https://www.seventnews.com/en/articles/bruno-40-lets-you-bring-your-own-ai-key-to-an-open-source-api-client SevenTnewS en 2026-07-26T13:38:00+02:00 Bruno 4.0 lets you bring your own AI key. Your data stays yours. https://www.seventnews.com/en/articles/the-08b-model-that-challenged-everything-about-document-parsing SevenTnewS en 2026-07-26T13:20:00+02:00 The 0.8B model that just broke document parsing's pipeline monopoly https://www.seventnews.com/en/articles/microsofts-4b-parameter-image-model-runs-on-a-single-a100-challenges-30b-systems SevenTnewS en 2026-07-26T13:00:00+02:00 A 4-billion-parameter model just did what 30-billion systems couldn't: fit on one GPU https://www.seventnews.com/en/articles/anthropic-launches-claude-science-an-ai-workbench-for-scientific-research SevenTnewS en 2026-07-26T11:00:00+02:00 Anthropic Launches Claude Science, an AI Workbench for Scientific Research https://www.seventnews.com/en/articles/better-llm-detectors-might-make-people-use-ai-more-not-less SevenTnewS en 2026-07-26T11:00:00+02:00 Better AI detectors might make people use AI more, not less https://www.seventnews.com/en/articles/alibabas-new-tts-model-speaks-16-languages-and-laughs-on-command SevenTnewS en 2026-07-26T11:00:00+02:00 Alibaba's new TTS model speaks 16 languages, laughs on command, and might actually ship https://www.seventnews.com/en/articles/the-continent-that-ai-forgot-how-grok-xiaomi-and-metas-billions-bypass-africa SevenTnewS en 2026-07-26T10:52:18.428263+02:00 The continent AI forgot: Grok, Xiaomi and Meta's billions bypass Africa https://www.seventnews.com/en/articles/sakanas-fugu-cyber-finds-the-vulnerabilities-claude-opus-5-was-designed-to-miss SevenTnewS en 2026-07-26T10:50:49.315953+02:00 Sakana's Fugu-Cyber finds the vulnerabilities Claude Opus 5 was designed to miss https://www.seventnews.com/en/articles/what-a-new-graph-fraud-detection-paper-reveals-about-ais-dirty-little-secret-label-leakage SevenTnewS en 2026-07-26T09:00:00+02:00 Your fraud detection model may be cheating. A new paper proves it. https://www.seventnews.com/en/articles/shared-routing-across-layers-gives-sparse-attention-a-7x-speed-boost-without-the-quality-hit SevenTnewS en 2026-07-26T09:00:00+02:00 One routing pass to rule them all: 7x faster LLM decoding without the quality trade-off https://www.seventnews.com/en/articles/microsoft-uncovers-gigawiper-a-backdoor-that-bundled-three-wiper-families-into-one SevenTnewS en 2026-07-26T08:00:00+02:00 The wiper that bundles three flavors of destruction into a single backdoor https://www.seventnews.com/en/articles/kling-drops-two-benchmarks-that-expose-how-bad-generative-video-actually-is SevenTnewS en 2026-07-26T07:10:00+02:00 Video generation's dirty secret is finally on the record https://www.seventnews.com/en/articles/anthropics-opus-5-is-here-and-the-government-is-watching SevenTnewS en 2026-07-26T02:59:05.166644+02:00 Anthropic's Opus 5 is here, and the government is watching https://www.seventnews.com/en/articles/the-memory-shortage-is-making-your-next-gadget-more-expensive SevenTnewS en 2026-07-26T02:58:17.550182+02:00 The memory shortage is making your next gadget more expensive https://www.seventnews.com/en/articles/openai-meta-and-40-others-sign-the-same-ai-policy-letter-heres-why-that-matters SevenTnewS en 2026-07-26T02:52:11.392203+02:00 OpenAI, Meta, and 40 others sign an AI policy letter. Here's why that matters. https://www.seventnews.com/en/articles/your-ai-coding-assistant-has-a-blind-spot-alibaba-is-trying-to-fix-it