{"api":"discoveria/1","endpoint":"/api/fil.json","genere_le":"2026-10-18T17:58:45Z","source":{"depot":null,"commit":"f47441d0f03f4d0804cda52a2f84062a6b9a228f"},"licence":{"id":"CC-BY-4.0","url":"https://creativecommons.org/licenses/by/4.0/","attribution":"DiscoverIA"},"schema_donnees":2,"donnees":{"items":[{"id":"2026-10-12-muse-wonder-and-muse-wonder-full-meta","date":"2026-10-12","famille":"modeles-labos","categorie":"modele","acteur":"Meta","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Muse Wonder et Muse Wonder Full (Meta)","en":"Muse Wonder and Muse Wonder Full (Meta)"},"resume":{"fr":"Meta présente Muse Wonder et Muse Wonder Full, ses modèles les plus capables, accessibles via la Meta Model API et Muse Code pour Muse Wonder, et aux abonnés de Muse Code pour la version Full.","en":"Meta introduces Muse Wonder and Muse Wonder Full, its most capable models, available via the Meta Model API and Muse Code (Wonder) and to Muse Code subscribers (Full)."},"sources":[{"url":"https://about.fb.com/news/2026/10/introducing-muse-wonder-models/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://about.fb.com/news/2026/10/introducing-muse-wonder-models/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/muse-wonder-full","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-11-conditional-integer-multiplication-bound-0-00139","date":"2026-10-11","famille":"sciences","categorie":"decouverte","acteur":"un contributeur (contributeur de CrocSwap/integer-mult-bounds)","titre":{"fr":"Borne conditionnelle κ = 0,001391523459591829 pour la multiplication entière (PR #514 d'integer-mult-bounds)","en":"Conditional integer-multiplication bound κ = 0.001391523459591829 (integer-mult-bounds PR #514)"},"resume":{"fr":"un contributeur propose le 11 octobre 2026 (PR #514 d'integer-mult-bounds) une borne conditionnelle κ = 0,001391523459591829 pour la multiplication de grands entiers. Certificat rejouable ; demande ouverte, non fusionnée. Statut : non évalué.","en":"a contributor proposed on 11 October 2026 (integer-mult-bounds PR #514) a conditional bound κ = 0.001391523459591829 for multiplying large integers. The certificate can be replayed; the request is open, not merged. Status: not yet evaluated."},"sources":[{"url":"https://github.com/CrocSwap/integer-mult-bounds/pull/514","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://github.com/CrocSwap/integer-mult-bounds","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://github.com/openai/math/blob/main/CONTENTS.md","nature":"communaute","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":["imb-pr514-kappa-2026-10-11"],"redige_par":"pipeline-redaction/0.1 (« En une phrase » de la fiche)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":{"mode":"complet","relu":["fr","en"]},"exemple":false},{"id":"2026-10-09-theophyseval-50-unpublished-theoretical-physics-problems","date":"2026-10-09","famille":"sciences","categorie":"decouverte","acteur":"Suvrat Raju (ICTS), Praneeth Netrapalli et coauteurs","statut":"non-evalue","constate_le":"2026-10-15","titre":{"fr":"TheoPhysEval : cinq agents d'IA face à 50 problèmes inédits de physique théorique","en":"TheoPhysEval: five AI agents on 50 unpublished theoretical physics problems"},"resume":{"fr":"Des physiciens publient sur arXiv 50 problèmes de physique théorique inédits, à réponse numérique, et y testent cinq agents d'IA. Le meilleur, GPT-6-Astra avec Codex, donne la bonne réponse à 49 problèmes sur 50 avec accès au web, 46 sans. Préprint non relu par des pairs ; une seule exécution par agent, pas d'humains chronométrés.","en":"Physicists post on arXiv 50 unpublished theoretical physics problems with numerical answers and test five AI agents on them. The best, GPT-6-Astra with Codex, gets 49 of 50 problems right with web access, 46 without. Preprint not peer-reviewed; one run per agent, no timed humans."},"sources":[{"url":"https://arxiv.org/abs/2610.16733","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-15","titre":"TheoPhysEval: Assessing AI agents at autonomous theoretical physics problem solving","editeur":"arXiv"},{"url":"https://huggingface.co/datasets/RameshChandra1/TheoPhysEval","nature":"communaute","fiabilite":"PAGE","consulte_le":"2026-10-15","titre":"TheoPhysEval (jeu de données, accès sous conditions)","editeur":"Hugging Face"}],"avancees":[],"redige_par":"Claude Code (session 2bbfc3a9)","redige_le":"2026-10-15","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-10-08-ai-driven-formal-proof-search-in-mathematics-sci","date":"2026-10-08","famille":"sciences","categorie":"decouverte","acteur":"Google DeepMind","titre":{"fr":"Recherche de preuves formelles par IA en mathématiques (Science)","en":"AI-driven formal proof search in mathematics (Science)"},"resume":{"fr":"Science publie un agent de recherche de preuves formelles (Lean) qui a résolu neuf problèmes ouverts sur 353 de la liste Erdős et 44 conjectures sur 492 de l'OEIS, selon le résumé.","en":"Science publishes a formal-proof-search agent (Lean) that resolved nine of 353 open Erdos problems and 44 of 492 OEIS conjectures, per the abstract."},"sources":[{"url":"https://api.crossref.org/works/10.1126/science.aej2213","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://arxiv.org/abs/2605.22763","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":["alphaproof-nexus-science-2026-10-08"],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":{"mode":"complet","relu":["fr","en"]},"exemple":false},{"id":"2026-10-08-claude-dashboards-and-claude-motion-in-beta","date":"2026-10-08","famille":"apps-saas","categorie":"produit","acteur":"Anthropic","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Claude Dashboards et Claude Motion en bêta","en":"Claude Dashboards and Claude Motion in beta"},"resume":{"fr":"Claude peut construire un tableau de bord connecté à une base de données et des animations explicatives exportables en MP4.","en":"Claude can build a live dashboard connected to a data platform and explainer animations exportable as MP4."},"sources":[{"url":"https://claude.com/resources/articles/dashboards-and-motion","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-08-midjourney-test-of-a-thinking-mode-for-image-gen","date":"2026-10-08","famille":"apps-saas","categorie":"produit","acteur":"Midjourney","statut":"demo-seulement","constate_le":"2026-10-12","titre":{"fr":"Midjourney : test d'un « Thinking Mode » pour la génération d'images","en":"Midjourney: test of a \"Thinking Mode\" for image generation"},"resume":{"fr":"Midjourney teste sur son site Alpha un mode qui relance une image avec davantage de raisonnement.","en":"Midjourney is testing on its Alpha site a mode that reruns an image with additional reasoning."