10.1002/3070-5436.70006","https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations"]},{"kind":"result","text":"In the team’s peer-reviewed paper, published October 12, the models were checked against trained observers counting bees in flower patches, and the two sets of counts tracked each other closely. At the models’ strictest setting, 93% to 97% of the sounds they flagged really were bees. Both models are now built into Raven Pro, a sound-analysis program from the Cornell Lab of Ornithology that many wildlife researchers already use, and the team’s recordings are public on Dryad.","sourceUrls":["https://resjournals.onlinelibrary.wiley.com/doi/10.1002/3070-5436.70006","https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations","https://datadryad.org/dataset/doi:10.5061/dryad.sbcc2frnh"]},{"kind":"quote","text":"I’ve been working with bees for over a decade and can tell a bumble bee from a honeybee in the field by sound. If my brain can easily distinguish between bee categories, then we should be able to train an AI model to do it, too.","speaker":"Laura Figueroa, assistant professor of environmental conservation, UMass Amherst","sourceUrls":["https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations"]},{"kind":"hooray","text":"Scientists can check on bees by leaving a recorder in a flower patch, and no bees have to die for the count. It also reaches places that are hard to survey in person. Figueroa points to tree canopies and fields during heat waves, and the paper suggests the bumble bee model could help growers of buzz-pollinated crops like tomatoes and blueberries decide how many bumble bee colonies to use and where to put them.","sourceUrls":["https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations","https://resjournals.onlinelibrary.wiley.com/doi/10.1002/3070-5436.70006"]},{"kind":"next","text":"The authors want the models to learn a wider range of bees and other pollinating insects. Figueroa says the tool is meant to work alongside traditional surveys. It hears best up close, and counting individual bee species still depends on traditional traps.","sourceUrls":["https://resjournals.onlinelibrary.wiley.com/doi/10.1002/3070-5436.70006","https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations"]}],"shareBlurb":"Ecologists at UMass Amherst and Cornell trained AI to pick bees out of field recordings by their buzz. The models’ counts tracked trained observers, and they now run in Raven Pro, so a recorder can stand in for soapy traps that kill bees."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/science-umass-ai-counts-bees-by-buzz","api_url":"https://hooraymachine.com/api/v1/stories/science-umass-ai-counts-bees-by-buzz","title":"AI listens to bee buzz to track pollinator activity","summary":"Ecologists at UMass Amherst and Cornell trained two AI sound models to pick bees out of field recordings, including telling honey bees from bumble bees. In a peer-reviewed study, the models’ counts closely tracked trained observers, and they now run inside Raven Pro, a sound tool many wildlife researchers already use. A recorder in a flower patch can stand in for soapy traps that kill bees.","category":{"id":"science-nature","name":"Science & Nature"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed, open-access paper in Approaches in Entomology: Methods, Philosophies, and Ethics (first published October 12, 2026) by researchers at UMass Amherst and Cornell, announced by UMass on October 21, 2026. The models were tested on held-out recordings and against trained field observers, and both are available in Raven Pro. No independent field results from other teams yet.","caveat":"At their strictest setting, the models are right 93% to 97% of the time when they flag a bee, but they miss many bees, catching 13% to 44% of bee sounds. Their surest detections came within about 50 cm (20 inches) of the recorder, and they measure bee activity, not a headcount. The training recordings came mostly from calm, sunny days, so noisy sites may trip them up. Species-level bee counts still depend on traditional traps, and Raven Pro carries a license fee in many high-income countries."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations"},"sources":[{"title":"What’s all the buzz about? Passive acoustic monitoring of bees","url":"https://resjournals.onlinelibrary.wiley.com/doi/10.1002/3070-5436.70006","published_date":"2026-10-12","source_type":"Primary: peer-reviewed open-access paper (Approaches in Entomology: Methods, Philosophies, and Ethics), Figueroa et al., UMass Amherst and Cornell","primary":true},{"title":"Listening to the Buzz: UMass Amherst Ecologist Develops New Tool for Monitoring Bee Populations","url":"https://www.umass.edu/news/article/listening-buzz-umass-amherst-ecologist-develop-new-tool-monitoring-bee-populations","published_date":"2026-10-21","source_type":"University news release (UMass Amherst)"},{"title":"Data: What’s all the buzz about? Passive acoustic monitoring of bees","url":"https://datadryad.org/dataset/doi:10.5061/dryad.sbcc2frnh","published_date":"2026-09-22","source_type":"Open dataset (Dryad)"}],"dates":{"edition_date":"2026-10-23","reviewed_at":"2026-10-23T12:04:06Z","first_curated_at":"2026-10-23T12:04:06.445Z","modified_at":"2026-10-25T13:30:43.149Z"},"image":{"url":"https://hooraymachine.com/images/ff301f4317afa64e491fbeea64614610bccc28dc71a330df8cbe3eac696e5ecf.webp","alt":"Clay stop-motion scene: a sunny clay meadow of chunky yellow and orange flowers under a cream sky, where four small faceless striped clay bees fly and rest on blossoms; on a wooden stake among the flowers stands a small square forest-green clay audio recorder with a round black microphone on top, the only face in the scene, with two little eyes and a smile, while soft yellow clay sound ripples float in the air beside it; conceptual illustration.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"health-childrens-national-er-surge-forecast","topics":[{"id":"health","name":"Health"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Pediatric emergency rooms get crowded, and when the wait drags on, some families give up and go home before a doctor sees their child. That used to happen mostly with milder cases, but staffing shortages and children stuck waiting for hospital beds mean seriously ill kids are now leaving too. In the D.C. area, patients spend an average of more than five hours in the emergency department, according to federal data cited by NBC4 Washington.","sourceUrls":["https://innovationdistrict.childrensnational.org/predicting-risk-before-it-becomes-harm-in-the-pediatric-ed/","https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/"]},{"kind":"clever-bit","text":"The clever bit is that it forecasts the whole waiting room a shift ahead. Doctors at Children’s National Hospital and data scientists at Virginia Tech trained machine-learning models on 512,616 emergency visits from 2018 to 2024, using numbers the hospital already tracks: arrivals, how full the department is, waiting times, children waiting for a bed upstairs, and the time and day. The tool predicts whether two or more high-acuity children will leave unseen over the next eight hours, a window chosen to match staffing shifts, and it updates every hour. In the team’s JACEP Open study, the best models scored 0.85 and 0.86 on a standard accuracy measure where 1.0 is perfect.","sourceUrls":["https://pmc.ncbi.nlm.nih.gov/articles/PMC12799923/","https://innovationdistrict.childrensnational.org/predicting-risk-before-it-becomes-harm-in-the-pediatric-ed/","https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/"]},{"kind":"result","text":"Then it left the lab. Children’s National started using the forecast in fall 2024. When it predicts trouble, Dr. Kenneth McKinley, an emergency physician and one of the study’s authors, gets an automatic notification, and the hospital calls in doctors and nurses it has lined up as backup. McKinley told NBC4 that wait times improved by more than 25% this September compared with September 2023 and 2024, on days when the ER treated more than 200 patients.","sourceUrls":["https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/","https://pmc.ncbi.nlm.nih.gov/articles/PMC12799923/"]},{"kind":"stat","value":"more than 25%","text":"better ER wait times on days with 200+ patients, Sept 2026 vs Sept 2023 and 2024, by the hospital’s count","sourceUrls":["https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/"]},{"kind":"hooray","text":"Hooray: the sickest kids get a better shot at being seen before their families give up and leave. McKinley calls the alert “a real sign that some truly sick patients might be going home,” so the extra hands arrive where the danger is. The forecast never makes decisions about a child’s care; it tells the hospital when to bring in help.","sourceUrls":["https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/"]},{"kind":"next","text":"Next, McKinley wants the forecast to fire earlier in the day so teams can move staff sooner. The study’s authors add that the models were tested on one hospital’s past data, that they still need prospective validation, and that they will need retraining as crowding patterns change.","sourceUrls":["https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/","https://pmc.ncbi.nlm.nih.gov/articles/PMC12799923/"]}],"shareBlurb":"Children’s National and Virginia Tech built an AI forecast that warns, hours ahead, when seriously ill kids may leave the ER unseen. Backup doctors and nurses get called in, and the hospital says busy-day waits improved more than 25%."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/health-childrens-national-er-surge-forecast","api_url":"https://hooraymachine.com/api/v1/stories/health-childrens-national-er-surge-forecast","title":"Children’s National’s AI calls in extra ER staff before sick kids give up waiting, and busy-day waits improved more than 25%","summary":"Doctors at Children’s National Hospital in Washington, D.C., and data scientists at Virginia Tech built machine-learning models that forecast, every hour, when seriously ill children are likely to leave the ER without being seen. When the forecast trips, backup doctors and nurses are called in. The hospital says waits on its busiest September days improved by more than 25%.","category":{"id":"health","name":"Health"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"In use at Children’s National Hospital since fall 2024 (NBC4 Washington, October 16, 2026). The models are described in a peer-reviewed JACEP Open study published online December 24, 2025 (512,616 visits from 2018 to 2024; AUROC 0.85 and 0.86 on 2024 test data). The more-than-25% wait improvement is the hospital’s own figure, and the linked sources cite no independent audit.","caveat":"The wait-time improvement is the hospital’s own before-and-after comparison for busy days, given in a TV interview; the linked sources cite no independent audit and don’t separate the forecast’s effect from other changes. The published study tested the models on one hospital’s past data, and its authors say prospective validation is still to come. The tool only flags when to add staff; clinicians make every care decision."