SevenTnewS en 2026-07-26T02:37:00+02:00 The blind spot in AI-generated code that Alibaba fixes mid-sentence https://www.seventnews.com/en/articles/the-boring-update-that-matters-why-polish-won-the-developer-tools-race SevenTnewS en 2026-07-26T00:12:44.461335+02:00 The boring update that matters: why polish won the developer tools race https://www.seventnews.com/en/articles/apple-opens-its-walled-garden-to-third-party-ai-models-in-xcode-27 SevenTnewS en 2026-07-25T21:00:00+02:00 Apple's walled garden just got a gate with an AI-shaped key https://www.seventnews.com/en/articles/grok-build-clis-latest-update-goes-for-polish-over-hype SevenTnewS en 2026-07-25T20:44:14.044751+02:00 Grok Build CLI fixed 20 bugs and proved polish beats features https://www.seventnews.com/en/articles/amds-new-ai-halo-chip-goes-after-the-one-thing-nvidia-doesnt-control-your-desk SevenTnewS en 2026-07-25T20:36:00+02:00 A 128 GB desktop that undercuts Nvidia cloud rentals by 6x https://www.seventnews.com/en/articles/a-158-ms-reasoning-model-just-rewrote-the-speed-quality-playbook SevenTnewS en 2026-07-25T20:31:52.996344+02:00 The 158 ms model that broke the speed-quality rule https://www.seventnews.com/en/articles/opus-5-crushes-fable-5-on-physics-test-while-costing-half-as-much SevenTnewS en 2026-07-25T20:20:44.195185+02:00 The $1.40 model that crushed its $2.82 sibling on physics https://www.seventnews.com/en/articles/nvidias-nemotron-4-is-fast-open-and-aimed-right-at-the-gulf-between-llama-and-gpt SevenTnewS en 2026-07-25T20:00:00+02:00 Nvidia's Nemotron-4 runs 30% cheaper than GPT-4. The catch is buried in the benchmark. https://www.seventnews.com/en/articles/apple-bundles-pixelmator-pro-into-creator-studio-with-ai-captioning-and-shape-generation SevenTnewS en 2026-07-25T18:00:00+02:00 Apple's Creator Studio update turns Pixelmator Pro into the bundle's secret weapon https://www.seventnews.com/en/articles/elevenlabs-brings-its-ai-summit-to-warsaw-and-makes-a-point-about-polish-tech SevenTnewS en 2026-07-25T15:00:00+02:00 ElevenLabs brought its AI summit to Warsaw. The message is bigger than voice tech. https://www.seventnews.com/en/articles/ai-agents-flunked-a-test-on-pathogen-surveillance-that-should-worry-us SevenTnewS en 2026-07-25T15:00:00+02:00 The 50 percent ceiling on AI pathogen surveillance that should keep us up at night https://www.seventnews.com/en/articles/meta-and-uiuc-researchers-train-vision-language-models-to-check-their-own-work-and-rethink-wrong-answers SevenTnewS en 2026-07-25T15:00:00+02:00 Meta's new training trick teaches AI to catch its own mistakes without a teacher https://www.seventnews.com/en/articles/your-mac-just-became-a-remote-gemini-agent-unless-youre-in-europe SevenTnewS en 2026-07-25T13:39:00+02:00 Your Mac just became a remote Gemini agent, unless you're in Europe https://www.seventnews.com/en/articles/the-robot-simulation-stack-is-fragmenting-and-recomposing-in-plain-view SevenTnewS en 2026-07-25T13:00:00+02:00 A dozen physics engines are fighting for the right to train your robot https://www.seventnews.com/en/articles/mistral-nano-runs-on-devices-with-under-1gb-of-ram-matching-85-of-its-bigger-siblings-reasoning-scores SevenTnewS en 2026-07-25T11:00:00+02:00 Mistral Nano runs on $80 hardware and matches 85% of its bigger sibling. That gap decides everything. https://www.seventnews.com/en/articles/when-a-coding-agent-fails-cheap-retries-beat-escalation-at-35-of-the-cost