},"sources":[{"url":"https://updates.midjourney.com/testing-thinking-mode/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"demo-seulement","fiche":null,"exemple":false},{"id":"2026-10-07-chatgpt-intelligent-ui-answers-as-interactive-in","date":"2026-10-07","famille":"apps-saas","categorie":"produit","acteur":"OpenAI","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"ChatGPT : « Intelligent UI », réponses sous forme d'interfaces interactives","en":"ChatGPT: \"Intelligent UI\" answers as interactive interfaces"},"resume":{"fr":"Avec GPT-6, ChatGPT combine texte, visuels et éléments interactifs (comparatifs, calculatrices, jeux) et choisit le format automatiquement.","en":"With GPT-6, ChatGPT combines text, visuals and interactive elements (comparisons, calculators, games) and picks the format itself."},"sources":[{"url":"https://openai.com/index/gpt-6-for-everyone/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://help.openai.com/en/articles/6825453-chatgpt-release-notes","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-07-claude-haiku-5-5-anthropic","date":"2026-10-07","famille":"modeles-labos","categorie":"modele","acteur":"Anthropic","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Claude Haiku 5.5 (Anthropic)","en":"Claude Haiku 5.5 (Anthropic)"},"resume":{"fr":"Anthropic publie Claude Haiku 5.5, son petit modèle, avec un réglage d'effort ajustable pour la première fois dans la gamme Haiku et un coût moyen d'environ 75 % inférieur à Haiku 4.5.","en":"Anthropic releases Claude Haiku 5.5, its small model, the first Haiku with an adjustable effort setting and an average cost about 75% lower than Haiku 4.5."},"sources":[{"url":"https://www.anthropic.com/claude-haiku-5-5","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.anthropic.com/claude-haiku-5-5","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/claude-haiku-5-5","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-07-d1-3b-and-d1-omni-600m-open-weight-decision-mode","date":"2026-10-07","famille":"modeles-labos","categorie":"modele","acteur":"Liquid AI","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"d1-3B et d1-omni-600M, modèles de décision à poids ouverts (Liquid AI)","en":"d1-3B and d1-omni-600M, open-weight decision models (Liquid AI)"},"resume":{"fr":"Liquid AI publie deux modèles de décision ouverts qui répondent en une seule passe avant, sans générer de jetons, pensés pour le matériel embarqué NVIDIA.","en":"Liquid AI releases two open decision models that answer in a single forward pass without generating tokens, aimed at NVIDIA edge hardware."},"sources":[{"url":"https://www.liquid.ai/blog/d1-open","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.liquid.ai/blog/d1-open","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-07-three-manuscripts-withdrawn-after-a-sign-error-w","date":"2026-10-07","famille":"sciences","categorie":"decouverte","acteur":"OpenAI","titre":{"fr":"Trois manuscrits retirés (erreur de signe, classes de Weil) : Algebraicity of Weil classes on split abelian eightfolds ; Kuga-Satake K3 ; rational Hodge conjecture for products of K3","en":"Three manuscripts withdrawn after a sign error (Weil classes, Kuga–Satake K3, rational Hodge)"},"resume":{"fr":"Le 7 octobre 2026, OpenAI a retiré trois manuscrits de sa publication de mathématiques, sur les classes de Weil, Kuga-Satake K3 et la conjecture de Hodge rationnelle, après une erreur de signe. Statut : retiré.","en":"On 7 October 2026, OpenAI withdrew three manuscripts from its mathematics release, on Weil classes, Kuga-Satake K3 and the rational Hodge conjecture, after a sign error. Status: withdrawn."},"sources":[{"url":"https://raw.githubusercontent.com/openai/math/main/history.md","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-11"},{"url":"https://retractionwatch.com/2026/10/08/openai-withdraws-preprints-722-manuscripts-unsolved-math-problems/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":["oai-withdrawn-weil-classes-2026-10-07"],"redige_par":"pipeline-redaction/0.1 (« En une phrase » de la fiche)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"retire","fiche":{"mode":"complet","relu":[]},"exemple":false},{"id":"2026-10-06-cursor-remote-control-of-local-agents-from-the-i","date":"2026-10-06","famille":"apps-saas","categorie":"produit","acteur":"Cursor","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Cursor : contrôle à distance des agents locaux depuis l'app iOS","en":"Cursor: remote control of local agents from the iOS app"},"resume":{"fr":"L'app iOS de Cursor permet de voir et de répondre aux agents qui tournent sur l'ordinateur de l'utilisateur.","en":"Cursor's iOS app lets users see and reply to agents running on their own computer."},"sources":[{"url":"https://cursor.com/changelog","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-06-google-geospatial-models-applied-to-five-public","date":"2026-10-06","famille":"sciences","categorie":"decouverte","acteur":"Google Research, Google Earth AI et partenaires de santé publique","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"Modèles géospatiaux de Google appliqués à cinq cas de santé publique","en":"Google geospatial models applied to five public-health cases"},"resume":{"fr":"Google Research publie un préprint et un billet où des plongements de lieux améliorent des modèles épidémiologiques (vaccination ROR, dengue au Mexique, choléra en RDC, etc.).","en":"Google Research posts a preprint and blog where place embeddings improve epidemiological models (MMR vaccination, dengue in Mexico, cholera in DRC, etc.)."},"sources":[{"url":"https://arxiv.org/abs/2610.05699","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://research.google/blog/earth-ais-planetary-geospatial-foundation-models-for-global-public-health/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-10-06-mistral-large-4-public-preview-mistral-ai","date":"2026-10-06","famille":"modeles-labos","categorie":"modele","acteur":"Mistral AI","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Mistral Large 4, préversion publique (Mistral AI)","en":"Mistral Large 4 public preview (Mistral AI)"},"resume":{"fr":"Mistral ouvre une préversion publique de Mistral Large 4, modèle multimodal de 1 000 milliards de paramètres dont 52 milliards actifs, avec poids annoncés pour la fin du mois.","en":"Mistral opens a public preview of Mistral Large 4, a natively multimodal 1-trillion-parameter model with 52 billion active parameters, weights promised by end of month."