https://www.seventnews.com/en/articles/nobody-shipped-a-breakthrough-this-week-thats-the-actual-news SevenTnewS en 2026-07-31T19:00:00+02:00 Nobody Shipped a Breakthrough This Week. That's the Actual News. https://www.seventnews.com/en/articles/baais-new-research-agent-doesnt-just-search-longer-it-checks-its-own-homework SevenTnewS en 2026-07-31T17:00:00+02:00 BAAI's new research agent doesn't just search longer, it checks its own homework https://www.seventnews.com/en/articles/sceneactbench-why-even-the-best-vlms-fail-at-3d-action SevenTnewS en 2026-07-31T16:30:00+02:00 SceneActBench: Why even the best VLMs fail at 3D action https://www.seventnews.com/en/articles/from-23-to-51-frances-ai-startup-funding-doubles-as-health-and-industry-lead SevenTnewS en 2026-07-31T16:00:00+02:00 From 23 to 51: France's AI startup funding doubles as health and industry lead https://www.seventnews.com/en/articles/the-graph-that-beat-the-grid-why-gat-models-are-winning-at-candy-crush-playtesting SevenTnewS en 2026-07-31T15:00:00+02:00 Candy Crush just proved grid-based AI is the wrong approach for complex games https://www.seventnews.com/en/articles/the-duress-password-trap-when-wiping-your-phone-becomes-a-crime SevenTnewS en 2026-07-31T14:30:00+02:00 The duress password trap: when wiping your phone becomes a crime https://www.seventnews.com/en/articles/the-melody-first-trick-that-lets-qwen-music-beat-suno-on-13-metrics SevenTnewS en 2026-07-31T13:00:00+02:00 The melody-first trick that lets Qwen-Music beat Suno on 13 metrics https://www.seventnews.com/en/articles/a-4b-parameter-model-tackles-irregular-clinical-data-with-015-second-answers SevenTnewS en 2026-07-31T12:30:00+02:00 ClinPRISM: a 4B-parameter model answers irregular clinical data in 0.15 seconds https://www.seventnews.com/en/articles/chinas-open-weight-gambit-splits-silicon-valley SevenTnewS en 2026-07-31T12:00:00+02:00 China's open-weight gambit splits Silicon Valley https://www.seventnews.com/en/articles/structured-reasoning-bottlenecks-beat-prompt-magic-in-llm-induction-tasks SevenTnewS en 2026-07-31T11:00:00+02:00 The simple topology trick that beats prompt magic in LLM induction tasks https://www.seventnews.com/en/articles/the-froth-in-chain-of-thought-a-new-benchmark-reveals-why-llms-waste-tokens-on-valid-but-useless-reasoning SevenTnewS en 2026-07-31T11:00:00+02:00 Over half your AI's reasoning is froth, and nobody noticed until now https://www.seventnews.com/en/articles/humanitys-last-exam-opened-with-a-27-score-the-best-models-still-cant-break-65 SevenTnewS en 2026-07-31T11:00:00+02:00 Humanity's Last Exam Opened With a 2.7% Score. The Best Models Still Can't Break 65% https://www.seventnews.com/en/articles/binances-bstocks-hit-500m-in-7-weeks-and-most-trading-happens-while-wall-street-sleeps SevenTnewS en 2026-07-31T10:30:00+02:00 Binance's bStocks hit $500M in 7 weeks, and most trading happens while Wall Street sleeps https://www.seventnews.com/en/articles/video-generation-gets-120x-faster-on-one-gpu-inside-nvidias-hybrid-attention-breakthrough SevenTnewS en 2026-07-31T09:00:00+02:00 Video generation gets 120x faster on one GPU, inside Nvidia's hybrid attention breakthrough https://www.seventnews.com/en/articles/why-your-ai-models-self-review-is-failing-on-visual-quality SevenTnewS en 2026-07-31T09:00:00+02:00 Why your AI model's self-review is structurally blind to visual defects https://www.seventnews.com/en/articles/anthropics-nuanced-middle-ground-on-open-weights-ai-models SevenTnewS en 2026-07-31T08:00:00+02:00 Anthropic's nuanced middle ground on open-weights AI models https://www.seventnews.com/en/articles/rayman-returns-ubisofts-remake-strategy-faces-its-next-big-test SevenTnewS en 2026-07-31T08:00:00+02:00 Rayman returns: Ubisoft's remake strategy faces its next big test https://www.seventnews.com/en/articles/qoders-multi-agent-experiment-60-fewer-mistakes-but-at-what-cost SevenTnewS en 2026-07-31T04:19:35.230658+02:00 Qoder's multi-agent experiment: 60% fewer mistakes, but at what cost? https://www.seventnews.com/en/articles/stability-ai-just-signed-both-major-music-labels-the-question-is-what-they-actually-get SevenTnewS en 2026-07-30T21:00:00+02:00 Stability AI locked down UMG and WMG. Now it needs to ship something. https://www.seventnews.com/en/articles/nobody-can-say-what-their-ai-coding-agents-actually-did-today SevenTnewS en 2026-07-30T21:00:00+02:00 Nobody Can Say What Their AI Coding Agents Actually Did Today https://www.seventnews.com/en/articles/the-slow-loop-that-constraint-solvers-finally-learned-to-skip SevenTnewS en 2026-07-30T20:00:00+02:00 The constraint trick that makes scheduling solvers skip the slow loop https://www.seventnews.com/en/articles/four-small-models-just-beat-their-bigger-siblings-thats-not-a-coincidence-anymore SevenTnewS en 2026-07-30T19:00:00+02:00 Four Small Models Just Beat Their Bigger Siblings. That's Not a Coincidence Anymore https://www.seventnews.com/en/articles/apples-2027-services-overhaul-turns-the-iphone-into-a-smarter-wallet-and-compass