SevenTnewS en 2026-07-25T11:00:00+02:00 Faible coût, pas d'échec : un correctif pour les agents de codage à 35 % du prix https://www.seventnews.com/en/articles/alibabas-happy-oyster-lets-you-drive-through-a-desert-and-rewrite-the-plot-as-you-go SevenTnewS en 2026-07-25T10:45:00+02:00 Alibaba just launched an interactive AI world where you can steer the plot https://www.seventnews.com/en/articles/smaller-models-that-keep-second-guessing-themselves-can-beat-10x-bigger-ones SevenTnewS en 2026-07-25T10:39:00+02:00 The 4B model that beats 32B ones by refusing to trust itself https://www.seventnews.com/en/articles/your-smart-glasses-remember-everything-but-cant-tell-you-when-it-happened SevenTnewS en 2026-07-25T10:38:00+02:00 Your smart glasses remember everything but can't tell you when it happened https://www.seventnews.com/en/articles/monitors-miss-sabotage-in-ai-research-agents-half-the-time-new-benchmark-finds SevenTnewS en 2026-07-25T09:00:00+02:00 Your AI research assistant is sabotaging you, and you won't catch it half the time https://www.seventnews.com/en/articles/the-gulf-between-llms-and-scientific-method-a-new-benchmark-finds-models-can-describe-data-but-cannot-reason-through-it SevenTnewS en 2026-07-25T09:00:00+02:00 LLMs can describe data. They cannot reason through it. A new benchmark proves the gap is real. https://www.seventnews.com/en/articles/cisas-kev-catalog-grows-again-two-flaws-one-pattern-and-a-directive-that-demands-more-than-patching SevenTnewS en 2026-07-25T08:00:00+02:00 Two new CISA alerts, one hard question: did they get in before you patched? https://www.seventnews.com/en/articles/the-supervision-gap-in-agentic-rl-just-got-a-fix-from-hindsight SevenTnewS en 2026-07-25T07:11:00+02:00 RL's oldest blind spot just met a fix that lets agents learn from their own mistakes https://www.seventnews.com/en/articles/opencode-hit-40m-arr-by-betting-developers-want-choices-not-lock-in SevenTnewS en 2026-07-24T21:54:54.011490+02:00 How an Anthropic ban backfired and drove OpenCode to $40M ARR https://www.seventnews.com/en/articles/claude-opus-5-matches-near-frontier-performance-at-half-the-cost-with-a-deliberate-blind-spot-on-cyber SevenTnewS en 2026-07-24T19:39:18.672675+02:00 Claude Opus 5 nearly matches Fable 5 across benchmarks, with cybersecurity as a deliberate blind spot https://www.seventnews.com/en/articles/microsofts-in-house-ai-models-cut-gpu-costs-up-to-89-in-production SevenTnewS en 2026-07-24T19:38:28.268015+02:00 The 89% GPU cost cut that changes the math for every AI company https://www.seventnews.com/en/articles/why-the-best-document-parser-now-fits-in-800-million-parameters SevenTnewS en 2026-07-24T19:19:36.643000+02:00 The 0.8B model that beat every pipeline document parser at its own game https://www.seventnews.com/en/articles/googles-cyber-fine-tune-finds-vulnerabilities-that-claude-opus-46-misses SevenTnewS en 2026-07-24T17:00:00+02:00 Google's cheap fine-tune found vulnerabilities that Claude Opus 4.6 missed https://www.seventnews.com/en/articles/graph-learning-models-arent-tested-on-data-this-ugly-a-new-benchmark-changes-that SevenTnewS en 2026-07-24T15:00:00+02:00 Your graph model breaks on dirty data. A new benchmark shows exactly where. https://www.seventnews.com/en/articles/hugging-face-pulls-the-infrastructure-lever-not-the-model-lever SevenTnewS en 2026-07-24T15:00:00+02:00 Hugging Face isn't fighting the model war. It's making deployment boring.