},"sources":[{"url":"https://mistral.ai/news/mistral-large-4/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://mistral.ai/news/mistral-large-4/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/mistral-large-4","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-06-openai-math-719-manuscripts-in-372-families-from","date":"2026-10-06","famille":"sciences","categorie":"decouverte","acteur":"OpenAI (modèle interne non diffusé)","titre":{"fr":"openai/math : 719 manuscrits (722 au lancement), 372 familles, modèle interne non publié","en":"openai/math: 719 manuscripts in 372 families from an unreleased internal model"},"resume":{"fr":"Le 6 octobre 2026, OpenAI publie le dépôt openai/math : 719 manuscrits (722 au lancement) en 372 familles, couvrant 17 disciplines, issus d'un modèle interne non diffusé. Statut : en examen ; presque toutes ses sous-entrées sont non évaluées, une est contestée, une retirée.","en":"On 6 October 2026, OpenAI published the openai/math repository: 719 manuscripts (722 at launch) in 372 families across 17 disciplines, from an unreleased internal model. Status: under review; almost all its entries are not evaluated, one is contested, one withdrawn."},"sources":[{"url":"https://raw.githubusercontent.com/openai/math/main/README.md","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-11"}],"avancees":["oai-math-release-2026-10-06"],"redige_par":"pipeline-redaction/0.1 (« En une phrase » de la fiche)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"en-examen","fiche":{"mode":"complet","relu":["fr","en"]},"exemple":false},{"id":"2026-10-06-visionary-ai-predicting-pre-eclampsia-from-retin","date":"2026-10-06","famille":"sciences","categorie":"decouverte","acteur":"Équipe Visionary AI (article Nature Biotechnology)","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"Visionary AI : prédire la pré-éclampsie par imagerie rétinienne (Nature Biotechnology)","en":"Visionary AI: predicting pre-eclampsia from retinal imaging (Nature Biotechnology)"},"resume":{"fr":"Nature Biotechnology publie un modèle fondé sur des graphes vasculaires rétiniens qui prédit la pré-éclampsie avant l'apparition clinique, avec une validation externe sans réentraînement.","en":"Nature Biotechnology publishes a retinal vascular graph model that predicts pre-eclampsia before clinical onset, with external validation without retraining."},"sources":[{"url":"https://www.nature.com/articles/s41587-026-03303-0","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false},{"id":"2026-10-05-pie-virtual-cell-model-for-unseen-perturbations","date":"2026-10-05","famille":"sciences","categorie":"decouverte","acteur":"Arc Institute (Hani Goodarzi, Rishi Verma et al.)","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"PIE : modèle de cellule virtuelle pour des perturbations inédites","en":"PIE: virtual cell model for unseen perturbations"},"resume":{"fr":"L'Arc Institute présente PIE, qui prédit quels gènes changent après une perturbation et dont le AUPRC est de 1,2 à 3,2 fois celui du meilleur concurrent sur le jeu Replogle-Nadig.","en":"Arc Institute presents PIE, which predicts which genes change after a perturbation, with AUPRC 1.2 to 3.2 times that of the strongest competitor on the Replogle-Nadig dataset."},"sources":[{"url":"https://arcinstitute.org/news/pie","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://arcinstitute.org/news/pie","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-10-03-kolibri-1-open-weights-aleph-alpha","date":"2026-10-03","famille":"modeles-labos","categorie":"modele","acteur":"Aleph Alpha","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Kolibri 1, poids ouverts (Aleph Alpha)","en":"Kolibri 1, open weights (Aleph Alpha)"},"resume":{"fr":"Aleph Alpha publie Kolibri 1, modèle de raisonnement à experts (78 milliards de paramètres, 3,46 milliards actifs) en allemand et anglais, sous licence Apache 2.0.","en":"Aleph Alpha releases Kolibri 1, a mixture-of-experts reasoning model (78B parameters, 3.46B active) for German and English, under Apache 2.0."},"sources":[{"url":"https://huggingface.co/Aleph-Alpha/kolibri-1","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://huggingface.co/Aleph-Alpha/kolibri-1","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-01-bootloops-llm-driven-toolkit-for-exact-quantitat","date":"2026-10-01","famille":"sciences","categorie":"decouverte","acteur":"Matthew D. Schwartz (Harvard)","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"BootLoops : boîte à outils d'IA pour le calcul exact en sciences","en":"BootLoops: LLM-driven toolkit for exact quantitative science"},"resume":{"fr":"Matthew Schwartz (Harvard) publie sur arXiv une boîte à outils de calcul exact (intégrales de Feynman, preuves de complétude) utilisable par des agents d'IA, appliquée à la génomique, l'écologie et d'autres domaines.","en":"Matthew Schwartz (Harvard) posts on arXiv a toolkit for exact computation (Feynman integrals, completeness proofs) operable by AI agents, applied to genomics, ecology and other fields."},"sources":[{"url":"https://arxiv.org/abs/2610.02286","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.science.org/content/article/problems-22-scientific-fields-had-solutions-hiding-plain-sight-ai-has-found-them","nature":"presse","fiabilite":"SNIPPET","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-10-01-boston-dynamics-new-13-degree-of-freedom-hand-fo","date":"2026-10-01","famille":"objets-robots","categorie":"produit","acteur":"Boston Dynamics","statut":"demo-seulement","constate_le":"2026-10-12","titre":{"fr":"Boston Dynamics : nouvelle main à 13 degrés de liberté pour Atlas","en":"Boston Dynamics: new 13-degree-of-freedom hand for Atlas"},"resume":{"fr":"Boston Dynamics présente la nouvelle main d'Atlas : 13 degrés de liberté, actionnement direct, conçue pour la simulation haute fidélité et la fabrication en série (7 degrés pour la génération précédente).","en":"Boston Dynamics presents Atlas's new hand: 13 degrees of freedom, directly actuated, built for high-fidelity simulation and mass manufacturing (7 degrees on the previous generation)."