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/"},"sources":[{"title":"How AI is helping Children’s National reduce ER wait times","url":"https://www.nbcwashington.com/news/health/how-ai-is-helping-childrens-national-reduce-emergency-room-wait-times/4168378/","published_date":"2026-10-16","source_type":"Primary: independent TV news report (NBC4 Washington, Eun Yang and Patricia Fantis)","primary":true},{"title":"A Machine Learning Strategy to Predict the Number of High-Acuity Children Who Leave Without Being Seen From the Emergency Department","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12799923/","published_date":"2025-12-24","source_type":"Peer-reviewed study (JACEP Open, open access via PubMed Central), Children’s National Hospital and Virginia Tech"},{"title":"Predicting risk before it becomes harm in the pediatric ED","url":"https://innovationdistrict.childrensnational.org/predicting-risk-before-it-becomes-harm-in-the-pediatric-ed/","published_date":"2026-01-21","source_type":"Hospital research news (Children’s National Innovation District)"}],"dates":{"edition_date":"2026-10-22","reviewed_at":"2026-10-22T11:56:03Z","first_curated_at":"2026-10-22T11:56:03.445Z","modified_at":"2026-10-22T11:56:03.445Z"},"image":{"url":"https://hooraymachine.com/images/6de56a8ed5fb3a1dfa9305b5604ce4083ed41dde1f342a9dd84a21ae586ac6ea.webp","alt":"Clay stop-motion scene: a cozy cream clay children’s hospital waiting room with forest-green benches where a few faceless clay parents sit with small children; on the wall, a small rounded green clay forecast lantern with a yellow beacon, the only smiling face in the scene, has its little eyes turned toward a doorway, where a faceless clay nurse and doctor in orange are hurrying in to help; conceptual illustration.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"public-northumbrian-water-ai-listens-for-leaks","topics":[{"id":"government-public-life","name":"Government & Public Life"},{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"context","text":"Most water leaks hide. Northumbrian Water, in north-east England, traditionally spotted them through visible signs like water pooling or changes in pressure, and it often relied on customers to flag them. “Finding leaks isn’t always straightforward, especially when there’s no visible signs at the surface,” said Jim Howey, its head of water networks.","sourceUrls":["https://www.nwg.co.uk/news-and-media/news-releases/ai-technology-helps-to-hunt-down-leaks/"]},{"kind":"clever-bit","text":"The clever bit is listening. In a trial that started in November 2025, the utility and FIDO Tech fixed more than 1,000 acoustic sensors to fittings along 149 km of water pipe. The sensors record the sound inside the network, and FIDO’s AI analyses it in real time for the noise of water escaping underground. It then gives technicians areas to investigate, even where nothing shows at the surface. The sensors can stay in place for long stretches or move daily to sweep a wider area.","sourceUrls":["https://www.nwg.co.uk/news-and-media/news-releases/ai-technology-helps-to-hunt-down-leaks/","https://www.nwg.co.uk/news-and-media/news-releases/northumbrian-water-projects-scoop-two-national-environmental-awards/"]},{"kind":"result","text":"In winter, temperature changes can raise the risk of bursts and leaks. Over those cold months, the project helped save around 6.4 million litres of water a day, Northumbrian Water says, which it puts at about 25,000 bathtubs. On September 16 the entry, “Reducing the Impact of Winter Water Main Breaks,” won Real-Time Monitoring & Control Project of the Year at the UK’s Environmental Monitoring & Measurement Achievement Awards in Birmingham.","sourceUrls":["https://www.nwg.co.uk/news-and-media/news-releases/northumbrian-water-projects-scoop-two-national-environmental-awards/","https://www.nwg.co.uk/news-and-media/news-releases/ai-technology-helps-to-hunt-down-leaks/","https://www.ess-expo.co.uk/environmental-monitoring-and-measurement-achievement-awards/2026-winners","https://www.northumberlandgazette.co.uk/business/northumbrian-water-projects-scoop-two-national-environmental-awards-9247767"]},{"kind":"stat","value":"6.4 million litres","text":"of water saved a day during last winter’s high-risk months, by Northumbrian Water’s own count","sourceUrls":["https://www.nwg.co.uk/news-and-media/news-releases/northumbrian-water-projects-scoop-two-national-environmental-awards/"]},{"kind":"hooray","text":"Hooray: hidden leaks get found and fixed early, before they grow into bigger bursts, which means less emergency work, fewer surprise interruptions for customers and more drinking water kept in the pipes.","sourceUrls":["https://www.nwg.co.uk/news-and-media/news-releases/ai-technology-helps-to-hunt-down-leaks/"]},{"kind":"next","text":"Northumbrian Water calls the trial part of its ongoing work to cut leakage with new technology. Its releases don’t say how far the sensor network will grow beyond 149 km, and they cite no independent check of the savings, so the next numbers to watch are its own leakage figures.","sourceUrls":["https://www.nwg.co.uk/news-and-media/news-releases/ai-technology-helps-to-hunt-down-leaks/","https://www.nwg.co.uk/news-and-media/news-releases/northumbrian-water-projects-scoop-two-national-environmental-awards/"]}],"shareBlurb":"Northumbrian Water put AI-backed listening sensors on 149 km of pipe. The AI hears hidden leaks, crews fix them early, and the utility says the project saved around 6.4 million litres of water a day last winter."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/public-northumbrian-water-ai-listens-for-leaks","api_url":"https://hooraymachine.com/api/v1/stories/public-northumbrian-water-ai-listens-for-leaks","title":"AI that listens to water pipes helped Northumbrian Water save about 6.4 million litres a day last winter","summary":"Northumbrian Water fixed more than 1,000 listening sensors to 149 km of pipe in north-east England, and AI picked out the sound of water escaping underground so technicians knew where to look. Over last winter’s high-risk months, the utility says, the project helped save around 6.4 million litres of water a day. In September it won a national UK monitoring award.","category":{"id":"public-life","name":"Public Life"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Utility-reported results: Northumbrian Water’s September 27, 2026 release reports the 6.4 million litres a day, the first figure in its June 1, 2026 trial release. The Northumberland Gazette (October 1, 2026) reported the award, and the 2026 EMMA Awards winners list confirms the FIDO Tech project’s win. The linked sources cite no independent audit of the savings.","caveat":"The 6.4 million litres a day is Northumbrian Water’s own figure for one winter on 149 km of pipe, first reported as an early trial result in June; the linked sources cite no independent audit. The AI only points to likely leak areas, and technicians still pinpoint and repair each one. FIDO Tech, which makes the sensors and the AI, entered the project for the award."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.nwg.co.uk/news-and-media/news-releases/northumbrian-water-projects-scoop-two-national-environmental-awards/"},"sources":[{"title":"Northumbrian Water projects scoop two national environmental awards","url":"https://www.nwg.co.uk/news-and-media/news-releases/northumbrian-water-projects-scoop-two-national-environmental-awards/","published_date":"2026-09-27","source_type":"Primary: utility news release (Northumbrian Water Group)","primary":true},{"title":"Northumbrian Water projects scoop two national environmental awards","url":"https://www.northumberlandgazette.co.uk/business/northumbrian-water-projects-scoop-two-national-environmental-awards-9247767","published_date":"2026-10-01","source_type":"Independent local news coverage (Northumberland Gazette, Andrew Coulson)"},{"title":"AI technology helps to hunt down leaks","url":"https://www.nwg.co.uk/news-and-media/news-releases/ai-technology-helps-to-hunt-down-leaks/","published_date":"2026-06-01","source_type":"Utility news release on the trial (Northumbrian Water Group)"},{"title":"EMMAs 2026 Winners","url":"https://www.ess-expo.co.uk/environmental-monitoring-and-measurement-achievement-awards/2026-winners","published_date":"2026-09-16","source_type":"Award organizer winners list (Environmental Monitoring & Measurement Achievement Awards)"}],"dates":{"edition_date":"2026-10-21","reviewed_at":"2026-10-21T11:58:48Z","first_curated_at":"2026-10-21T11:58:48.577Z","modified_at":"2026-10-22T03:44:55.369Z"},"image":{"url":"https://hooraymachine.com/images/0efb9decdc10eb268ad0956a0a8f4b11d5267d2b3993ef434e302890cbaf4772.webp","alt":"Clay stop-motion scene: a cutaway of a quiet cream clay street shows a forest-green water pipe running underground; a small round cream clay listening sensor clamped to the pipe, the only smiling face in the scene, has its little eyes turned toward a tiny bead of water escaping from a pipe joint, with soft yellow sound ripples between them, while above ground a little orange clay repair van has already pulled up over the spot; conceptual illustration.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"safety-alertcalifornia-ai-first-to-spot-51-percent","topics":[{"id":"safety-disaster","name":"Safety & Disaster Response"},{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"On a July day this year, a small crew drove a dusty mountain road to an old fire lookout on Shadequarter Mountain, 4,000 feet above Visalia, California. The tower was built in 1964 and abandoned in the late 1980s. The crew turned an off-the-shelf deer stand into a tripod, bolted on solar panels, two 80-pound batteries and a Starlink dish, and fixed two cameras in front of the tower’s broken windows. A rancher sometimes runs cattle up there, so local officials agreed to add a short fence. “Cows are heavy,” said Colby Nicholson, the crew’s senior engineering supervisor.","sourceUrls":["https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html"]},{"kind":"context","text":"The crew works for Alert California, a UC San Diego program that runs nearly 1,300 cameras on peaks, towers and rooftops across the state’s most fire-prone land. Its budget can approach $14 million a year, mostly from Cal Fire and large utilities.","sourceUrls":["https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html"]},{"kind":"clever-bit","text":"The clever bit is the software that watches the feeds around the clock. At each site, one camera spins a full circle every two minutes. Since 2023, AI has scanned that footage for smoke and alerted fire officials, who check the live view to rule out fog or a rancher burning debris, then decide how many crews to send. Anyone can watch the same feeds on the Alert California website.","sourceUrls":["https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html","https://insider.govtech.com/california/news/ai-cameras-are-spotting-california-wildfires-before-humans-call-911"]},{"kind":"result","text":"The