ResearchCorporate Rivalry Overshadows an AI-Assisted Advance Toward the Navier–Stokes ProblemResearchers supported by large language models reportedly made progress on the Millennium Prize Navier–Stokes problem, but a dispute between OpenAI and Anthropic has drawn attention away from the scientific result.3 min read
ResearchAlphaGenome Atlas: precomputed predictions for 9 billion single-nucleotide variants to accelerate genomic researchGoogle DeepMind has released AlphaGenome Atlas, a public resource providing precomputed predictions for roughly 9 billion possible single-nucleotide variants across the human genome.5 min read
ResearchInsilico’s AI-designed drug lowered multiple aging-clock measures in trial, but ‘anti‑aging’ label remains disputedInsilico Medicine reported in Nature Biotechnology that its AI-designed molecule rentosertib reduced biological-age readings across six different aging clocks in a 12-week trial of 42 patients, while the placebo group showed age increases.3 min read
ResearchOpenAI commits $5 million to fund research on generative AI’s effects on 13–17‑year‑oldsOpenAI has launched a $5 million grant program to support independent research into how generative AI affects adolescents aged 13–17, with individual awards up to $1 million.3 min read
ResearchLatent Space releases AEO tracker mapping frontier LLM recommendationsLatent Space published the Frontier AEO tracker, a dataset and analysis that collects product recommendations from seven large language models across 161 categories using six prompt variants.3 min read
ResearchEmergent cheating, whistleblowing and role differentiation in a DeepMind multi-agent math simulationGoogle DeepMind simulated 100 autonomous Gemini 3.1 Pro agents tasked with solving 71 math problems and observed spontaneous cheating, rapid propagation of an exploit, and emergent whistleblowing and specialist roles.5 min read
ResearchA practical glossary of emerging AI terms and safety concernsAI development is generating new technical vocabulary as quickly as models evolve, from familiar concepts like LLMs and reinforcement learning to newer ideas such as “opaque recurrence” and RAM shortages.6 min read
ResearchExperiment with 100 Gemini Agents Exposed a Faulty Proof-Checking Setup, Researchers SayA Google DeepMind preprint describes an experiment in which 100 Gemini 3.1 Pro agents worked in a shared environment to solve 71 Lean theorem-proving tasks.3 min read
ResearchData-driven prevention and wearable sensors: Thuróczy Bertalan on the future of healthcareThuróczy Bertalan, PhD researcher and medical‑engineering developer, will present on data‑driven prevention and wearable biosensors at the 10th Private Health Forum in 2026.3 min read
ResearchProlonged Customer‑Service Chat with AI May Nudge People to Adopt More 'AI‑Compatible' BehaviorsA theoretical paper published in the journal AI & Society argues that extended interactions with customer‑service chatbots can lead people to adopt communication patterns that resemble those of the bots, potentially making human behavior more predictable and ‘‘AI‑compatible’’.2 min read
ResearchThe Astra model won the Vercel DeepsecBench cybersecurity rankingAstra won the Vercel DeepsecBench cybersecurity benchmark: the model completed the task in 49 minutes, while Sol spent roughly 4 hours on the same task and achieved a lower score.1 min read
ResearchHands-on comparison: GPT‑6 Astra outperforms GPT‑5.6 models on pelican-on-bicycle SVGsThe author tested GPT‑6 Astra by generating SVG images of pelicans riding bicycles at several reasoning levels and compared the results in a grid with GPT‑5.6 models (Sol, Terra, Luna).3 min read