},"sources":[{"url":"https://bostondynamics.com/blog/robot-hands-for-modern-ai-and-real-work/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.therobotreport.com/boston-dynamics-drops-pinkie-on-new-humanoid-hand/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"demo-seulement","fiche":null,"exemple":false},{"id":"2026-10-01-chatgpt-virtual-try-on-of-clothing-from-a-selfie","date":"2026-10-01","famille":"apps-saas","categorie":"produit","acteur":"OpenAI","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"ChatGPT : essayage virtuel de vêtements à partir d'un selfie","en":"ChatGPT: virtual try-on of clothing from a selfie"},"resume":{"fr":"Un bouton « Try on » sur les fiches produits de ChatGPT génère un essayage virtuel à partir d'un selfie, avec ChatGPT Images.","en":"A \"Try on\" button on ChatGPT product listings generates a virtual try-on from a selfie using ChatGPT Images."},"sources":[{"url":"https://help.openai.com/en/articles/6825453-chatgpt-release-notes","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-01-gemini-live-guided-vision-real-time-visual-assis","date":"2026-10-01","famille":"apps-saas","categorie":"produit","acteur":"Google","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Gemini Live : « Guided Vision », assistance visuelle en temps réel pour personnes aveugles ou malvoyantes","en":"Gemini Live: \"Guided Vision\", real-time visual assistance for blind and low-vision users"},"resume":{"fr":"Avec la caméra partagée, Gemini Live décrit l'environnement à voix haute et guide le cadrage par indications verbales.","en":"With the camera shared, Gemini Live describes surroundings aloud and guides framing with spoken cues."},"sources":[{"url":"https://blog.google/innovation-and-ai/products/gemini-app/guided-vision-gemini-live/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-01-github-copilot-computer-use-in-the-cli-and-app-i","date":"2026-10-01","famille":"apps-saas","categorie":"produit","acteur":"GitHub (Microsoft)","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"GitHub Copilot : utilisation d'ordinateur (computer use) dans la CLI et l'app, en aperçu public","en":"GitHub Copilot: computer use in the CLI and app, in public preview"},"resume":{"fr":"Copilot peut piloter des applications de bureau (cliquer, saisir, faire défiler) sur macOS et Windows, y compris des logiciels sans API.","en":"Copilot can operate desktop applications (click, type, scroll) on macOS and Windows, including software without an API."},"sources":[{"url":"https://github.blog/changelog/2026-10-01-github-copilot-can-now-interact-with-desktop-apps","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-10-01-suno-speech-beta-voice-and-music-generated-toget","date":"2026-10-01","famille":"apps-saas","categorie":"produit","acteur":"Suno","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Suno : « Speech » (bêta), voix et musique générées ensemble","en":"Suno: \"Speech\" (beta), voice and music generated together"},"resume":{"fr":"Suno ouvre à tous une bêta qui génère un texte parlé avec une musique originale en une seule piste.","en":"Suno opens a beta to everyone that generates spoken audio with original background music as a single track."},"sources":[{"url":"https://suno.com/blog/introducing-speech-beta","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-30-gemini-4-argon-google-deepmind","date":"2026-09-30","famille":"modeles-labos","categorie":"modele","acteur":"Google DeepMind","statut":"annonce","constate_le":"2026-10-12","titre":{"fr":"Gemini 4 Argon (Google DeepMind)","en":"Gemini 4 Argon (Google DeepMind)"},"resume":{"fr":"Google annonce Gemini 4 Argon, nouveau modèle de pointe, déployé d'abord auprès de défenseurs en cybersécurité de confiance, avec une limite de sortie portée à 1 million de jetons.","en":"Google announces Gemini 4 Argon, a new frontier model rolling out first to trusted cyber defenders, with output limit raised to 1 million tokens."},"sources":[{"url":"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/gemini-4-argon","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{"fr":"Antoine Olivier","en":"Antoine Olivier"},"relu_le":{"fr":"2026-10-13","en":"2026-10-13"},"statut_courant":"annonce","fiche":null,"exemple":false},{"id":"2026-09-30-gemini-reusable-skills-rolled-out-in-chat","date":"2026-09-30","famille":"apps-saas","categorie":"produit","acteur":"Google","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Gemini : des « skills » réutilisables déployés dans le chat","en":"Gemini: reusable \"skills\" rolled out in chat"},"resume":{"fr":"Gemini peut créer des skills à partir des conversations, les lancer automatiquement quand une consigne correspond, et en empiler plusieurs.","en":"Gemini can build skills from chats, run them automatically on matching prompts, and stack several at once."},"sources":[{"url":"https://gemini.google/release-notes/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-30-pretrained-model-for-neutrino-detectors-nature-m","date":"2026-09-30","famille":"sciences","categorie":"decouverte","acteur":"Équipe de physique des neutrinos (article Nature Machine Intelligence)","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"Modèle pré-entraîné pour détecteurs de neutrinos (Nature Machine Intelligence)","en":"Pretrained model for neutrino detectors (Nature Machine Intelligence)"},"resume":{"fr":"Nature Machine Intelligence publie un transformeur épars pré-entraîné qui, sur événements simulés du concept FASERCAL, égale avec environ 10^3 événements étiquetés l'entraînement classique avec dix fois plus de données.","en":"Nature Machine Intelligence publishes a pretrained sparse Vision Transformer that, on simulated FASERCAL events, matches scratch training with an order of magnitude more data using about 10^3 labelled events."},"sources":[{"url":"https://www.nature.com/articles/s42256-026-01309-6","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false},{"id":"2026-09-30-space-weather-risk-forecast-for-66-935-power-sub","date":"2026-09-30","famille":"sciences","categorie":"decouverte","acteur":"Microsoft Research (stagiaire Rohan Kannan)","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"Prévision du risque météo spatial pour 66 935 sous-stations électriques","en":"Space-weather risk forecast for 66,935 power substations"},"resume":{"fr":"Un billet de Microsoft Research décrit un système qui estime le risque météo spatial par sous-station américaine et détecte près de 80 % des événements majeurs de la période d'évaluation.","en":"A Microsoft Research post describes a system estimating space-weather risk per US substation and detecting nearly 80% of major events in the evaluation period."