Archive
Every reason for hope, month by month
16 stories since March 2026.
November 2026
5 storiesAI nudges helped North American homes skip a gigawatt of peak power this year
Utilities using Oracle's AI-driven Opower programs asked customers to ease off electricity at the busiest moments, with personalised tips and rebates for doing it. Across a record 73 events in 2026, the savings added up to 1 gigawatt of peak demand.
Read the storyFifty blind veterans in Georgia got AI glasses that tell them what's in front of them
The Georgia Center for the Visually Impaired gave about 50 blind veterans free AI-powered smart glasses, made by Meta, at an event marking White Cane Safety Day. The Atlanta-based center also trains people to use them.
Read the storyA Montreal app lets people with ALS talk to family in a voice like their own
Anne-Marie Poirier lost her ability to speak to ALS. Now she types a few keywords into Vocality, an app built by two young Montreal founders, and it turns them into full sentences spoken in a recreation of her voice.
Read the storyDeepMind's new AI predicts how tiny DNA changes affect specific cells
Google DeepMind released AlphaGenome 2, trained on more than 100 million individual cells from 68 tissues. It predicts how a single-letter change in DNA alters what happens inside particular cell types, such as heart muscle or immune cells.
Read the storyA regular smartwatch spotted more hidden high blood pressure than a check-up did
Google researchers built an AI that reads the pulse and motion sensors already in a smartwatch, with no blood pressure cuff needed. In a study of 196 people at five US sites, it flagged 65.8% of those with high blood pressure, compared with 55.3% for a standard reading at the doctor's office.
Read the storyOctober 2026
1 storyJuly 2026
1 storyJune 2026
3 storiesCanada wants rules for AI chatbots, and safety advocates are cautiously pleased
Canada's federal government has introduced legislation aimed at AI chatbots, the kind people now use for homework help, advice and customer service.
Read the storyIllinois could make big AI companies prove their safety plans to outside auditors
A bill in Illinois, SB 315, would be the first US law requiring independent audits of how frontier AI developers manage safety.
Read the storyA man who lost his voice to ALS is back at a full-time job, speaking through a brain implant
UC Davis researchers fitted a man with ALS with a brain implant whose AI turns his attempted speech into words with about 99% accuracy. He has used it for more than 3,800 hours since 2023 and now works full time as an environmental advocate.
Read the storyApril 2026
2 storiesAI spotted pancreatic cancer on scans that looked normal, over a year early
Mayo Clinic researchers ran their AI over nearly 2,000 routine CT scans, including older scans from people later diagnosed with pancreatic cancer that had looked clear at the time. It caught about 73% of those hidden cases, roughly 16 months before diagnosis on average.
Read the storyAn AI built at McMaster has already designed a brand-new antibiotic
Researchers at McMaster University in Canada built a generative AI model that assembles new drug candidates from roughly 150,000 chemical building blocks. In early tests, it has already designed a new antibiotic.
Read the storyMarch 2026
4 storiesAI read years of old news to learn how to forecast flash floods
Google researchers used AI to comb years of public reports and catalogue more than 2.6 million past floods. That record trained a model that can warn of urban flash floods up to 24 hours ahead, shown free on Google's Flood Hub.
Read the storyAI helped UK breast screening find about 10% more cancers
The UK's first full evaluation of AI in breast cancer screening found that adding it to the process raised cancer detection by 10.4%, while also cutting the workload on doctors.
Read the storyAn AI call filter has warned UK phone users about a billion scam calls
Virgin Media O2 says its AI system has now labelled more than a billion suspected scam and spam calls, about 70 million a month, and blocks known fraud numbers outright.
Read the storyAI models turn out to be bad at hiding what they're thinking
Some AI models work through problems step by step in writing before they answer. New research found they struggle to control what shows up in that written reasoning.
Read the story