numbers keep growing. In 2024 the cameras picked up 1,668 of the 7,553 wildfires in Cal Fire’s territory and caught 636 of them before anyone called 911, Cal Fire’s Brian York said. In 2025 the system helped Cal Fire respond to about 3,600 wildfire incidents, and in more than half of them it alerted officials before a 911 caller did, program director Neal Driscoll told Route Fifty. This summer Driscoll told The New York Times that 51% of California wildfires are now first spotted through AI.","sourceUrls":["https://insider.govtech.com/california/news/ai-cameras-are-spotting-california-wildfires-before-humans-call-911","https://www.route-fifty.com/public-safety/2026/07/what-california-has-learned-ai-enabled-early-wildfire-detection-system/414566/","https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html"]},{"kind":"quote","text":"The greatest success stories are the fires you’ve never heard about.","speaker":"Falko Kuester, co-director of Alert California and engineering professor, UC San Diego","sourceUrls":["https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html"]},{"kind":"hooray","text":"Hooray: more than half the time, the first alarm about a new California wildfire now comes from AI watching a hilltop camera, which gives crews a head start.","sourceUrls":["https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html","https://www.route-fifty.com/public-safety/2026/07/what-california-has-learned-ai-enabled-early-wildfire-detection-system/414566/"]},{"kind":"next","text":"Driscoll wants to grow the network to 1,800 cameras to fill blind spots like Soda Canyon in Napa County, where the crew added a camera this June because local fire officials wanted eyes on it.","sourceUrls":["https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html"]}],"shareBlurb":"A crew hauls cameras up to old fire lookouts, and AI watches the feeds for smoke. Alert California’s director says AI is now first to spot 51% of California’s wildfires, often before anyone calls 911."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/safety-alertcalifornia-ai-first-to-spot-51-percent","api_url":"https://hooraymachine.com/api/v1/stories/safety-alertcalifornia-ai-first-to-spot-51-percent","title":"AI watching nearly 1,300 hilltop cameras is now first to spot 51% of California’s wildfires, its director says","summary":"Alert California, a UC San Diego program, runs nearly 1,300 cameras on the state’s fire-prone peaks, and since 2023 AI has scanned the footage for smoke. In 2024 the cameras beat 911 callers to 636 fires. In 2025 they beat them on more than half of about 3,600 incidents. This summer the program’s director said AI now spots 51% of California’s wildfires first.","category":{"id":"public-life","name":"Public Life"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Independent reporting: The New York Times (John Branch, October 12, 2026, syndicated by The Philadelphia Inquirer on October 15) reports program director Neal Driscoll’s 51% figure and the camera network’s size and budget. Route Fifty (July 1, 2026) reports Driscoll’s 2025 figures, and Bay Area News Group (via Industry Insider California, September 23, 2025) reports Cal Fire’s 2024 counts.","caveat":"The 51% figure comes from the program’s director, as reported by The New York Times; the linked reporting cites no independent audit, and each year’s figure is counted a little differently. Cameras can’t see through mountains or into some canyon bottoms. The AI, built by Chico company DigitalPath, can mistake dust, clouds or geothermal steam for smoke, so people check every alert before crews roll. Early warning can’t stop every fire: rooftop cameras caught the 2025 Palisades fire, but high winds and very dry land still made it a disaster."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html"},"sources":[{"title":"How California is using AI and cameras to stop wildfires","url":"https://www.inquirer.com/news/nation-world/remote-cameras-ai-spot-wildfires-crew-satellite-wifi-california-20261015.html","published_date":"2026-10-15","source_type":"Primary independent news reporting (The New York Times, John Branch; syndicated by The Philadelphia Inquirer; NYT original October 12, 2026)","primary":true},{"title":"What California has learned from an AI-enabled early wildfire detection system","url":"https://www.route-fifty.com/public-safety/2026/07/what-california-has-learned-ai-enabled-early-wildfire-detection-system/414566/","published_date":"2026-07-01","source_type":"Secondary independent news coverage (Route Fifty)"},{"title":"AI Cameras Are Spotting California Wildfires Before Humans Call 911","url":"https://insider.govtech.com/california/news/ai-cameras-are-spotting-california-wildfires-before-humans-call-911","published_date":"2025-09-23","source_type":"Secondary independent news coverage (Bay Area News Group, via Industry Insider California)"}],"dates":{"edition_date":"2026-10-20","reviewed_at":"2026-10-20T12:03:44Z","first_curated_at":"2026-10-20T12:03:44.674Z","modified_at":"2026-10-22T03:44:55.563Z"},"image":{"url":"https://hooraymachine.com/images/0a3b4fa05f4fe19293846066d5ee2dee21206017edda5b0a03225f5f08ff71fa.webp","alt":"Clay stop-motion scene: on a soft green clay mountaintop stands an old orange-and-cream clay fire lookout tower, with a small tripod of solar panels beside it; a round cream clay camera on the tower, the only smiling face in the scene, looks out across the valley at a thin wisp of cream smoke rising from a far ridge, while a little orange clay fire engine already heads down a winding road toward it; conceptual illustration.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"human-baroness-campbell-ai-voice-lords","topics":[{"id":"government-public-life","name":"Government & Public Life"},{"id":"accessibility","name":"Accessibility"}],"article":{"version":1,"sections":[{"kind":"scene","text":"On the afternoon of October 13, the House of Lords was starting a long debate on the Health Bill. First, the Chief Whip announced that Baroness Jane Campbell would speak using an AI-generated voice, and that she should get some leeway on time. The chamber answered: “Hear, hear!”","sourceUrls":["https://hansard.parliament.uk/lords/2026-10-13/debates/E93C0EEB-C2C2-4E81-9566-C10EB6EA4AF8/HealthBill"]},{"kind":"context","text":"Campbell, 67, has spinal muscular atrophy and has campaigned for disabled people’s rights for more than 40 years. Her voice has weakened over time. Since 2012 she has been the only peer allowed to bring a personal assistant into the chamber to read her words, a break from a rule dating to 1707.","sourceUrls":["https://today.rtl.lu/news/world/in-a-first-ai-helps-uk-lawmaker-find-her-voice-in-parliament-270802641","https://www.disabilitynewsservice.com/disabled-peer-uses-ai-tech-to-recreate-the-voice-she-used-to-have-as-she-delivers-speech-to-house-of-lords/","https://www.independentliving.co.uk/industry-news/historic-first-for-ai/"]},{"kind":"clever-bit","text":"The clever bit is where the voice came from. Campbell uploaded old recordings of herself, including a 2012 BBC radio interview, to ElevenLabs’ voice-cloning tool, which copied her tone, rhythm and pitch. She wrote the speech, and the tool turned it into audio that played in the chamber. Older methods meant recording almost 400 sentences before losing your voice, and the result sounded robotic, an ElevenLabs speech therapist told AFP.","sourceUrls":["https://www.disabilitynewsservice.com/disabled-peer-uses-ai-tech-to-recreate-the-voice-she-used-to-have-as-she-delivers-speech-to-house-of-lords/","https://today.rtl.lu/news/world/in-a-first-ai-helps-uk-lawmaker-find-her-voice-in-parliament-270802641"]},{"kind":"quote","text":"In 2012, I needed another person to speak my words for me, when I tired. In 2026, I can use technology to speak them directly.","speaker":"Baroness Campbell of Surbiton, crossbench peer, House of Lords","sourceUrls":["https://www.independentliving.co.uk/industry-news/historic-first-for-ai/"]},{"kind":"result","text":"She opened by introducing peers to “Baroness Campbell mark 2,” then made her case for personal health budgets, the NHS funding she relies on for her own support. The next speaker, the Bishop of Newcastle, noted “the historic nature of her excellent speech.”","sourceUrls":["https://hansard.parliament.uk/lords/2026-10-13/debates/E93C0EEB-C2C2-4E81-9566-C10EB6EA4AF8/HealthBill","https://www.disabilitynewsservice.com/disabled-peer-uses-ai-tech-to-recreate-the-voice-she-used-to-have-as-she-delivers-speech-to-house-of-lords/"]},{"kind":"hooray","text":"Hooray: after years of having others read her words, Baroness Campbell argued her case to Parliament in her own voice.","sourceUrls":["https://hansard.parliament.uk/lords/2026-10-13/debates/E93C0EEB-C2C2-4E81-9566-C10EB6EA4AF8/HealthBill","https://today.rtl.lu/news/world/in-a-first-ai-helps-uk-lawmaker-find-her-voice-in-parliament-270802641"]},{"kind":"next","text":"The Lords Procedure Committee has agreed she can record future speeches in advance with the AI voice, as a reasonable adjustment under the Equality Act. Campbell hopes other disabled people who struggle with public speaking will see that it “does not have to be a barrier to inclusion at the highest level.”","sourceUrls":["https://www.disabilitynewsservice.com/disabled-peer-uses-ai-tech-to-recreate-the-voice-she-used-to-have-as-she-delivers-speech-to-house-of-lords/"]}],"shareBlurb":"Baroness Jane Campbell, whose voice has weakened with spinal muscular atrophy, wrote her own House of Lords speech and had an AI copy of her once-clear voice deliver it. The Lords Chief Whip called it a first in a British Parliament."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/human-baroness-campbell-ai-voice-lords","api_url":"https://hooraymachine.com/api/v1/stories/human-baroness-campbell-ai-voice-lords","title":"Baroness Campbell spoke to the House of Lords in her own once-clear voice, rebuilt by AI from old recordings","summary":"Baroness Jane Campbell has spinal muscular atrophy, and since 2012 an assistant has read parts of her House of Lords speeches. On October 13 she wrote her Health Bill speech herself. An AI copy of her voice, built from older recordings, delivered it. The Lords Chief Whip called it a first in a British Parliament.","category":{"id":"public-life","name":"Public Life"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Official record: Lords Hansard for the Health Bill second reading on October 13, 2026, where the Chief Whip announced the AI-generated voice as “a first in a British Parliament,” followed by Baroness Campbell’s full speech. Independent coverage: AFP (October 13) and Disability News Service (October 15). Independent Living’s October 11 preview carries quotes from Baroness Campbell and the Lord Speaker.","caveat":"This is one person’s accommodation, approved by the Lords Procedure Committee. Her speeches have to be prepared in advance, and the method needs good recordings of a voice from before it changed. The voice comes from a commercial tool, ElevenLabs, and the company’s figures about how it works are its own. The Lord Speaker welcomed the move while noting that some uses of AI raise justifiable concerns."