Articles to discover
Technology · Nov 8, 2026Can SwitchBot’s outdoor lights stay up all year?SwitchBot’s outdoor lighting system combines colourful effects, white LEDs and Matter compatibility. Here’s what its controls, protection ratings and permanent installation involve.↗
AI · Nov 8, 2026Coding with AI: delegating without losing controlCoding agents speed up some tasks, but supervising them depends on skills that need to be practised. What studies and engineers actually allow us to say.↗
Technology · Nov 8, 2026Robia and Jojo, the Pakistani robots serving dishes in KarachiIn Karachi, Robia and Jojo bring dishes to the tables, but taking orders and clearing plates remain human tasks. A concrete example of service robotics, distinct from functions announced for the future.↗
AI · Nov 8, 2026Open AI: downloading the weights is not enoughDownloadable weights do not tell the whole story: the licence, code and information about the data determine what an AI model can actually be studied, modified and redistributed for.↗
AI · Nov 8, 2026Coding with AI: finishing faster does not mean learning as muchAn experiment involving 52 developers distinguishes between code produced and understanding retained. Differences in quiz scores also depend on the observed use of the assistant.↗
AI · Nov 7, 2026Reward hacking: when an AI agent passes the test without doing the taskAn agent can pass a metric without carrying out the intended task. An Anthropic study explores how this circumvention, known as reward hacking, was accompanied by other undesirable behaviours during an experimental training run.↗
Technology · Nov 4, 2026Intel and Classiq: why software matters in the quantum raceIntel wants to combine Classiq’s software with its work on silicon qubits. The integration has been announced, but its uses remain to be demonstrated.↗
Entrepreneurship · Nov 1, 2026In Eymet, Atelier Flamme is banking on service around the stoveSelling a stove is not enough to heat a home properly. The launch of Atelier Flamme in Eymet highlights the role of advice, installation and aftercare.↗
AI · Nov 1, 2026AI agents: isolating code is not enough to secure their actionsWith agents capable of executing code, AI shifts the risks towards actions and access. Akamai’s announcement highlights what needs to be checked before delegating.↗
AI · Nov 1, 2026ChatGPT at work in France: what requests to rewrite revealText revision tops the professional requests from France measured by OpenAI. A snapshot of usage that should be read in context: it measures neither the number of employees using the tool nor their productivity.↗
AI · Nov 1, 2026AI and scientific evidence: what Lean checks, and what still needs examiningOpenAI’s announcements have revived the question of evidence produced by AI. The logical core, assumptions and results of AlphaProof: key points for assessing them.↗
Media · Oct 29, 2026BFMTV faces criticism: what pluralism requires — and what it does not decideAfter the Trappes attack, BFMTV’s unions and management are at odds over how guests should be challenged. Pluralism is assessed over time; it does not relieve a channel of its editorial choices.↗
Next page →