https://aiholics.com/nvidia-cosmos3-h-embodied-nano-surgical-world-model/Aiholicsen2026-10-20T18:20:32+00:00Nvidia’s surgical robot world model learns to predict video and actions https://aiholics.com/gpt-oss-20b-multilingual-reasoning-sagemaker-hyperpod/Aiholicsen2026-10-20T17:54:58+00:00Amazon shows GPT-OSS-20B can learn to reason in French, but only after SFT https://aiholics.com/chatgpt-locked-chats-4-digit-pin/Aiholicsen2026-10-20T17:29:44+00:00ChatGPT locked chats hide conversations behind a 4-digit PIN https://aiholics.com/anthropic-public-benefit-fund-5-percent-shares-2/Aiholicsen2026-10-20T17:11:18+00:00Anthropic puts 5% of its shares into a public benefit fund https://aiholics.com/amd-rocopt-gpu-decision-optimization-rocm/Aiholicsen2026-10-20T17:05:16+00:00AMD rocOpt brings Nvidia cuOpt-style GPU solvers to Instinct, minus one key method https://aiholics.com/gemini-small-business-hub-shopify-canva-xero/Aiholicsen2026-10-20T16:29:20+00:00Gemini gets a small business hub with Shopify, Canva and Xero links https://aiholics.com/mistral-voxtral-mini-transcribe-3-pricing-diarization/Aiholicsen2026-10-20T16:10:15+00:00Mistral’s Voxtral Mini Transcribe 3 costs $0.0035 a minute and ranks 3rd on speaker labeling https://aiholics.com/google-vllm-sglang-torchtpu-inference-engines/Aiholicsen2026-10-20T15:09:31+00:00Google brings vLLM and SGLang to TPUs in PyTorch, with Meta already running it https://aiholics.com/elevenlabs-eleven-realtime-speech-to-speech-voice-agents/Aiholicsen2026-10-20T14:15:39+00:00ElevenLabs bets on one-model voice agents, with 98.8% on a speech reasoning test https://aiholics.com/deepseek-deepinfer-open-source-inference-stack/Aiholicsen2026-10-20T13:58:15+00:00DeepSeek open-sources DeepInfer, the stack behind its own services, minus key parts https://aiholics.com/apple-american-manufacturing-program-iphone-duo-titanium/Aiholicsen2026-10-20T12:10:18+00:00Apple’s foldable iPhone Duo gets titanium from a Pennsylvania mill https://aiholics.com/llm-d-1-0-production-inference-self-hosting-cost/Aiholicsen2026-10-20T12:05:53+00:00Llm-d 1.0 pitches self-hosted AI inference at 5 to 10 times lower cost https://aiholics.com/amazon-16-billion-germany-investment-plan/Aiholicsen2026-10-20T11:10:44+00:00Amazon’s €16 billion German plan goes well beyond data centers https://aiholics.com/google-gemini-for-government-services-preview/Aiholicsen2026-10-20T09:55:50+00:00Google puts its Gemini agent to work on FOIA requests and grant audits https://aiholics.com/whatsapp-help-pages-hidden-ai-instructions-regulator/Aiholicsen2026-10-20T05:19:47+00:00WhatsApp help pages held hidden AI notes against naming regulators, a report says https://aiholics.com/open-weight-models-explained/Aiholicsen2026-10-20T05:13:44+00:00Open-weight models: What you get when an AI model can be downloaded but not rebuilt https://aiholics.com/disney-character-ai-bankruptcy-deal-lawsuits/Aiholicsen2026-10-20T00:54:52+00:00Disney’s Character.AI deal takes the staff and tech but leaves the lawsuits behind https://aiholics.com/hacker-ai-agents-break-into-scores-of-sites-reuters/Aiholicsen2026-10-20T00:51:44+00:00A hacker reportedly ran AI agents against scores of sites for a few thousand dollars https://aiholics.com/tenstorrent-1-5-billion-series-e-6-2-billion-valuation/Aiholicsen2026-10-20T00:39:26+00:00Tenstorrent’s $1.5 billion round values the AI chip startup at $6.2 billion https://aiholics.com/samsung-art-store-art-basel-paris-2026-collection/Aiholicsen2026-10-20T00:36:17+00:00Samsung’s Art TVs get 19 Art Basel Paris works, for subscribers only https://aiholics.com/claude-structured-outputs-amazon-bedrock-json-schema/Aiholicsen2026-10-19T23:40:51+00:00Claude on Amazon Bedrock can now be forced to return valid JSON https://aiholics.com/github-copilot-model-policy-api-enterprise-teams/Aiholicsen2026-10-19T22:15:36+00:00Copilot admins can now script which AI models each team gets https://aiholics.com/grok-4-7-amazon-bedrock-aws-govcloud-us/Aiholicsen2026-10-19T21:09:19+00:00Grok 