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://hansard.parliament.uk/lords/2026-10-13/debates/E93C0EEB-C2C2-4E81-9566-C10EB6EA4AF8/HealthBill"},"sources":[{"title":"Health Bill, Second Reading, House of Lords, Volume 860, debated on Tuesday 13 October 2026","url":"https://hansard.parliament.uk/lords/2026-10-13/debates/E93C0EEB-C2C2-4E81-9566-C10EB6EA4AF8/HealthBill","published_date":"2026-10-13","source_type":"Primary official record (UK Parliament, Lords Hansard)","primary":true},{"title":"In a first, AI helps UK lawmaker find her voice in parliament","url":"https://today.rtl.lu/news/world/in-a-first-ai-helps-uk-lawmaker-find-her-voice-in-parliament-270802641","published_date":"2026-10-13","source_type":"Secondary independent news coverage (AFP, via RTL Today)"},{"title":"Disabled peer uses AI tech to recreate the voice she used to have, as she delivers speech to House of Lords","url":"https://www.disabilitynewsservice.com/disabled-peer-uses-ai-tech-to-recreate-the-voice-she-used-to-have-as-she-delivers-speech-to-house-of-lords/","published_date":"2026-10-15","source_type":"Secondary independent news coverage (Disability News Service)"},{"title":"Historic First for AI: Baroness Campbell to make history with first AI-generated voice speech in the House of Lords","url":"https://www.independentliving.co.uk/industry-news/historic-first-for-ai/","published_date":"2026-10-11","source_type":"Secondary coverage with quotes from Baroness Campbell and the Lord Speaker (Independent Living)"}],"dates":{"edition_date":"2026-10-19","reviewed_at":"2026-10-19T12:03:55Z","first_curated_at":"2026-10-19T12:03:56.256Z","modified_at":"2026-10-22T03:44:55.173Z"},"image":{"url":"https://hooraymachine.com/images/cf2f5adc0968cd1b8ec2c39276cec5751cd1b2ddd61fe7d3310d6622b9a897b7.webp","alt":"Clay stop-motion scene: in a cozy clay version of the House of Lords with rows of orange-red benches and forest-green wood panels, a calm clay woman in a forest-green power wheelchair sits at the end of a bench with her blank cream speech pages on her lap, while a small cream clay loudspeaker beside her, the only smiling face in the scene, sends soft yellow ribbon-shaped sound waves across the chamber; a little orange clay reel of old recording tape sits next to the speaker, and two clay peers on the bench opposite listen calmly; conceptual illustration.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"delight-birdwatcher-notes-ai-wildlife-encounters","topics":[{"id":"everyday","name":"Delight"},{"id":"climate-wildlife","name":"Climate & Wildlife"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Birdwatchers who log sightings on eBird often add a quick note. “Mixed with bohemians,” one wrote, meaning a flock that included Bohemian Waxwings. Notes like that record which birds flock or feed together, or gang up on a predator. Ecologists struggle to collect that kind of detail because nobody can watch one patch of ground around the clock.","sourceUrls":["https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full","https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/","https://wildlife.org/ai-mines-comments-in-ebird-inaturalist-for-ecological-info/"]},{"kind":"context","text":"A University of Michigan team pointed AI at those notes. A GPT-4o model, tuned on 544 hand-sorted comments from Texas, screened more than 18 million Michigan sightings logged from 2019 to early 2024. Of the 632,169 that came with comments, it flagged 29,345 that may describe birds interacting.","sourceUrls":["https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full","https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/"]},{"kind":"clever-bit","text":"The clever bit is the side-by-side test. Undergraduate researcher Olivia Stein hand-sorted 485 of the flagged notes by type of encounter, which took her about eight hours. Claude, given the same rulebook and no practice examples, matched her on 395 of them in minutes. When birds ganged up on a possible predator, it caught every case.","sourceUrls":["https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full","https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/"]},{"kind":"result","text":"Then the team tried a butterfly, the West Coast lady. Of 8,999 sightings in a global biodiversity database, 996 had comments in English or Spanish. The AI found 299 describing interactions, including 100 flower visits, and those visits named plants missing from a major database’s list for the butterfly. Many were not native to its range.","sourceUrls":["https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full","https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/"]},{"kind":"quote","text":"I was surprised, and delighted, by the wide range of natural history observations that members of the public noted.","speaker":"Brian Weeks, associate professor, University of Michigan School for Environment and Sustainability","sourceUrls":["https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/"]},{"kind":"hooray","text":"Hooray: quick notes from people who simply love watching birds and butterflies are becoming wildlife data scientists can use, and AI handles the first read in minutes.","sourceUrls":["https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full","https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/","https://wildlife.org/ai-mines-comments-in-ebird-inaturalist-for-ecological-info/"]},{"kind":"next","text":"People stay in the loop to fix names and catch misses. Lead author Heng-Xing Zou says there is “definitely a lot more information that could be recovered from these kinds of comments,” and the team’s code and data are public for others to try.","sourceUrls":["https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/","https://wildlife.org/ai-mines-comments-in-ebird-inaturalist-for-ecological-info/","https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full"]}],"shareBlurb":"Birdwatchers leave quick notes like “Mixed with bohemians.” University of Michigan researchers used AI to screen 632,169 eBird comments and flag 29,345 that may describe bird encounters, turning casual notes into wildlife data. People still check the work."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/delight-birdwatcher-notes-ai-wildlife-encounters","api_url":"https://hooraymachine.com/api/v1/stories/delight-birdwatcher-notes-ai-wildlife-encounters","title":"AI combed birdwatchers’ casual notes and flagged 29,345 that may describe wildlife encounters","summary":"Birdwatchers often add a quick note to a sighting, like “Mixed with bohemians.” University of Michigan researchers had AI screen 632,169 of those comments from Michigan eBird records and flag 29,345 that may describe birds flocking, feeding or chasing off predators together. Checked against a student’s hand sorting, a second AI sorted about four in five correctly, in minutes. People still check the work.","category":{"id":"everyday","name":"Delight"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed study published in PNAS on October 8, 2026 (doi:10.1073/pnas.2602244123) by University of Michigan researchers and colleagues. The PNAS paper is linked first; the free full text also linked is the authors’ bioRxiv version (v2, September 11, 2026). The team checked the AI against people: 485 hand-sorted Michigan eBird comments and 400 hand-labeled butterfly comments. University of Michigan news release (October 12); independent coverage: Earth.com (October 12) and The Wildlife Society (October 16).","caveat":"This is a proof of concept covering birds in one US state and one butterfly species. The screening model missed about a quarter of the comments that did describe encounters, people had to fix nicknames like “bohemians,” and accuracy fell to about 62% on Chinese-language notes from Taipei. A butterfly on a flower does not prove pollination, and more comments about an encounter do not mean it happens more often."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.1073/pnas.2602244123"},"sources":[{"title":"Large language models unlock the ecology of species interactions","url":"https://www.pnas.org/doi/10.1073/pnas.2602244123","published_date":"2026-10-08","source_type":"Primary peer-reviewed research paper, version of record (PNAS 123(41) e2602244123)","primary":true},{"title":"Large language models unlock the ecology of species interactions","url":"https://www.biorxiv.org/content/10.64898/2026.02.06.704115v2.full","published_date":"2026-09-11","source_type":"Primary research paper, authors’ full-text preprint (bioRxiv v2); peer-reviewed version published in PNAS on October 8, 2026"},{"title":"How nature lovers can aid ecology with an assist from AI","url":"https://news.umich.edu/how-nature-lovers-can-aid-ecology-with-an-assist-from-ai/","published_date":"2026-10-12","source_type":"Secondary institution news release (University of Michigan News)"},{"title":"AI reveals hidden wildlife relationships in millions of birdwatching records","url":"https://www.earth.com/science/ai-reveals-hidden-wildlife-relationships-in-millions-of-birdwatching-records/","published_date":"2026-10-12","source_type":"Secondary independent news coverage (Earth.com)"},{"title":"AI mines comments in eBird, iNaturalist for ecological info","url":"https://wildlife.org/ai-mines-comments-in-ebird-inaturalist-for-ecological-info/","published_date":"2026-10-16","source_type":"Secondary independent news coverage (The Wildlife Society)"}],"dates":{"edition_date":"2026-10-18","reviewed_at":"2026-10-18T12:04:05Z","first_curated_at":"2026-10-18T12:04:06.113Z","modified_at":"2026-10-22T03:44:54.203Z"},"image":{"url":"https://hooraymachine.com/images/cb97a5a34166207db12a5900666f7913ab8ade2f27d30d9bf5fd54da8d07427a.webp","alt":"Clay stop-motion scene: a birdwatcher in an orange knit hat, seen from behind on a green park bench, looks through binoculars while an open cream notebook with eyes and a smile stands on the bench, a small orange clay robot reads it through a yellow magnifying glass beside a row of tiles stamped with bird shapes, and three clay birds sit together on a branch; conceptual illustration.","credit":"Original AI-assisted clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"delight-hobbyist-ai-super-earth-candidate","topics":[{"id":"everyday","name":"Delight"},{"id":"science-nature","name":"Science & Nature"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Pavel Rabtsevich is a product manager in Spain. He has no formal astronomy training and no telescope. What he did have was free public data from TESS, NASA’s planet-hunting space telescope, and a few AI coding tools.","sourceUrls":["https://gizmodo.com/did-claude-help-an-amateur-astronomer-discover-a-new-planet-2000824393","https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/"]},{"kind":"context","text":"TESS finds planets by catching tiny, regular dips in a star’s light as a world crosses in front of it. Rabtsevich pointed AI agents, including Claude Code and Codex, at TESS light curves for 126,246 stars. One star, TIC 