ai 41
- 722 papers in one night: is OpenAI harming mathematics?
- GPT-6.1 Sol brings flagship reasoning at half the price
- OpenAI splits the difference with GPT-6 Sol and Luna
- Inside the refusal circuit of a large language model
- OpenAI releases GPT-6 Astra, its most capable model yet
- NVIDIA takes a $3.5 billion stake in MediaTek
- Two thousand humanoid robots hold their own Olympics
- Summer of agentic work: a labour market reckoning
- DeepSeek raises prices: the price war has a price after all
- AI that designed its own viruses
- Shanghai hosts the AI summit that created a global body
- OpenAI previews GPT-5.6 Sol, a model aimed at the hardest work
- WWDC makes Apple's devices the interface
- Google I/O bets that cheaper is the future
- DeepSeek V4 opens the million-token era at a discount
- Testing models at the frontier of human knowledge: Humanity's Last Exam
- GPT-5.5 makes agents the product
- Models are told to name their sources
- March was the month AI stopped being just models
- Two Sessions puts AI inside the work report
- Gemini 3.1 Ultra is Google's answer to the agent era
- Brussels issues its first fine under the AI Act
- First drug designed by an AI reaches the clinic
- GPT-5.4 arrives with a million-token memory
- Milano Cortina closes on a high-tech note
- Rare-disease AI and an in-brain gene editor
- An AI model joins the Olympic Games
- One year of the AI safety board: what the first rulings settled
- Diffusion models take on every molecule of life: AlphaFold 3
- The year the robots stood up: humanoid machines meet the factory floor
- Flow matching for robots: one backbone, every body
- One model, one router: what GPT-5's unification says about the frontier
- AI race enters the summer with a new phase
- DeepSeek R1: a reminder that intelligence is not the same as cost
- Mamba: linear-time language models without attention
- Apple lets the software carry the phone
- Apple Intelligence arrives at WWDC
- GPT-4o opens its voice to everyone
- M4 debuts in the thinnest iPad Pro yet
- Teaching a model to grade its own homework
- Preference learning without the reward model