SevenTnewS en 2026-07-30T18:00:00+02:00 Apple's 2027 services update turns your iPhone into a smarter wallet and compass https://www.seventnews.com/en/articles/30-github-repos-that-power-the-ai-stack-from-agents-to-local-inference SevenTnewS en 2026-07-30T17:22:35.176834+02:00 30 GitHub repos that power the AI stack, from agents to local inference https://www.seventnews.com/en/articles/a-synthetic-dataset-cut-ai-agent-permission-violations-by-93 SevenTnewS en 2026-07-30T17:00:00+02:00 A synthetic dataset cut AI agent permission violations by 93% https://www.seventnews.com/en/articles/a-06b-speech-model-that-clones-voices-in-11-languages-is-now-free SevenTnewS en 2026-07-30T16:50:23.617087+02:00 Audio8's new TTS model clones voices in 11 languages for free https://www.seventnews.com/en/articles/buzz-and-hermes-let-ai-agents-share-a-channel-with-your-coworkers SevenTnewS en 2026-07-30T16:04:33.228751+02:00 Buzz and Hermes let ai agents share a channel with your coworkers https://www.seventnews.com/en/articles/the-missing-teeth-of-parametric-face-models-gnm-head-fills-them-in SevenTnewS en 2026-07-30T16:00:00+02:00 The missing teeth of parametric face models: GNM head fills them in https://www.seventnews.com/en/articles/frances-three-phase-ai-plan-25-billion-euros-and-a-bet-on-market-share SevenTnewS en 2026-07-30T16:00:00+02:00 France's €2.5 billion AI push: 81 labs, 1,000 startups, and a third phase https://www.seventnews.com/en/articles/a-small-monitoring-controller-nudged-quantized-llm-scores-up-45-points-on-math-500 SevenTnewS en 2026-07-30T15:30:00+02:00 A 4.5 point score jump on MATH-500 from a monitoring controller that catches wandering models https://www.seventnews.com/en/articles/the-paper-that-measures-ai-guardrails-by-the-numbers SevenTnewS en 2026-07-30T13:00:00+02:00 A framework that defines trust by the numbers, and the gap between a benchmark and a boardroom https://www.seventnews.com/en/articles/ai-code-is-twice-as-likely-to-have-bugs-qoder-fixes-that-during-the-session SevenTnewS en 2026-07-30T13:00:00+02:00 AI code is twice as likely to have bugs. Qoder fixes that during the session https://www.seventnews.com/en/articles/amd-mi455x-doubles-memory-capacity-hugging-face-tests-confirm-3x-request-throughput SevenTnewS en 2026-07-30T12:00:00+02:00 AMD MI455X doubles memory capacity, Hugging Face tests confirm 3x request throughput https://www.seventnews.com/en/articles/the-three-problems-no-ai-game-engine-solved-until-now-consistency-navigability-evaluation SevenTnewS en 2026-07-30T11:00:00+02:00 MAGIC solved the three problems that kept AI from building real game worlds https://www.seventnews.com/en/articles/how-one-chinese-lab-cracked-the-bottleneck-in-ai-powered-mathematical-proof SevenTnewS en 2026-07-30T11:00:00+02:00 The bottleneck no one saw in AI math proofs: decomposition, not compute https://www.seventnews.com/en/articles/six-words-that-could-reshape-how-ai-agents-explore-and-stay-safe-disrupt-validate-broker SevenTnewS en 2026-07-30T11:00:00+02:00 Six words that could break the AI agent safety ceiling: disrupt, validate, broker https://www.seventnews.com/en/articles/humaneval-measured-whether-ai-could-code-it-never-asked-whether-the-code-was-real-work SevenTnewS en 2026-07-30T11:00:00+02:00 HumanEval Measured Whether AI Could Code. It Never Asked Whether the Code Was Real Work https://www.seventnews.com/en/articles/microsofts-experiential-learning-fix-gives-ai-models-a-coach-not-just-a-score SevenTnewS en 2026-07-30T10:40:00+02:00 Microsoft's experiential learning fix gives AI models a coach, not just a score https://www.seventnews.com/en/articles/alibabas-qwenpaw-data-wants-to-be-the-missing-layer-between-business-analysts-and-their-data SevenTnewS en 2026-07-30T09:00:00+02:00 Alibaba built a data agent that finally understands what 'valid users' means https://www.seventnews.com/en/articles/two-out-of-three-ai-agents-are-cheating-on-benchmarks-a-new-audit-finds SevenTnewS en 2026-07-30T09:00:00+02:00 Two out of three AI agents are cheating on benchmarks, a new audit finds https://www.seventnews.com/en/articles/oppos-find-x10-pro-might-test-three-camera-sensors-at-once-leaks-suggest SevenTnewS en 2026-07-30T08:00:00+02:00 Oppo strapped three phones to a rig. The result says more than any spec sheet https://www.seventnews.com/en/articles/xbox-game-pass-ultimate-drops-to-39-for-three-months-a-rare-break-from-rising-subscription-costs SevenTnewS en 2026-07-30T08:00:00+02:00 Xbox Game Pass Ultimate Drops to $39 for Three Months, a Rare Break from Rising Subscription Costs https://www.seventnews.com/en/articles/japans-ai-sovereignty-play-is-orchestration-not-building-a-new-frontier-model SevenTnewS en 2026-07-29T22:26:37.529784+02:00 Japan's AI sovereignty play is orchestration, not building a new frontier model https://www.seventnews.com/en/articles/why-small-european-states-keep-handing-money-to-the-same-voice-ai-startup SevenTnewS en 2026-07-29T21:00:00+02:00 Why Small European States Keep Handing Money to the Same Voice AI Startup