},"sources":[{"url":"https://www.microsoft.com/en-us/research/blog/forecasting-space-weather-risks-on-power-grids/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-09-30-synthid-bio-watermarking-ai-designed-proteins-na","date":"2026-09-30","famille":"sciences","categorie":"decouverte","acteur":"Google DeepMind","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"SynthID Bio : filigrane pour protéines conçues par IA (Nature)","en":"SynthID Bio: watermarking AI-designed proteins (Nature)"},"resume":{"fr":"Google DeepMind publie dans Nature des méthodes de filigrane pour séquences et structures de protéines générées par IA ; des liants filigranés testés en laboratoire égalent les non filigranés.","en":"Google DeepMind publishes in Nature watermarking methods for AI-generated protein sequences and structures; watermarked binders tested in the lab matched unwatermarked ones."},"sources":[{"url":"https://www.nature.com/articles/s41586-026-10965-y","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://deepmind.google/blog/introducing-synthid-bio/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false},{"id":"2026-09-29-gpt-6-1-sol-openai","date":"2026-09-29","famille":"modeles-labos","categorie":"modele","acteur":"OpenAI","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"GPT-6.1 Sol (OpenAI)","en":"GPT-6.1 Sol (OpenAI)"},"resume":{"fr":"OpenAI publie GPT-6.1 Sol, mise à niveau de GPT-6 Sol qui, selon elle, approche l'intelligence d'Astra en codage agentique, usage d'ordinateur et travail professionnel pour un cinquième du prix.","en":"OpenAI releases GPT-6.1 Sol, an upgrade of GPT-6 Sol that it says nearly matches Astra on agentic coding, computer use and professional work at one fifth of the price."},"sources":[{"url":"https://openai.com/index/introducing-gpt-6-1-sol","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://openai.com/index/introducing-gpt-6-1-sol","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/gpt-6-1-sol","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-29-openai-launches-dots-always-on-agents-in-chatgpt","date":"2026-09-29","famille":"apps-saas","categorie":"produit","acteur":"OpenAI","statut":"teste-par-un-tiers","constate_le":"2026-10-12","titre":{"fr":"OpenAI lance les « dots », agents permanents dans ChatGPT avec leur propre ordinateur cloud","en":"OpenAI launches \"dots\", always-on agents in ChatGPT with their own cloud computer"},"resume":{"fr":"ChatGPT propose des agents permanents qui poursuivent un travail entre les conversations, avec un ordinateur cloud dédié et des applications connectées.","en":"ChatGPT now offers always-on agents that keep working between conversations, each with its own cloud computer and connected apps."},"sources":[{"url":"https://help.openai.com/en/articles/6825453-chatgpt-release-notes","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.macstories.net/stories/openais-dots-a-first-look/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{"fr":"Antoine Olivier","en":"Antoine Olivier"},"relu_le":{"fr":"2026-10-13","en":"2026-10-13"},"statut_courant":"teste-par-un-tiers","fiche":null,"exemple":false},{"id":"2026-09-29-quine-ai-system-for-biology-tested-on-pancreatic","date":"2026-09-29","famille":"sciences","categorie":"decouverte","acteur":"Microsoft Research, Broad Institute","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"Quine : système d'IA pour la biologie testé sur un cancer du pancréas","en":"Quine: AI system for biology tested on pancreatic cancer"},"resume":{"fr":"Microsoft Research présente Quine, un modèle du monde de la biologie couplé à un harnais d'outils, et indique que les composés les mieux classés ont produit les plus grands changements d'état de cellules tumorales.","en":"Microsoft Research introduces Quine, a biology world model with a tool harness, and says its top-ranked compounds produced the largest shifts in tumour cell states."},"sources":[{"url":"https://www.microsoft.com/en-us/research/blog/introducing-quine-an-ai-research-system-designed-for-the-complexity-of-biology/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.microsoft.com/en-us/research/blog/introducing-quine-an-ai-research-system-designed-for-the-complexity-of-biology/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-09-28-agent-ai-designed-mrna-vaccines-stable-at-37-c-n","date":"2026-09-28","famille":"sciences","categorie":"decouverte","acteur":"Équipe AGENT (article Nature Biotechnology)","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"AGENT : vaccins ARNm stables à 37 °C conçus avec l'IA (Nature Biotechnology)","en":"AGENT: AI-designed mRNA vaccines stable at 37 °C (Nature Biotechnology)"},"resume":{"fr":"Nature Biotechnology publie un cadre d'optimisation bayésienne qui identifie en six itérations des formulations solides d'ARNm-LNP conservant leur bioactivité à 37 °C.","en":"Nature Biotechnology publishes a Bayesian-optimisation framework that found solid-state mRNA-LNP formulations in six iterations that kept bioactivity at 37 °C."},"sources":[{"url":"https://www.nature.com/articles/s41587-026-03331-w","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false},{"id":"2026-09-28-claude-sonnet-5-5-anthropic","date":"2026-09-28","famille":"modeles-labos","categorie":"modele","acteur":"Anthropic","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Claude Sonnet 5.5 (Anthropic)","en":"Claude Sonnet 5.5 (Anthropic)"},"resume":{"fr":"Anthropic publie Claude Sonnet 5.5, deuxième modèle de la famille 5.5, plus rapide de plus de 30 % que Sonnet 5, au même prix par token.","en":"Anthropic releases Claude Sonnet 5.5, second model of the 5.5 family, over 30% faster than Sonnet 5 at the same per-token price."},"sources":[{"url":"https://www.anthropic.com/claude-sonnet-5-5","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.anthropic.com/claude-sonnet-5-5","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/claude-sonnet-5-5","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-28-genfocal-fine-scale-weather-from-climate-project","date":"2026-09-28","famille":"sciences","categorie":"decouverte","acteur":"Équipe GenFocal (article Nature Machine Intelligence)","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"GenFocal : météo fine générée à partir de projections climatiques (Nature Machine Intelligence)","en":"GenFocal: fine-scale weather from climate projections (Nature Machine Intelligence)"},"resume":{"fr":"Nature Machine Intelligence publie GenFocal, un cadre génératif qui produit des canicules et cyclones à fine échelle à partir de projections climatiques grossières, sans paires d'entraînement alignées.","en":"Nature Machine Intelligence publishes GenFocal, a generative framework producing fine-scale heatwaves and cyclones from coarse climate projections without temporally aligned training pairs."