4206066, about 116 light-years away, kept dimming for about two hours every 3.18 days.","sourceUrls":["https://zenodo.org/records/22967456","https://gizmodo.com/did-claude-help-an-amateur-astronomer-discover-a-new-planet-2000824393","https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/"]},{"kind":"clever-bit","text":"The clever bit is the public bet. The models wrote the code, ran it and checked each other’s work, while he set the rules and made the calls. Then he froze his code and posted the exact times the star should dim during the next TESS look, before any of that data exists. Anyone can check him against it.","sourceUrls":["https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/","https://zenodo.org/records/23175179"]},{"kind":"result","text":"Going back through TESS data from 2018 and 2020, he found the same dip arriving on schedule: 23 transits in all. If it is a planet, it is probably rocky, about 1.4 times Earth’s width, on a three-day orbit hot enough to reach roughly 1,000°F. The TESS team approved his request to watch the star every two minutes from October 31 to November 26.","sourceUrls":["https://zenodo.org/records/22967456","https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/","https://tess.mit.edu/science/ddt/","https://zenodo.org/records/23175179"]},{"kind":"quote","text":"There’s every reason to expect that this candidate super-Earth will pass further tests and make it into the official census.","speaker":"Greg Laughlin, astronomer, Yale University","sourceUrls":["https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/"]},{"kind":"hooray","text":"Hooray: a hobbyist with AI helpers turned free telescope data into a planet candidate good enough to win a closer TESS watch, and he put his predictions on the record before the test.","sourceUrls":["https://tess.mit.edu/science/ddt/","https://zenodo.org/records/23175179","https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/"]},{"kind":"next","text":"If the dips show up on time and at the right depth between October 31 and November 26, the case gets much stronger. Even then, calling it a planet would still take sharper imaging and other measurements to rule out a hidden companion star. Rabtsevich says he will post the result either way.","sourceUrls":["https://zenodo.org/records/23175179","https://zenodo.org/records/22967456","https://gizmodo.com/did-claude-help-an-amateur-astronomer-discover-a-new-planet-2000824393"]}],"shareBlurb":"A product manager in Spain had AI agents search public NASA TESS data on 126,246 stars and found a possible super-Earth. TESS takes a closer look starting Oct. 31, and he posted his predicted dimming times first."},"they_wrote_back":true,"subject_reply":{"text":"I'm really grateful for all the attention [the find] has received. What I want most is for all of us to find things out together and get answers to as many questions as we can. And I feel lucky to live at a time when this is open to anyone.","speaker":"Pavel Rabtsevich","role":"independent researcher in Spain","channel":"email","received_date":"2026-10-17","attribution":"— Pavel Rabtsevich, independent researcher in Spain, via email to Hooray Machine","source":"private_correspondence"},"url":"https://hooraymachine.com/story/delight-hobbyist-ai-super-earth-candidate","api_url":"https://hooraymachine.com/api/v1/stories/delight-hobbyist-ai-super-earth-candidate","title":"A hobbyist and his AI helpers spotted a possible super-Earth. NASA’s planet hunter takes a closer look Oct. 31","summary":"Pavel Rabtsevich, a product manager in Spain with no astronomy training, had AI coding agents search public TESS data on 126,246 stars. One star dims for about two hours every 3.18 days, a hint of a planet about 1.4 times Earth’s width. The TESS team approved a closer two-minute watch starting October 31, and he posted his predicted dimming times first. It is still a candidate.","category":{"id":"everyday","name":"Delight"},"evidence":{"label":"Testing","meaning":"Early research, prototypes or preprints. Encouraging questions, not established benefits.","details":"Unreviewed research note posted to Zenodo on September 25, 2026 by independent researcher Pavel Rabtsevich (doi:10.5281/zenodo.22967456), plus his pre-registered Sector 110 test protocol (v1.1, October 6, 2026; doi:10.5281/zenodo.23175179). The TESS Science Office at MIT lists his Director’s Discretionary Time program #100 (two-minute cadence, active). Independent coverage: Gizmodo (October 9) and Scientific American (October 14), with comment from Yale astronomer Greg Laughlin and MIT TESS team member Pablo Mercader Perez.","caveat":"It is still a candidate; nobody has confirmed a planet yet. The main alternative is a planet circling an unseen companion star, and the note’s own false-positive estimate falls short of formal validation. Some 2018 dips sat near thruster firings. TESS’s own tools had also flagged the signal, and an MIT team member says it would likely have reached the official list within months. A second, 11.13-day signal is tentative."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.5281/zenodo.22967456"},"sources":[{"title":"Two Transit-like Signals in TESS Photometry of the Nearby K Dwarf TIC 4206066","url":"https://zenodo.org/records/22967456","published_date":"2026-09-25","source_type":"Primary research note preprint (Zenodo; prepared for Research Notes of the AAS; not yet peer reviewed)","primary":true},{"title":"A Pre-registered TESS Sector 110 Test of Two Transit-like Signals in TIC 4206066","url":"https://zenodo.org/records/23175179","published_date":"2026-10-06","source_type":"Primary pre-registration protocol (Zenodo, v1.1)"},{"title":"Director’s Discretionary Time programs (DDT 100: TIC 4206066)","url":"https://tess.mit.edu/science/ddt/","published_date":"2026-10-17","source_type":"Official program listing (TESS Science Office, MIT), checked October 17, 2026"},{"title":"Did Claude Help an Amateur Astronomer Discover a New Planet?","url":"https://gizmodo.com/did-claude-help-an-amateur-astronomer-discover-a-new-planet-2000824393","published_date":"2026-10-09","source_type":"Secondary independent news coverage (Gizmodo)"},{"title":"Amateur astronomer harnesses AI to find a possible hidden exoplanet","url":"https://www.scientificamerican.com/article/amateur-astronomer-harnesses-ai-to-find-a-possible-hidden-exoplanet/","published_date":"2026-10-14","source_type":"Secondary independent news coverage (Scientific American)"}],"dates":{"edition_date":"2026-10-17","reviewed_at":"2026-10-17T12:02:46Z","first_curated_at":"2026-10-17T12:02:46.982Z","modified_at":"2026-10-22T03:44:54.410Z"},"image":{"url":"https://hooraymachine.com/images/7134fe9a190755f2ad7905e75d7566da5dac20a0ea2077bf6e495921b1f04bbd.webp","alt":"Clay stop-motion scene: a calm clay hobbyist at a cream desk with a forest-green laptop showing a simple wavy line with one dip, two small clay robot helpers with round eyes beside him, and through a round window a small orange clay planet with eyes and a smile passing in front of a big yellow star; conceptual illustration.","credit":"Original AI-assisted clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"health-penn-epilepsy-patient-ai-chat","topics":[{"id":"health","name":"Health"}],"article":{"version":1,"sections":[{"kind":"scene","text":"In an epilepsy monitoring unit, brain waves stream all day to machines that mostly talk to clinicians later. Patients living through those days usually get little of that stream in the moment—even though they are the ones who feel the seizures, the sleep, and the false alerts.","sourceUrls":["https://www.nih.gov/news-events/news-releases/patients-engage-epilepsy-monitor-real-time-improving-seizure-detection","https://www.nature.com/articles/s41591-026-04680-4"]},{"kind":"context","text":"Researchers at the University of Pennsylvania linked each patient’s live EEG to a HIPAA-compliant cloud and a smartphone chat. Thirteen adults in Penn’s epilepsy monitoring unit could ask about seizures, sleep, and trends, answer thrice-daily check-ins, and mark seizure alerts as true, false, or uncertain. The system answered most data questions accurately; investigators reviewed every message daily.","sourceUrls":["https://www.nature.com/articles/s41591-026-04680-4","https://www.pennmedicine.org/news/ai-platform-allows-patients-to-text-with-their-medical-device","https://www.medicaleconomics.com/view/patients-with-epilepsy-texted-questions-to-their-brain-monitors-in-penn-pilot-study"]},{"kind":"clever-bit","text":"The clever bit is the feedback loop. When the detector cried wolf, patients said so. Those labels later retrained the model every few hours in a simulation that mirrored a live rollout—without a clinician in the loop.","sourceUrls":["https://www.nature.com/articles/s41591-026-04680-4","https://www.nih.gov/news-events/news-releases/patients-engage-epilepsy-monitor-real-time-improving-seizure-detection"]},{"kind":"result","text":"Patients and the system exchanged 1,307 messages. Of 58 real-time seizure alerts, patients answered about two-thirds. In the simulation that reused those answers, the median false-alarm rate fell about 80%—from roughly 0.74 an hour to 0.10—while sensitivity held. Five of thirteen asked, unprompted, to take the system home.","sourceUrls":["https://www.nature.com/articles/s41591-026-04680-4","https://www.nih.gov/news-events/news-releases/patients-engage-epilepsy-monitor-real-time-improving-seizure-detection","https://www.medicaleconomics.com/view/patients-with-epilepsy-texted-questions-to-their-brain-monitors-in-penn-pilot-study"]},{"kind":"quote","text":"More importantly, both the patient and the device can learn together over time. The device learns what matters to the patient, and the patient learns how their behavior affects their health.","speaker":"Brian Litt, MD, professor of Neurology, Neurosurgery, and Bioengineering, University of Pennsylvania","sourceUrls":["https://www.pennmedicine.org/news/ai-platform-allows-patients-to-text-with-their-medical-device"]},{"kind":"hooray","text":"Hooray: people in the ward taught the seizure detector to quiet down—about an 80% drop in false alarms in simulation—and several wanted the chat to come home with them.","sourceUrls":["https://www.nature.com/articles/s41591-026-04680-4","https://www.nih.gov/news-events/news-releases/patients-engage-epilepsy-monitor-real-time-improving-seizure-detection"]},{"kind":"next","text":"Penn and NIH frame this as a hospital proof of concept. Larger outpatient studies are planned, including work that would let patients help tune deep-brain stimulation for Parkinson’s. For now, families should keep their usual care plans; this app is not a discharged product.","sourceUrls":["https://www.pennmedicine.org/news/ai-platform-allows-patients-to-text-with-their-medical-device","https://www.nih.gov/news-events/news-releases/patients-engage-epilepsy-monitor-real-time-improving-seizure-detection"]}],"shareBlurb":"In a Penn epilepsy ward, patients texted their brain monitors and flagged false alarms. In simulation, those labels cut false alarms by about 80%."