Pangram also notes that it has improved the model’s accuracy over the past few months, especially against edited images, low-resolution pictures, and pictures made by the most advanced AI models out there
Researchers in Japan have developed a compact smartphone-connected sensor that detects extremely low concentrations of toxic forever chemicals in water, potentially making contamination testing faster, cheaper, and more accessible.
A Cambridge-led study suggests comets rarely delivered useful quantities of life’s molecular ingredients to early Earth because impacts destroyed them and surviving material wouldn’t have stayed concentrated at the surface for long.
Dimensional advertised $20-an-hour robot dog walkers who might need to carry the machines home. Its software and Anthropic's research show how much human help useful AI-powered robots can still require.
Sleeptite and RMIT University have secured a US patent for flexible sensors built into mattress protectors, designed to monitor body position and help caregivers identify potential fall risks.
A cybersecurity startup managed to uncover weaknesses in Anthropic's internal systems using AI agents. The company says its core AI technology and customer conversations remained secure, but the findings raise bigger questions about how AI companies protect sensitive information.
AI-powered visual assistance can help blind users understand their surroundings, but a University of Maryland study finds that lengthy responses and limited user controls can make getting a simple answer unnecessarily complicated.
Researchers at the National University of Singapore have developed a garment with a touchscreen woven into its fibres. The flexible display can show images, play videos, respond to touch, and survive washing.
Researchers have developed Scentronic, an AI-powered system that detects and recreates smells remotely. It currently supports five aromas, with potential applications in online shopping and virtual reality.
Could a selfie reveal clues about a cancer patient’s survival risk? Researchers have developed an AI model that estimates biological age and predicts mortality risk, using facial photographs of a patient.
When it comes to taking pictures, however, it doesn't matter whether I use a voice command or a quick tap on the temple. The whole act makes my stomach churn.
Anthropic claims that these self-replicating AI swarms will need something with as little computing power as an average Internet-connected device, like a smart oven or an Android TV box.
Fold CEO Will Reeves explains why making Bitcoin boring beats attempts to use volatile crypto as everyday cash. One such scenario? Accumulating modest rewards on mortgages and daily spending.
Johns Hopkins engineers built rice-sized magnetic robots that deliver antibiotics straight to infected heart tissue, potentially replacing weeks of IV drugs or risky open-heart surgery.
Meta has just launched its latest frontier AI models in the Muse Wonder series, rubbing shoulders against the likes of Claude Opus 5.5, Google Gemini 4 Argon, and OpenAI's GPT-6 Astra.