discovery · Source published 2026-06-15
A voice in everyday life.
Casey Harrell used a brain–computer interface at home to communicate and operate his computer without researchers present.
Find your way in.
86 records. Source reports remain distinct from discoveries.
discovery · Source published 2025-12-02
More time to decide when to plant.
India's Agriculture Ministry reported that a Kharif 2025 pilot sent local monsoon-onset forecasts by SMS to 38,845,214 farmers in 13 states, using a blend that included NeuralGCM, ECMWF AIFS, and 125 years of India Meteorological Department rainfall data.
discovery · Source published 2025-01-24
A field image that asks for less tilling.
A study combined image-based tillage classification, soil sensors, weather, slope, and crop inputs to recommend tilling intensity and fertilizer amount; authors report stationary prototypes deployed on 155 farms across five countries.
discovery · Source published 2025-11-21
Finding bacteria among the crumbs.
Researchers trained a deep-learning detector on microscopic bacterial microcolonies and visually similar chicken, spinach, and cheese debris, aiming to make food-safety screening faster and more reliable in messy samples.
discovery · Source published 2025-05-19
A teacher and an AI tutor after school.
A World Bank randomized trial studied a six-week, teacher-supported after-school English program in Edo State, Nigeria, where first-year senior secondary students used Microsoft Copilot powered by GPT-4.
discovery · Source published 2025-06-25
When the answer arrives too soon.
A preregistered randomized trial in a Turkish high school compared a standard GPT-4 chat interface, a teacher-informed GPT tutor with guardrails, and no generative-AI resource during mathematics study sessions.
discovery · Source published 2025-06-04
A walk that chooses what to describe.
VIPTour used the FocusFormer model and portable consumer devices to prioritize visual information for blind and low-vision visitors exploring unfamiliar environments, supporting exploration, recollection, and sharing.
discovery · Source published 2025-05-15
A screen reader that can take a follow-up question.
Google announced a rollout that let Android TalkBack users ask Gemini follow-up questions about images and about the whole screen, extending beyond one-shot image descriptions.
discovery · Source published 2025-05-14
Making room in a data center
Google DeepMind reported that an AlphaEvolve-discovered scheduling heuristic had been in production for more than a year and continuously recovered an average 0.7% of Google's worldwide compute resources.
discovery · Source published 2026-05-07
Finding more workable grid plans
Google DeepMind reported that AlphaEvolve improved a trained graph neural network's ability to find feasible solutions to the AC Optimal Power Flow problem from 14% to more than 88%.
discovery · Source published 2025-10-16
Rehearsing a fusion pulse before it happens
Google DeepMind and Commonwealth Fusion Systems announced a partnership in which CFS uses the open-source TORAX plasma simulator in daily workflows to test and refine SPARC operating plans before the tokamak is turned on.
discovery · Source published 2025-02-14
Personalizing an ablation for persistent AF
A multicenter randomized, controlled, double-blind superiority trial tested whether adding AI-detected spatio-temporal electrogram-dispersion targets to pulmonary-vein isolation improves treatment of drug-refractory persistent atrial fibrillation.
discovery · Source published 2026-01-28
Reading more of a DNA change's consequences
AlphaGenome is a unified DNA-sequence model that takes up to 1 Mb of DNA and predicts thousands of functional genomic tracks, including gene expression, splicing, chromatin features, transcription-factor binding, and contact maps.
discovery · Source published 2025-09-04
Quieting the mirrors that listen to space
Researchers developed Deep Loop Shaping, a reinforcement-learning feedback controller tested on the LIGO Livingston Observatory to reduce control noise in a difficult mirror-control loop.
discovery · Source published 2025-03-13
A model whose full recipe is visible.
Ai2 released OLMo 2 32B, a language model accompanied by publicly available data, code, weights, training details, and a reproducible training recipe.
discovery · Source published 2025-03-18
Reasoning models with more of the toolkit in view.
NVIDIA announced the Llama Nemotron Nano and Super reasoning models, alongside plans to make the post-training tools, datasets and optimization techniques openly available.
discovery · Source published 2025-08-05
Reasoning weights people can run themselves.
OpenAI released gpt-oss-120b and gpt-oss-20b as Apache-2.0 open-weight reasoning models, with downloadable weights and reference implementations for local deployment.
discovery · Source published 2025-04-29
Choosing when a model thinks longer.
The Qwen team released and described Qwen3, a family of dense and mixture-of-experts language models with selectable thinking and non-thinking modes.
discovery · Source published 2025-03-03
A warehouse inventory check that rolls itself.
GXO reported the successful completion of a pilot at its Oosterhout warehouse in the Netherlands, where a Dexory robot autonomously scanned inventory and produced real-time digital inventory models.
discovery · Source published date unknown
A warehouse robot that knows when to stop squeezing.
Amazon describes Vulcan, a fulfillment-center robot that combines vision with force-feedback sensors to pick and stow items in crowded inventory bins and hand difficult cases to a person.