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/health-penn-epilepsy-patient-ai-chat","api_url":"https://hooraymachine.com/api/v1/stories/health-penn-epilepsy-patient-ai-chat","title":"Patients texted their seizure monitors, and in a simulation their feedback cut false alarms by about 80%","summary":"In a Penn Medicine epilepsy ward, 13 adults used a smartphone chat that could see their live EEG. They asked about seizures and sleep, flagged false alarms, and in a later simulation those labels cut the detector’s false-alarm rate by about 80%. Five people asked to take the app home. It is a hospital proof of concept, not outpatient care yet.","category":{"id":"health","name":"Health"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed prospective observational study · 13 adults in the Hospital of the University of Pennsylvania epilepsy monitoring unit. Goldblum, Shi, Litt and colleagues, Nature Medicine, published October 13, 2026 (doi:10.1038/s41591-026-04680-4). NIH and Penn Medicine releases same day. False-alarm reduction assessed in a pseudoprospective simulation using patient labels; live care used the baseline detector only.","caveat":"Only 13 inpatients, for a few days each, in a unit where medications are often lowered to provoke seizures. The ~80% false-alarm drop came from a simulation after the stay, not from a detector already running live on those labels. Sensitivity did not improve. About two to four false alarms a day may still remain. Outpatient benefit is unproven."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.1038/s41591-026-04680-4"},"sources":[{"title":"Human–artificial intelligence collaboration for comanaging epilepsy in real time","url":"https://www.nature.com/articles/s41591-026-04680-4","published_date":"2026-10-13","source_type":"Primary peer-reviewed research paper (Nature Medicine)","primary":true},{"title":"Patients engage with epilepsy monitor in real time, improving seizure detection","url":"https://www.nih.gov/news-events/news-releases/patients-engage-epilepsy-monitor-real-time-improving-seizure-detection","published_date":"2026-10-13","source_type":"Secondary funder news release (NIH / NINDS)"},{"title":"AI platform allows patients to text with their medical device","url":"https://www.pennmedicine.org/news/ai-platform-allows-patients-to-text-with-their-medical-device","published_date":"2026-10-13","source_type":"Secondary institution news release (Penn Medicine)"},{"title":"Patients with epilepsy texted questions to their brain monitors in Penn pilot study","url":"https://www.medicaleconomics.com/view/patients-with-epilepsy-texted-questions-to-their-brain-monitors-in-penn-pilot-study","published_date":"2026-10-14","source_type":"Secondary independent trade coverage (Medical Economics)"}],"dates":{"edition_date":"2026-10-16","reviewed_at":"2026-10-16T12:35:16Z","first_curated_at":"2026-10-16T12:24:59.366Z","modified_at":"2026-10-22T03:44:54.604Z"},"image":{"url":"https://hooraymachine.com/images/24c72c5b62fb8cd41184d45c9130a752180a82dc2c4f0bb376ba88efe8e264b2.webp","alt":"Clay stop-motion scene: a clay patient sitting up in a forest-green hospital bed wearing a soft cream headband and holding a tablet, a green monitor showing a gentle wavy line, and a smiling orange clay phone on a bedside table with a yellow speech bubble between them; conceptual illustration.","credit":"Original AI-assisted clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"health-tanzania-dr-screening-referrals","topics":[{"id":"global-health","name":"Global Health"},{"id":"health","name":"Health"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Diabetic retinopathy gives little warning. High blood sugar damages tiny blood vessels at the back of the eye, and by the time vision blurs, some sight can be gone for good. In northern Tanzania, people with diabetes can get their eyes photographed at their regular clinic. The hard part has been the next step: actually getting to the eye specialist.","sourceUrls":["https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care","https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(26)00224-X/fulltext"]},{"kind":"context","text":"In the usual setup, the photos went back to Kilimanjaro Christian Medical Centre for a doctor to grade, and results came by phone, typically two to four weeks later. Researchers from the London School of Hygiene & Tropical Medicine and KCMC, working with Tanzania's Ministry of Health, randomly assigned 2,378 adults at 24 hospitals to that pathway or an AI-supported one.","sourceUrls":["https://pmc.ncbi.nlm.nih.gov/articles/PMC10824006/","https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(26)00224-X/fulltext","https://streamlinefeed.co.ke/news/tanzania-trial-tests-ai-role-diabetic-eye-care"]},{"kind":"clever-bit","text":"The clever bit is timing. An AI grading program, picked partly because it works offline, read the retinal photos during the visit. So staff could sit down with each person before they left, explain the result, answer questions and hand over a referral on the spot.","sourceUrls":["https://pmc.ncbi.nlm.nih.gov/articles/PMC10824006/","https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care"]},{"kind":"result","text":"Among the 904 people whose photos showed sight-threatening disease, 157 of 437 in the AI group (35.9%) reached the eye clinic within eight weeks, compared with 85 of 467 (18.2%) in the usual pathway. People who went after AI screening got there in about 11 days on average, versus about 37.","sourceUrls":["https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(26)00224-X/fulltext","https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care"]},{"kind":"quote","text":"Using AI in a diabetic eye screening programme in Tanzania enabled patients to receive their result along with face-to-face counselling at the point of screening, giving people the opportunity to ask questions, understand treatment options and feel supported in their care.","speaker":"Dr Charles Cleland, ophthalmologist and clinical research fellow, International Centre for Eye Health, London School of Hygiene & Tropical Medicine","sourceUrls":["https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care"]},{"kind":"hooray","text":"Hooray: nearly twice as many people with sight-threatening diabetic eye disease made it to the specialist, and they got there weeks sooner.","sourceUrls":["https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(26)00224-X/fulltext","https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care"]},{"kind":"next","text":"Tanzania's Ministry of Health, a partner on the study, has committed to adding AI diagnostics to diabetic eye screening in several regions, with no timetable announced yet. The researchers say specialist clinics and treatment need to grow alongside screening. The open question is whether more clinic visits turn into more treatment and saved sight.","sourceUrls":["https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care","https://streamlinefeed.co.ke/news/tanzania-trial-tests-ai-role-diabetic-eye-care"]}],"shareBlurb":"In a Tanzania trial, people with diabetes who got an instant AI eye-screening result and a face-to-face talk were nearly twice as likely to reach the eye specialist: 35.9% vs 18.2% within eight weeks."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/health-tanzania-dr-screening-referrals","api_url":"https://hooraymachine.com/api/v1/stories/health-tanzania-dr-screening-referrals","title":"An on-the-spot AI eye check nearly doubled specialist visits for people with diabetes in Tanzania","summary":"At 24 hospitals in northern Tanzania, people with diabetes got their eye-screening result from AI before leaving the room, plus a face-to-face talk about what it meant. Among those who needed a specialist, 35.9% went within eight weeks, compared with 18.2% of people who heard by phone weeks later. Tanzania's health ministry plans to expand AI eye screening.","category":{"id":"health","name":"Health"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed randomized controlled trial · 2,378 adults with diabetes at 24 hospitals in Tanzania's Kilimanjaro and Arusha regions (March 2023–August 2024). Cleland, Burton and colleagues (LSHTM and Kilimanjaro Christian Medical Centre), The Lancet Global Health, published October 12, 2026; ISRCTN18317152. Primary outcome: specialist eye clinic attendance within 8 weeks among 904 people with true referable diabetic retinopathy, as judged by masked UK-certified graders.","caveat":"Clinic attendance within eight weeks is an early step, not proof that people got treatment or kept their sight; the trial was not powered to test those outcomes. The AI group also got an immediate in-person talk, so the trial can't separate the software from the faster conversation. About 64% of people in the AI group who needed a specialist still hadn't gone by eight weeks."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.1016/j.langlo.2026.104062"},"sources":[{"title":"Artificial intelligence-supported diabetic retinopathy screening in Tanzania: a randomised controlled trial","url":"https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(26)00224-X/fulltext","published_date":"2026-10-12","source_type":"Primary peer-reviewed research paper (The Lancet Global Health)","primary":true},{"title":"AI screening helps more people with diabetes access sight-saving eye care","url":"https://www.lshtm.ac.uk/newsevents/news/2026/ai-screening-helps-more-people-diabetes-access-sight-saving-eye-care","published_date":"2026-10-13","source_type":"Secondary institution news release (London School of Hygiene & Tropical Medicine)"},{"title":"Tanzania Trial Tests AI's Role in Diabetic Eye Care","url":"https://streamlinefeed.co.ke/news/tanzania-trial-tests-ai-role-diabetic-eye-care","published_date":"2026-10-14","source_type":"Secondary independent news analysis (Streamline Feed, Kenya)"},{"title":"Artificial intelligence-supported diabetic retinopathy screening in Tanzania: rationale and design of a randomised controlled trial","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10824006/","published_date":"2024-01-25","source_type":"Primary trial protocol (BMJ Open)"}],"dates":{"edition_date":"2026-10-15","reviewed_at":"2026-10-15T12:29:23Z","first_curated_at":"2026-10-15T11:55:32.427Z","modified_at":"2026-10-22T03:44:54.981Z"},"image":{"url":"https://hooraymachine.com/images/31cb76a6fb7eedbb431018dcf1d7287080b23ced9929e793faa38835587b5ca3.webp","alt":"Clay stop-motion scene: a forest-green clay retinal camera on a cream table, and an orange clay referral card with eyes and a smile standing on a yellow clay path that leads to a small clay eye clinic; conceptual illustration.","credit":"Original AI-assisted clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"safety-dane-ava-nonemergency","topics":[{"id":"safety-disaster","name":"Safety & Disaster Response"}],"article":{"version":1,"sections":[{"kind":"scene","text":"A noise complaint, a lost dog, a permit question—none of them need lights and sirens, but each one still ties up a Dane County dispatcher who also answers 911. Last year the Public Safety