},"sources":[{"url":"https://www.nature.com/articles/s42256-026-01308-7","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false},{"id":"2026-09-28-google-assistant-replaced-by-gemini-on-android-r","date":"2026-09-28","famille":"apps-saas","categorie":"produit","acteur":"Google","statut":"abandonne","constate_le":"2026-10-12","titre":{"fr":"Google Assistant remplacé par Gemini sur Android (retrait constaté)","en":"Google Assistant replaced by Gemini on Android (removal observed)"},"resume":{"fr":"Selon la presse spécialisée, Google Assistant a disparu des appareils Android vérifiés aux États-Unis et en Europe au profit de Gemini.","en":"Per specialist press, Google Assistant is gone from the Android devices checked in the US and Europe, replaced by Gemini."},"sources":[{"url":"https://9to5google.com/2026/09/28/google-assistant-gemini-android/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"abandonne","fiche":null,"exemple":false},{"id":"2026-09-28-grok-team-bots-shared-agents-that-learn-alongsid","date":"2026-09-28","famille":"apps-saas","categorie":"produit","acteur":"SpaceXSI (Grok)","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Grok : « Team Bots », agents partagés qui apprennent avec l'équipe","en":"Grok: \"Team Bots\", shared agents that learn alongside a team"},"resume":{"fr":"Un Team Bot reçoit fichiers, plugins et compétences, puis est partagé avec toute l'équipe qui travaille sur le même contexte.","en":"A Team Bot is given files, plugins and skills, then shared so a whole team works from the same context."},"sources":[{"url":"https://x.ai/news/team-bots","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-26-derived-invariance-of-the-signature-and-hodge-nu","date":"2026-09-26","famille":"sciences","categorie":"decouverte","acteur":"Ben Antieau et collègues (FRG) avec LLM","titre":{"fr":"Invariance dérivée de la signature et des nombres de Hodge des quadridimensionnelles (hexagon:2609.00104), texte marqué \"Primarily AI-generated\" par la plateforme","en":"Derived invariance of the signature and Hodge numbers of fourfolds"},"resume":{"fr":"Ben Antieau et des collègues, avec l'aide d'un modèle de langage, ont déposé le 26 septembre 2026 sur Hexagon un texte de géométrie algébrique. La plateforme l'a marqué « Primarily AI-generated ». Statut : non évalué.","en":"Ben Antieau and colleagues, working with a language model, posted an algebraic geometry text on Hexagon on 26 September 2026. The platform labelled it \"Primarily AI-generated\". Status: not yet evaluated."},"sources":[{"url":"https://hexagonmath.org/2609.00104v1","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-11"}],"avancees":["antieau-hodge-numbers-hexagon"],"redige_par":"pipeline-redaction/0.1 (« En une phrase » de la fiche)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":{"mode":"complet","relu":[]},"exemple":false},{"id":"2026-09-25-microsoft-new-copilot-app-organized-as-home-code","date":"2026-09-25","famille":"apps-saas","categorie":"produit","acteur":"Microsoft","statut":"annonce","constate_le":"2026-10-12","titre":{"fr":"Microsoft : nouvelle app Copilot organisée en Home, Code et Autopilot","en":"Microsoft: new Copilot app organized as Home, Code and Autopilot"},"resume":{"fr":"Microsoft annonce un Copilot unifié avec un agent persistant (Autopilot), un créateur d'apps en langage naturel (Code) et un point d'entrée Home.","en":"Microsoft announces a unified Copilot with a persistent agent (Autopilot), a natural-language app builder (Code) and a Home entry point."},"sources":[{"url":"https://blogs.microsoft.com/blog/2026/09/25/introducing-the-new-copilot-with-home-code-and-autopilot/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"annonce","fiche":null,"exemple":false},{"id":"2026-09-24-alphafold-database-predicted-protein-complexes-f","date":"2026-09-24","famille":"sciences","categorie":"decouverte","acteur":"EMBL-EBI, Google DeepMind, NVIDIA et partenaires académiques","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"Base AlphaFold : complexes protéiques prédits pour plus de 2 800 virus","en":"AlphaFold Database: predicted protein complexes for over 2,800 viruses"},"resume":{"fr":"L'EMBL-EBI, Google DeepMind et NVIDIA ajoutent à la base AlphaFold des complexes protéiques viraux prédits, via un portail de préparation aux pandémies.","en":"EMBL-EBI, Google DeepMind and NVIDIA add predicted viral protein complexes to the AlphaFold Database through a pandemic preparedness portal."},"sources":[{"url":"https://www.embl.org/news/?p=82309","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.nature.com/articles/d41586-026-03022-1","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-09-24-skydio-f10-lightrunner-fixed-wing-drone-deliveri","date":"2026-09-24","famille":"objets-robots","categorie":"produit","acteur":"Skydio","statut":"annonce","constate_le":"2026-10-12","titre":{"fr":"Skydio : drone à voilure fixe F10 Lightrunner, livraison prévue en 2027","en":"Skydio: F10 Lightrunner fixed-wing drone, deliveries planned for 2027"},"resume":{"fr":"Skydio annonce le F10 Lightrunner, son premier drone à voilure fixe, conçu pour un lancement en quelques secondes et jusqu'à 100 mph, avec un dock MegaDock.","en":"Skydio announces the F10 Lightrunner, its first fixed-wing drone, designed to launch in seconds and fly up to 100 mph, with a MegaDock."},"sources":[{"url":"https://www.skydio.com/blog/ascend-2026-f10-lightrunner-megadock-commands","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"annonce","fiche":null,"exemple":false},{"id":"2026-09-23-claude-identifies-an-enzyme-system-with-crispr-l","date":"2026-09-23","famille":"sciences","categorie":"decouverte","acteur":"Anthropic (groupe sciences de la vie)","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"Claude identifie un système enzymatique à répétitions de type CRISPR","en":"Claude identifies an enzyme system with CRISPR-like repeats"},"resume":{"fr":"Anthropic annonce que des agents Claude ont repéré, dans des données d'ADN, un système (ART) associant une transcriptase inverse, un gène partenaire et un réseau de répétitions, dont la fonction est inconnue.","en":"Anthropic reports that Claude agents spotted, in DNA data, a system (ART) pairing a reverse transcriptase, a partner gene and a repeat array, whose function is unknown."