4.7 reaches AWS GovCloud, a month after its commercial Bedrock debut https://aiholics.com/cern-mistral-large-physics-model-lhc/Aiholicsen2026-10-19T20:59:42+00:00CERN and Mistral plan one AI model to read every LHC collision https://aiholics.com/hugging-face-torchneuron-trainium-transformers-native/Aiholicsen2026-10-19T20:30:58+00:00Hugging Face drops its Trainium middleman: Transformers now runs on Neuron directly https://aiholics.com/chroma-awards-season-2-elevenlabs-450000-prizes/Aiholicsen2026-10-19T19:55:27+00:00Chroma Awards Season 2 offers $450,000+ in prizes, and ElevenLabs adds its own $22,000 https://aiholics.com/github-copilot-app-session-continue-vs-code/Aiholicsen2026-10-19T18:55:22+00:00GitHub Copilot app sessions can now be opened in VS Code, branch and all https://aiholics.com/deepgram-nova-3-aws-trainium2-audio-per-dollar/Aiholicsen2026-10-19T18:25:39+00:00Deepgram’s Nova-3 now runs on Trainium2 and gets up to 70% more audio per dollar https://aiholics.com/claude-code-smart-tv-ads-tracking-removal/Aiholicsen2026-10-19T18:22:48+00:00Developers are using Claude Code to strip ads and tracking from smart TVs https://aiholics.com/gemini-notebook-review-toggle-slides-video-overviews/Aiholicsen2026-10-19T17:05:12+00:00Gemini Notebook lets you edit the plan before it builds your slides or video https://aiholics.com/twelvelabs-pegasus-1-5-amazon-bedrock-video-json/Aiholicsen2026-10-19T16:58:43+00:00Amazon Bedrock’s Pegasus 1.5 turns two-hour videos into timestamped JSON https://aiholics.com/nvidia-nemotron-3-5-super-vl-bolt-open-models/Aiholicsen2026-10-19T16:49:43+00:00Nvidia’s 120B Super VL agent model has open weights and costs $0.10 per million tokens https://aiholics.com/minimax-m3-1-flash-price-speed-open-weights/Aiholicsen2026-10-19T16:07:45+00:00MiniMax M3.1 Flash costs half of M3 per token, and its weights are promised https://aiholics.com/microsoft-project-perception-ai-agents-insider-threat/Aiholicsen2026-10-19T16:05:50+00:00Microsoft’s new agent security assumes your AI agents may turn on you https://aiholics.com/dhs-orders-deepseek-removed-civilian-agency-systems/Aiholicsen2026-10-19T15:23:38+00:00US civilian agencies must drop DeepSeek, and so must vendors built on it https://aiholics.com/amazon-torchneuron-pytorch-trainium-no-rewrite/Aiholicsen2026-10-19T14:09:23+00:00Amazon says PyTorch code now runs on Trainium chips with one setting changed https://aiholics.com/claude-quantum-circuits-beat-published-designs/Aiholicsen2026-10-19T13:01:31+00:00Claude wrote quantum circuits that beat published designs on all eight tasks https://aiholics.com/ai-agent-harness-explained/Aiholicsen2026-10-19T12:43:02+00:00AI agent harness: The setup around a model that decides how well it works https://aiholics.com/presumption-of-conformity-eu-ai-act-harmonised-standards/Aiholicsen2026-10-19T10:28:47+00:00Presumption of conformity: How EU harmonised standards shortcut AI Act compliance https://aiholics.com/first-eu-ai-act-harmonised-standard-en-18286-published/Aiholicsen2026-10-19T08:19:30+00:00First EU AI Act standard is official: One of ten promised is out https://aiholics.com/speech-to-speech-models-explained/Aiholicsen2026-10-19T03:10:14+00:00Speech-to-speech models: How voice AI listens and talks without turning speech into text https://aiholics.com/google-harness2-self-improving-ai-agents-open-source/Aiholicsen2026-10-19T03:10:03+00:00Google’s Harness2 lets a weaker AI model beat a stronger one by rewriting its own setup https://aiholics.com/anthropic-public-benefit-fund-5-percent-shares/Aiholicsen2026-10-19T00:59:31+00:00Anthropic puts 5% of its shares into a public benefit fund, with more to follow https://aiholics.com/nvfp4-4-bit-format-explained/Aiholicsen2026-10-18T20:56:56+00:00NVFP4: How 4-bit numbers let huge AI models fit in far less memory https://aiholics.com/ai-agent-memory-how-assistants-remember/Aiholicsen2026-10-18T18:45:46+00:00AI agent memory: How a chatbot learns to remember you between conversations