Communications Department took about 375,000 calls; more than 200,000 were non-emergency.","sourceUrls":["https://www.danecounty.gov/PressDetail/11940","https://captimes.com/news/government/dane-county-s-new-ai-chatbot-ava-is-answering-non-emergency-calls/article_ac257151-a740-4e80-8694-c92b3ebea756.html"]},{"kind":"context","text":"On August 18, 2026 the county put AVA—the Aurelian Virtual Assistant—on the non-emergency line at 608-255-2345. Callers talk to an AI voice that asks what is going on, gathers details for a service request, and works in more than 45 languages. A communicator still reviews the transcript afterward. If AVA hears an emergency, or the caller asks for a person, the call moves to a human.","sourceUrls":["https://www.danecounty.gov/PressDetail/11940","https://dane911.com/AVA"]},{"kind":"clever-bit","text":"The clever bit is the split: AVA takes the routine non-emergency queue first, while trained communicators stay ready for 911. It listens for emergency keywords and can hand off mid-call. The answers are saved for a trained call-taker to review, and they can follow up by phone or text if they need more.","sourceUrls":["https://dane911.com/AVA","https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls","https://captimes.com/news/government/dane-county-s-new-ai-chatbot-ava-is-answering-non-emergency-calls/article_ac257151-a740-4e80-8694-c92b3ebea756.html"]},{"kind":"result","text":"About a month in, Spectrum News reported that AVA had handled more than 24,000 non-emergency calls. Deputy Director Johnny Leonard said there had been no major mistakes, and that one strength was how quickly AVA spots emergencies arriving on the non-emergency line and passes them to a person. Staff take turns monitoring AVA so a human can call back if something looks off.","sourceUrls":["https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls"]},{"kind":"hooray","text":"Hooray: more than 24,000 non-emergency calls handled in roughly the first month—so 911 communicators can stay on the line that needs them most.","sourceUrls":["https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls"]},{"kind":"stat","value":"24,000","text":"non-emergency calls handled by the AI assistant in about its first month","sourceUrls":["https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls"]},{"kind":"next","text":"Eleven months of the trial remained when Spectrum News checked in. Callers who dislike the AI can ask for a communicator. Residents should still dial 911 for police, fire, or ambulance. The ~60 percent non-emergency relief figure quoted at launch was a goal from Director Luis Bixler, not a measured month-one result—so the honest scorecard so far is call volume handled and “no major mistakes,” not a finished outcome study.","sourceUrls":["https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls","https://captimes.com/news/government/dane-county-s-new-ai-chatbot-ava-is-answering-non-emergency-calls/article_ac257151-a740-4e80-8694-c92b3ebea756.html","https://dane911.com/AVA"]}],"shareBlurb":"Dane County's AVA assistant took 24,000+ non-emergency calls in its first month so 911 dispatchers stay free for emergencies."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/safety-dane-ava-nonemergency","api_url":"https://hooraymachine.com/api/v1/stories/safety-dane-ava-nonemergency","title":"Dane County's AI assistant took 24,000 non-emergency calls in its first month","summary":"Since mid-August 2026, Dane County's AVA voice assistant has answered the non-emergency public-safety line so human dispatchers can stay free for 911. Spectrum News reported more than 24,000 calls handled in about a month, with county officials saying there were no major mistakes. Callers can still ask for a person; emergencies still go to 911.","category":{"id":"public-life","name":"Public Life"},"evidence":{"label":"In use","meaning":"Being used outside a research-only setting. This describes deployment, not proof of effectiveness.","details":"Operational deployment · Dane County Public Safety Communications. Aurelian Virtual Assistant (AVA) launched Aug 18, 2026 on the county non-emergency line. Spectrum News 1 Wisconsin one-month update (Sept 28, 2026): >24,000 non-emergency calls handled; Deputy Director Johnny Leonard reported no major mistakes. Cap Times (Aug 19) and county press release corroborate launch design. Not a controlled outcome study.","caveat":"Early trial results from county officials and local news, not a peer-reviewed study. The ~60% non-emergency volume relief cited at launch was a design goal, not a measured one-month outcome. Vendor is Aurelian; humans still monitor AVA and review transcripts. Eleven months of the trial remained as of late September."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls"},"sources":[{"title":"Dane County shares one month update on AI non-emergency line","url":"https://spectrumnews1.com/wi/milwaukee/news/2026/09/28/dane-county-public-safety-ai-non-emergency-calls","published_date":"2026-09-28","source_type":"Primary independent local news (Spectrum News 1 Wisconsin)","primary":true},{"title":"Dane County's new AI chatbot 'Ava' is answering non-emergency calls","url":"https://captimes.com/news/government/dane-county-s-new-ai-chatbot-ava-is-answering-non-emergency-calls/article_ac257151-a740-4e80-8694-c92b3ebea756.html","published_date":"2026-08-19","source_type":"Secondary independent local news (Cap Times)"},{"title":"Dane County Launches Virtual Assistant for Non-Emergency Calls","url":"https://www.danecounty.gov/PressDetail/11940","published_date":"2026-08-18","source_type":"Secondary government press release"},{"title":"AVA | Dane County 911","url":"https://dane911.com/AVA","published_date":"2026-08-18","source_type":"Secondary official service page"}],"dates":{"edition_date":"2026-10-14","reviewed_at":"2026-10-14T11:51:39Z","first_curated_at":"2026-10-14T11:51:39.782Z","modified_at":"2026-10-22T03:44:55.767Z"},"image":{"url":"https://hooraymachine.com/images/ab82d104ffb1ee8d9b850cba6e0fe90d79011288da6ea5ffb1ff1931718b40e8.webp","alt":"Clay stop-motion scene: an orange clay assistant with eyes and a smile wears a forest-green headset beside a green dispatch console with yellow buttons; a calm clay hand rests nearby; conceptual illustration.","credit":"Original AI-assisted clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"health-save-o2-oxygen-titration","topics":[{"id":"health","name":"Health"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Hospital oxygen is usually turned up or down by hand, every few hours. Between those checks, patients can drift too high or too low. The SAVE-O2 AI trial asked whether a closed-loop dial that watches a pulse oximeter and nudges the flow on its own could keep people in the middle more of the time.","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401"]},{"kind":"context","text":"Researchers enrolled 300 adults at four U.S. hospitals who were on 1 to 10 liters of oxygen for acute respiratory illness, trauma, burns, or surgery. Half received the O2matic PRO100 autonomous titrator for up to 72 hours; half stayed on usual manual titration. Both groups aimed for the same SpO2 target of 93 percent, with an acceptable range of 90 to 96 percent.","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401","https://clinicaltrials.gov/study/NCT06374225"]},{"kind":"clever-bit","text":"The clever bit is a trend-based controller: the device reads SpO2 from its own oximeter and continuously adjusts oxygen flow between 0 and 15 liters per minute to stay inside the target band, using the lowest flow that works. Clinicians still set the target; the dial does the frequent nudging between bedside checks.","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401","https://bmjopen.bmj.com/content/16/1/e110739"]},{"kind":"result","text":"Patients with the automatic dial spent a mean 85 percent of observed time in the target range, versus 63 percent with usual care—an adjusted difference of 21 percentage points. Time in hypoxemia (SpO2 under 88 percent) fell from 3.6 percent to 2.0 percent. Time above 96 percent also dropped sharply.","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401"]},{"kind":"hooray","text":"Hooray: 85 percent of the time in the oxygen sweet spot, up from 63 percent with hand titration—in a 300-person multicenter randomized trial.","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401"]},{"kind":"stat","value":"85%","text":"of the time in the target oxygen range, up from 63% with usual care","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401"]},{"kind":"next","text":"The trial was unblinded, used one device model, and did not find differences in mortality or hospital-free days. The PRO100 was not FDA-cleared during the study and ran under an investigational device exemption. Pulse oximeters have known limits; the controller tracks the meter, not true arterial oxygen. What comes next is regulatory review and tests of other devices in everyday wards.","sourceUrls":["https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401","https://bmjopen.bmj.com/content/16/1/e110739"]}],"shareBlurb":"In a 300-person U.S. trial, an automatic oxygen dial kept patients in range 85% of the time—up from 63% with hand titration."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/health-save-o2-oxygen-titration","api_url":"https://hooraymachine.com/api/v1/stories/health-save-o2-oxygen-titration","title":"An automatic oxygen dial keeps patients in range 85% of the time","summary":"In a 300-person U.S. hospital trial, an automatic oxygen dial kept patients in the target oxygen range 85% of the time, versus 63% with usual nurse and therapist adjustments. Time spent too low on oxygen also fell. The device was studied under an investigational exemption and is not yet cleared for routine U.S. use; the trial did not show fewer deaths or shorter stays.","category":{"id":"health","name":"Health"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed multicenter randomized clinical trial · 300 adults at 4 U.S. hospitals. Reported by Douin, Ginde and colleagues, JAMA Internal Medicine, published online August 3, 2026. Primary outcome: proportion of time in SpO2 90–96%.","caveat":"Unblinded trial of one device (O2matic PRO100) under FDA investigational device exemption; not FDA-cleared for routine U.S. care during the study. No significant difference in mortality or hospital-free days. Benefits are time-in-range, not proven longer-term outcomes."