},"sources":[{"url":"https://www.anthropic.com/news/claude-discovers-novel-enzyme-system","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.anthropic.com/news/claude-discovers-novel-enzyme-system","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.nature.com/articles/d41586-026-03039-6","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.biorxiv.org/content/10.64898/2026.09.22.753630v1","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-09-23-cursor-rollouts-and-security-review-bots-for-dep","date":"2026-09-23","famille":"apps-saas","categorie":"produit","acteur":"Cursor","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Cursor : bots « Rollouts » et « Security Review » pour le suivi de déploiement et la revue de sécurité","en":"Cursor: \"Rollouts\" and \"Security Review\" bots for deploy monitoring and security review"},"resume":{"fr":"Rollouts surveille chaque changement jusqu'au déploiement et Security Review signale les bugs exploitables sur chaque pull request.","en":"Rollouts watches each change through deployment and Security Review flags exploitable bugs on every pull request."},"sources":[{"url":"https://cursor.com/changelog","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-23-fda-pronova-virtue-qa-cleared-via-510-k-with-a-p","date":"2026-09-23","famille":"objets-robots","categorie":"produit","acteur":"ProNova Solutions ; FDA","statut":"annonce","constate_le":"2026-10-12","titre":{"fr":"FDA : VIRTUE QA de ProNova autorisé en 510(k) avec plan de modification prédéfini","en":"FDA: ProNova VIRTUE QA cleared via 510(k) with a predetermined change control plan"},"resume":{"fr":"La FDA indique une décision « Substantially Equivalent » le 2026-09-23 pour VIRTUE QA (ProNova, K260597), assurance qualité de protonthérapie par IA, avec plan de contrôle des modifications autorisé.","en":"The FDA lists a Substantially Equivalent decision on 2026-09-23 for VIRTUE QA (ProNova, K260597), AI quality assurance for proton therapy, with an authorized change control plan."},"sources":[{"url":"https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K260597","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K260597","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://innolitics.com/articles/september-2026-ai-ml-fda-clearances/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"annonce","fiche":null,"exemple":false},{"id":"2026-09-23-meta-ray-ban-display-glasses-landmark-based-navi","date":"2026-09-23","famille":"objets-robots","categorie":"produit","acteur":"Meta","statut":"annonce","constate_le":"2026-10-12","titre":{"fr":"Meta : lunettes Ray-Ban Display, navigation visuelle et hologramme pour appels","en":"Meta: Ray-Ban Display glasses, landmark-based navigation and call hologram"},"resume":{"fr":"Meta annonce pour les lunettes Ray-Ban Display une navigation citant des repères visibles, un hologramme du visage piloté par la voix pour les appels WhatsApp, et une extension à cinq pays.","en":"Meta announces for Ray-Ban Display glasses navigation that cites visible landmarks, a voice-driven face hologram for WhatsApp calls, and expansion to five countries."},"sources":[{"url":"https://about.fb.com/news/2026/09/new-features-for-meta-ray-ban-display-navigation-hologram/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://about.fb.com/news/2026/09/new-features-for-meta-ray-ban-display-navigation-hologram/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"annonce","fiche":null,"exemple":false},{"id":"2026-09-23-meta-ray-ban-meta-audio-glasses-and-muse-agent-a","date":"2026-09-23","famille":"objets-robots","categorie":"produit","acteur":"Meta ; EssilorLuxottica","statut":"annonce","constate_le":"2026-10-12","titre":{"fr":"Meta : lunettes audio Ray-Ban Meta Audio et agent Muse annoncé pour les lunettes IA","en":"Meta: Ray-Ban Meta Audio glasses and Muse agent announced for AI glasses"},"resume":{"fr":"Meta présente des lunettes audio de 43 g avec assistant IA (précommande dès 349 USD, livraison le 13 octobre) et annonce l'arrivée de l'agent Muse sur ses lunettes IA dans les prochains mois.","en":"Meta presents 43 g audio glasses with an AI assistant (pre-order from USD 349, shipping October 13) and announces its Muse agent coming to its AI glasses in the coming months."},"sources":[{"url":"https://about.fb.com/news/2026/09/introducing-ray-ban-meta-audio-glasses-new-styles-plus-muse/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://about.fb.com/news/2026/09/the-biggest-news-from-connect-2026/","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://about.fb.com/news/2026/10/announcing-ranveer-singh-as-brand-ambassador-for-ray-ban-and-ray-ban-meta-in-india-along-with-exciting-new-updates-to-our-ai-glasses/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"annonce","fiche":null,"exemple":false},{"id":"2026-09-23-minerva-genome-language-model-to-find-non-coding","date":"2026-09-23","famille":"sciences","categorie":"decouverte","acteur":"Stanford University, Arc Institute, DOE JGI (David B. Li et al.)","statut":"non-evalue","constate_le":"2026-10-12","titre":{"fr":"Minerva : modèle de langage génomique pour trouver des ARN non codants","en":"Minerva: genome language model to find non-coding RNAs"},"resume":{"fr":"Un préprint Stanford/Arc présente Minerva, qui prédit des interactions à partir de séquences et décrit des systèmes à transcriptase inverse de prophages produisant des réseaux d'ARN non codants.","en":"A Stanford/Arc preprint introduces Minerva, which predicts interactions from sequence and describes prophage reverse transcriptase systems encoding arrays of non-coding RNAs."},"sources":[{"url":"https://www.biorxiv.org/content/10.64898/2026.09.22.753630v1","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.nature.com/articles/d41586-026-03039-6","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"non-evalue","fiche":null,"exemple":false},{"id":"2026-09-23-skild-ai-physical-self-play-a-humanoid-learns-so","date":"2026-09-23","famille":"objets-robots","categorie":"produit","acteur":"Skild AI","statut":"demo-seulement","constate_le":"2026-10-12","titre":{"fr":"Skild AI : auto-jeu physique, un humanoïde apprend le football en simulation puis en réel","en":"Skild AI: physical self-play, a humanoid learns soccer in simulation then on a robot"},"resume":{"fr":"Skild AI annonce un résultat de post-entraînement par auto-jeu : une politique apprise en simulation (Isaac Sim) est transférée sur un humanoïde pour un match de football.","en":"Skild AI announces a self-play post-training result: a policy learned in simulation (Isaac Sim) is transferred to a humanoid for a soccer match."