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.1001/jamainternmed.2026.4023"},"sources":[{"title":"Autonomous Oxygen Titration for Maintaining Normoxemia in Acutely Ill Adults: The SAVE-O2 AI Randomized Clinical Trial","url":"https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2852401","published_date":"2026-08-03","source_type":"Primary peer-reviewed research paper","primary":true},{"title":"Statistical analysis plan for the SAVE-O2 AI trial: protocol","url":"https://bmjopen.bmj.com/content/16/1/e110739","published_date":"2026-01-01","source_type":"Primary trial protocol (BMJ Open)"},{"title":"Strategy to Avoid Excessive Oxygen Using an Autonomous Oxygen Titration Intervention (NCT06374225)","url":"https://clinicaltrials.gov/study/NCT06374225","published_date":"2026-01-09","source_type":"Secondary trial registry record"}],"dates":{"edition_date":"2026-10-13","reviewed_at":"2026-10-13T11:55:33Z","first_curated_at":"2026-10-13T11:55:33.611Z","modified_at":"2026-10-22T03:44:54.795Z"},"image":{"url":"https://hooraymachine.com/images/030be8b9edaf4953cb6289e1895084fd8a97d1443693565642d543def134d476.webp","alt":"Clay stop-motion scene: a calm clay patient asleep in a green hospital bed, linked by oxygen tubing to an orange clay dial with a sculpted smile, beside a gauge whose needle sits near a yellow target band; conceptual illustration.","credit":"Original AI-assisted clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"delight-instructmesh-printable-creations","topics":[{"id":"everyday","name":"Delight"}],"article":{"version":1,"sections":[{"kind":"scene","text":"Ask an AI model for a 3D-printable mug and you may get a lovely cup that can't hold your coffee. As MIT researchers put it, these models understand how an object should look, but not how it works. Fixing the result usually means wrestling with 3D design software, which is a steep climb for anyone who has never used it.","sourceUrls":["https://news.mit.edu/2026/instructmesh-tool-lets-users-repair-ai-3d-models-then-fabricate-them-1001"]},{"kind":"clever-bit","text":"InstructMesh, built by researchers at MIT CSAIL, Google and Northeastern University, lets people point at the part that's wrong and say what they want. A user highlights a region of the AI-made model, then asks for a change in plain words or nudges a slider. Behind the scenes, one AI system generates the shape and a language model turns each request into a fixed set of editing steps, which the user approves.","sourceUrls":["https://arxiv.org/abs/2608.28534","https://news.mit.edu/2026/instructmesh-tool-lets-users-repair-ai-3d-models-then-fabricate-them-1001"]},{"kind":"result","text":"In the team's first user study, 12 beginners worked through AI-generated designs with 156 flaws between them. An independent expert judged that 89.7% of those flaws were repaired. A second 12-person study compared ways of giving instructions, and the team printed playful examples to show the edited designs could leave the screen.","sourceUrls":["https://arxiv.org/abs/2608.28534"]},{"kind":"hooray","text":"People with no 3D modeling background were able to take a flawed AI design and steer it toward something they could actually print, using the words they already have. That's a friendlier on-ramp to making things, from household items to small accessories.","sourceUrls":["https://arxiv.org/abs/2608.28534","https://news.mit.edu/2026/instructmesh-tool-lets-users-repair-ai-3d-models-then-fabricate-them-1001"]},{"kind":"next","text":"The paper is set for the UIST 2026 conference and is available as an August author manuscript. People still have to notice the flaws themselves. The study didn't compare InstructMesh with other repair workflows, the printed objects were demonstrations rather than systematic tests, and hidden defects and moving parts were outside what the team evaluated.","sourceUrls":["https://arxiv.org/abs/2608.28534"]}],"shareBlurb":"MIT, Google and Northeastern researchers built InstructMesh so beginners can fix flawed AI-made 3D designs with words or sliders. In a 12-person study, an expert judged 89.7% of 156 flaws repaired."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/delight-instructmesh-printable-creations","api_url":"https://hooraymachine.com/api/v1/stories/delight-instructmesh-printable-creations","title":"Beginners fixed about 90% of flaws in AI-made 3D print designs with InstructMesh","summary":"In a UIST 2026 study, 12 beginners used InstructMesh to mark flaws in AI-generated 3D models and ask for repairs with words or sliders. An independent expert judged 89.7% of the 156 flaws fixed. People still have to spot the problems themselves, and the tool doesn't guarantee a safe, successful print.","category":{"id":"everyday","name":"Delight"},"evidence":{"label":"Testing","meaning":"Early research, prototypes or preprints. Encouraging questions, not established benefits.","details":"First user study: 12 novices, 156 annotated flaws, 89.7% repaired by expert assessment. A separate 12-person study compared interfaces. UIST 2026 paper available as an August preprint. AI role: TRELLIS generates geometry; GPT-4 maps instructions to fixed editing operations. Users select regions and approve changes.","caveat":"No comparison with alternative repair workflows; printed examples are demonstrations, not systematic validation. Hidden defects and moving mechanisms remain outside the evaluated scope."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.1145/3830398.3830647"},"sources":[{"title":"InstructMesh: Selective Refinement of Generative 3D Models for Fabrication","url":"https://arxiv.org/abs/2608.28534","published_date":"2026-08-28","source_type":"Primary research paper, UIST 2026 author manuscript"},{"title":"New tool lets users repair AI-generated 3D models, then fabricate them just the way they want","url":"https://news.mit.edu/2026/instructmesh-tool-lets-users-repair-ai-3d-models-then-fabricate-them-1001","published_date":"2026-10-01","source_type":"Research institution feature"}],"dates":{"edition_date":"2026-10-11","reviewed_at":"2026-10-11T17:00:47.555Z","first_curated_at":"2026-10-11T13:00:13.466Z","modified_at":"2026-10-27T21:01:51.448Z"},"image":{"url":"https://hooraymachine.com/images/4d0cb48dc085ab471b94aa8bd4f8770de4d9fd8aa90386fbb9d18bb5eb5cbf43.webp","alt":"Conceptual clay dragon-tail mug beside a laptop design and yellow modeling tool. A tiny smile is impressed into the tool handle.","credit":"Original AI-generated clay editorial illustration for Hooray Machine. Conceptual artwork, not a photograph or documentary evidence.","ai_generated":true}},{"id":"health-eye-care-history-interviews","topics":[{"id":"health","name":"Health"}],"article":{"version":1,"sections":[{"kind":"scene","text":"A good eye appointment starts with a good history: what's wrong, when it started, what makes it worse, which medicines the patient takes. In a busy clinic, that conversation often happens in a hurry, and details slip through that a doctor could have used.","sourceUrls":["https://www.nature.com/articles/s41746-026-03232-x"]},{"kind":"clever-bit","text":"Researchers at a Guangzhou eye hospital built an AI interview agent that interviews patients about their symptoms and background before their appointment and writes up a history for the doctor. They then ran a randomized trial comparing it with ophthalmology residents doing the same pre-visit interview. Half the patients were interviewed by the AI agent and half by a resident, chosen at random.","sourceUrls":["https://www.nature.com/articles/s41746-026-03232-x"]},{"kind":"result","text":"The trial enrolled 172 patients, 86 in each group, and assessors who didn't know which interviewer was used scored each history out of 100. The AI agent's histories scored 17.7 points higher on average. The AI conversations also ran longer, with a median of 11.2 minutes against 3.1 for residents.","sourceUrls":["https://www.nature.com/articles/s41746-026-03232-x"]},{"kind":"hooray","text":"Patients got a more complete hearing before they ever sat down with the doctor, and the doctor got a fuller picture to start from. The trial's assessors were masked, so they scored the histories without knowing who took them.","sourceUrls":["https://www.nature.com/articles/s41746-026-03232-x"]},{"kind":"next","text":"The history score was designed by the study team, and it's hard to separate the benefit of longer interviews from the AI's questioning style. The trial was at one hospital, in Chinese, in non-emergency care, and it didn't show better health outcomes or faster visits. The paper is peer-reviewed in npj Digital Medicine. The authors declared no competing interests, and the work was supported by public and university grants.","sourceUrls":["https://www.nature.com/articles/s41746-026-03232-x"]}],"shareBlurb":"In a 172-patient randomized trial at a Guangzhou eye hospital, an AI pre-visit interviewer took patient histories that scored 17.7 points higher out of 100 than ophthalmology residents'."},"they_wrote_back":false,"url":"https://hooraymachine.com/story/health-eye-care-history-interviews","api_url":"https://hooraymachine.com/api/v1/stories/health-eye-care-history-interviews","title":"AI pre-visit interviews took better eye-patient histories than residents in a randomized trial","summary":"In a 172-patient randomized trial at a Guangzhou eye hospital, an AI interview agent scored 17.7 points higher out of 100 than ophthalmology residents on the quality of the patient histories it gathered before appointments. Its interviews took longer, 11.2 minutes against 3.1, and the study didn't show better health outcomes or faster visits.","category":{"id":"health","name":"Health"},"evidence":{"label":"Promising","meaning":"Results measured in a study or bounded trial. Limitations and wider testing still matter.","details":"Peer-reviewed single-center randomized clinical trial, 86 patients per arm, with masked outcome assessment. The adjusted patient-history score difference was 17.67 points out of 100. Median interview duration was 11.2 minutes for AI versus 3.1 for residents.","caveat":"The MedHistory score was created in-house. Longer interviews and questioning strategy cannot be disentangled. Single-site, Chinese-language, non-emergency setting. Authors declare no competing interests; public/university grant support. No improved health outcome or faster workflow was established."},"provenance":{"kind":"editorial_summary","original_research":false,"curated_by":"Hooray Machine","canonical_source_id":"doi:10.1038/s41746-026-03232-x"},"sources":[{"title":"Standardized pre-consultation by a large language model agent vs ophthalmology residents: a randomized clinical trial","url":"https://www.nature.com/articles/s41746-026-03232-x","published_date":"2026-10-05","source_type":"Primary peer-reviewed clinical trial"}],"dates":{"edition_date":"2026-10-08","reviewed_at":"2026-10-08T18:26:12.669225+00:00","first_curated_at":"2026-10-08T13:08:01.201Z","modified_at":"2026-10-27T21:01:52.114Z"},"image":{"url":"https://hooraymachine.com/images/eye-history.webp","alt":"Conceptual handmade clay clipboard, tablet and eye symbol, with one little sculpted smile on the clipboard clip. Not a diagnostic image.","credit":"Original AI-generated clay editorial illustration for Hooray Machine; conceptual, not documentary evidence.","ai_generated":true}}],"pagination":{"page":1,"limit":20,"total":27,"total_pages":2,"next":"https://hooraymachine.com/api/v1/stories?limit=20&page=2","previous":null},"links":{"self":"https://hooraymachine.com/api/v1/stories?limit=20&page=1","documentation":"https://hooraymachine.com/api/docs","schema":"https://hooraymachine.com/openapi.json","rss":"https://hooraymachine.com/feed.xml","json_feed":"https://hooraymachine.com/feed.json"}}