},"sources":[{"url":"https://www.skild.ai/blogs/physical-self-play","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.skild.ai/blogs/physical-self-play","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"demo-seulement","fiche":null,"exemple":false},{"id":"2026-09-22-claude-opus-5-5-anthropic","date":"2026-09-22","famille":"modeles-labos","categorie":"modele","acteur":"Anthropic","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Claude Opus 5.5 (Anthropic)","en":"Claude Opus 5.5 (Anthropic)"},"resume":{"fr":"Anthropic présente Claude Opus 5.5, premier modèle de la famille Claude 5.5, qu'elle dit au niveau de Fable 5.1 sur la plupart des tâches pour un coût inférieur de 40 % à Opus 5.","en":"Anthropic introduces Claude Opus 5.5, first model of the Claude 5.5 family, which it says matches Fable 5.1 on most work at 40% lower cost than Opus 5."},"sources":[{"url":"https://www.anthropic.com/news/claude-opus-5-5","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.anthropic.com/news/claude-opus-5-5","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/claude-opus-5-5","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-22-eagle-ai-detection-of-oesophageal-cancer-on-ct-n","date":"2026-09-22","famille":"sciences","categorie":"decouverte","acteur":"Équipe EAGLE (article Nature Medicine)","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"EAGLE : détection du cancer de l'œsophage par IA sur scanner (Nature Medicine)","en":"EAGLE: AI detection of oesophageal cancer on CT (Nature Medicine)"},"resume":{"fr":"Nature Medicine publie un modèle d'IA de dépistage du cancer de l'œsophage sur scanner thoracique sans contraste, validé sur 80 612 patients dans 12 centres de trois pays.","en":"Nature Medicine publishes an AI model for oesophageal cancer screening on non-contrast chest CT, validated on 80,612 patients in 12 centres across three countries."},"sources":[{"url":"https://www.nature.com/articles/s41591-026-04656-4","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false},{"id":"2026-09-22-gpt-6-sol-and-gpt-6-luna-openai","date":"2026-09-22","famille":"modeles-labos","categorie":"modele","acteur":"OpenAI","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"GPT-6 Sol et GPT-6 Luna (OpenAI)","en":"GPT-6 Sol and GPT-6 Luna (OpenAI)"},"resume":{"fr":"OpenAI ajoute à la famille GPT-6 deux modèles plus rapides et moins chers, Sol et Luna, entraînés comme Astra, avec des prix API réduits de 50 % par rapport aux tarifs promotionnels de GPT-5.6.","en":"OpenAI adds two faster, cheaper models to the GPT-6 family, Sol and Luna, trained like Astra, with API prices cut 50% versus GPT-5.6 promotional pricing."},"sources":[{"url":"https://openai.com/index/introducing-gpt-6-sol-and-luna","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://openai.com/index/introducing-gpt-6-sol-and-luna","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://openai.com/index/gpt-6-astra/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/gpt-6-sol","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-21-grok-4-7-xai","date":"2026-09-21","famille":"modeles-labos","categorie":"modele","acteur":"xAI (page signée « SpaceXSI »)","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"Grok 4.7 (xAI)","en":"Grok 4.7 (xAI)"},"resume":{"fr":"xAI publie Grok 4.7, présenté comme son modèle le plus capable en codage et travail de connaissance, au même prix que Grok 4.6, avec un nouveau modèle de base plus grand.","en":"xAI releases Grok 4.7, described as its most capable model for coding and knowledge work, at the same price as Grok 4.6, on a new, larger base model."},"sources":[{"url":"https://x.ai/news/grok-4-7","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://x.ai/news/grok-4-7","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/grok-4-7","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-21-mimo-v2-6-pro-and-flash-open-weights-xiaomi","date":"2026-09-21","famille":"modeles-labos","categorie":"modele","acteur":"Xiaomi (MiMo)","statut":"disponible","constate_le":"2026-10-12","titre":{"fr":"MiMo-V2.6-Pro et Flash, poids ouverts (Xiaomi)","en":"MiMo-V2.6-Pro and Flash, open weights (Xiaomi)"},"resume":{"fr":"Xiaomi publie sur Hugging Face, sous licence MIT, MiMo-V2.6-Pro-RL, modèle omnimodal de 1,02 billion de paramètres dont 42 milliards actifs, avec 1 million de jetons de contexte.","en":"Xiaomi publishes on Hugging Face, under MIT licence, MiMo-V2.6-Pro-RL, an omnimodal model of 1.02T total and 42B active parameters with 1M-token context."},"sources":[{"url":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://artificialanalysis.ai/models/mimo-v2-6-pro","nature":"mesure-tierce","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"disponible","fiche":null,"exemple":false},{"id":"2026-09-21-retrochimera-chemical-retrosynthesis-model-natur","date":"2026-09-21","famille":"sciences","categorie":"decouverte","acteur":"Microsoft Research","statut":"relu-par-pairs","constate_le":"2026-10-12","titre":{"fr":"RetroChimera : modèle de rétrosynthèse chimique (Nature)","en":"RetroChimera: chemical retrosynthesis model (Nature)"},"resume":{"fr":"Microsoft Research publie dans Nature un modèle de rétrosynthèse qui combine deux modèles complémentaires ; des chimistes jugent environ 90 % de ses routes complètes acceptables.","en":"Microsoft Research publishes in Nature a retrosynthesis model that ensembles two complementary models; chemists accepted about 90% of its complete routes."},"sources":[{"url":"https://www.nature.com/articles/s41586-026-11160-9","nature":"primaire","fiabilite":"PAGE","consulte_le":"2026-10-12"},{"url":"https://www.microsoft.com/en-us/research/blog/improving-synthesis-prediction-of-small-molecules-at-scale-with-retrochimera/","nature":"presse","fiabilite":"PAGE","consulte_le":"2026-10-12"}],"avancees":[],"redige_par":"recherche-fil/0.1 + claude-opus-5-5 (4 chercheurs web, 12/10)","redige_le":"2026-10-13","relu_par":{},"relu_le":{},"statut_courant":"relu-par-pairs","fiche":null,"exemple":false}]}}