https://www.a17news.com AI news in a flash - the fastest way to know what's happening in AI. Short, sharp, and always up to date. Live bulletins on models, startups, big players, tools, and research. en-us Sat, 16 May 2026 18:03:59 GMT 60 A17News / Next.js https://www.a17news.com/og.png https://www.a17news.com https://www.a17news.com/news/aurora-ceo-discusses-scaling-self-driving-truck-operations-1778356186 https://www.a17news.com/news/aurora-ceo-discusses-scaling-self-driving-truck-operations-1778356186 Aurora co-founder and CEO Chris Urmson discusses the journey and current scaling phase of their commercial driverless truck operations, marking a shift from early trials to fleet expansion. Sat, 09 May 2026 20:15:00 GMT Startups TechCrunch AI https://www.a17news.com/news/china-s-moonshot-ai-achieves-20b-valuation-amid-open-source-ai-surge-1778356223 https://www.a17news.com/news/china-s-moonshot-ai-achieves-20b-valuation-amid-open-source-ai-surge-1778356223 Moonshot AI has achieved a $20B valuation fueled by booming demand for open source AI, with annualized recurring revenue exceeding $200 million in April. Sat, 09 May 2026 20:05:00 GMT Startups TechCrunch AI https://www.a17news.com/news/grok-code-fast-1-deprecation-coming-for-github-copilot-features-1778356252 https://www.a17news.com/news/grok-code-fast-1-deprecation-coming-for-github-copilot-features-1778356252 Grok Code Fast 1 will be deprecated across all GitHub Copilot experiences and code features on May 15th. Sat, 09 May 2026 20:03:11 GMT Models GitHub Changelog https://www.a17news.com/news/perplexity-500m-series-d https://www.a17news.com/news/perplexity-500m-series-d The AI answer engine doubles its valuation in eight months as enterprise revenue triples year over year. Sat, 09 May 2026 17:30:22 GMT Startups TechCrunch https://www.a17news.com/news/openai-gptx-reasoning https://www.a17news.com/news/openai-gptx-reasoning A major upgrade focused on multi-step reasoning and 4x faster inference, now live in the API. Sat, 09 May 2026 17:00:22 GMT Models OpenAI Blog https://www.a17news.com/news/figure-03-humanoid-pilot https://www.a17news.com/news/figure-03-humanoid-pilot 100 units begin live production work in Spartanburg - the largest humanoid deployment in a US factory to date. Sat, 09 May 2026 16:00:22 GMT Robotics Figure https://www.a17news.com/news/microsoft-copilot-azure-foundry https://www.a17news.com/news/microsoft-copilot-azure-foundry A unified billing, deployment, and policy plane for every Copilot surface across Microsoft's stack. Sat, 09 May 2026 16:00:22 GMT Big Players Microsoft https://www.a17news.com/news/langchain-v1-stable-runtime https://www.a17news.com/news/langchain-v1-stable-runtime After three years and 200K stars, LangChain commits to a frozen API surface and long-term support. Sat, 09 May 2026 15:00:22 GMT Tools LangChain https://www.a17news.com/news/anthropic-claude-sonnet-4-7 https://www.a17news.com/news/anthropic-claude-sonnet-4-7 A mid-tier model with new computer-use reliability and a 1M token context window. Sat, 09 May 2026 14:00:22 GMT Models Anthropic https://www.a17news.com/news/openai-superalignment-eval https://www.a17news.com/news/openai-superalignment-eval A new benchmark to test whether frontier models help, hide from, or sabotage their evaluators. Sat, 09 May 2026 14:00:22 GMT Safety OpenAI https://www.a17news.com/news/deepmind-linear-attention-32k https://www.a17news.com/news/deepmind-linear-attention-32k A new state-space hybrid matches softmax attention quality with O(n) memory at long context. Sat, 09 May 2026 13:00:22 GMT Research arXiv https://www.a17news.com/news/cursor-1b-arr https://www.a17news.com/news/cursor-1b-arr The AI code editor becomes one of the fastest-growing developer tools in history. Sat, 09 May 2026 12:00:22 GMT Startups The Information https://www.a17news.com/news/huggingface-spaces-pro https://www.a17news.com/news/huggingface-spaces-pro Persistent GPU instances, custom domains, and zero cold-starts for production AI showcases. Sat, 09 May 2026 11:00:22 GMT Tools Hugging Face https://www.a17news.com/news/eu-ai-act-gpai-codes-deadline https://www.a17news.com/news/eu-ai-act-gpai-codes-deadline Major AI providers must now comply with transparency and copyright rules for general-purpose AI models under the Act's first enforcement milestone. Sat, 09 May 2026 10:00:22 GMT Regulation EU AI Office https://www.a17news.com/news/mistral-200b-moe-open https://www.a17news.com/news/mistral-200b-moe-open A 22B-active mixture-of-experts model with Apache 2.0 weights, matching GPT-4-class quality. Sat, 09 May 2026 09:00:22 GMT Models Mistral AI https://www.a17news.com/news/waymo-nyc-public-launch https://www.a17news.com/news/waymo-nyc-public-launch After a year of mapped testing, Waymo One goes live to all NYC riders below 110th Street. Sat, 09 May 2026 08:00:22 GMT Robotics Waymo https://www.a17news.com/news/apple-foundation-models-ios-19 https://www.a17news.com/news/apple-foundation-models-ios-19 A 3B parameter on-device LLM is now exposed to third-party apps via a new Swift framework. Sat, 09 May 2026 07:00:22 GMT Big Players Apple https://www.a17news.com/news/vercel-ai-gateway https://www.a17news.com/news/vercel-ai-gateway One endpoint, 100+ models, automatic fallback, and unified billing. Sat, 09 May 2026 06:00:22 GMT Tools Vercel https://www.a17news.com/news/agentic-rl-survey https://www.a17news.com/news/agentic-rl-survey A comprehensive review of reinforcement-learned agents - from RLHF descendants to self-play loops. Sat, 09 May 2026 05:00:22 GMT Research arXiv https://www.a17news.com/news/sakana-evolutionary-merging-api https://www.a17news.com/news/sakana-evolutionary-merging-api Automatically discover high-performing model merges across thousands of open-weight checkpoints. Sat, 09 May 2026 04:00:22 GMT Startups Sakana AI https://www.a17news.com/news/amd-mi400-inference-chip https://www.a17news.com/news/amd-mi400-inference-chip A new data-center GPU with 288GB HBM3e and ROCm 7 software stack targets LLM inference workloads at hyperscale. Sat, 09 May 2026 03:00:22 GMT Hardware AMD https://www.a17news.com/news/deepfake-political-ad-ban https://www.a17news.com/news/deepfake-political-ad-ban Generative voice and video clones of candidates now require sworn human-actor disclosure on every paid spot. Sat, 09 May 2026 02:00:22 GMT Safety FCC https://www.a17news.com/news/nvidia-blackwell-ultra https://www.a17news.com/news/nvidia-blackwell-ultra 288GB of HBM3e and 1.5x training throughput vs B200, shipping to hyperscalers in Q3. Sat, 09 May 2026 02:00:22 GMT Big Players Nvidia https://www.a17news.com/news/google-gemini-3-10m-context https://www.a17news.com/news/google-gemini-3-10m-context A new flagship with native video understanding and a 10M token effective context window. Sat, 09 May 2026 00:00:22 GMT Models Google DeepMind https://www.a17news.com/news/1x-neo-home-beta https://www.a17news.com/news/1x-neo-home-beta $20K consumer humanoid begins shipping to invited households in SF, Austin, and NYC. Fri, 08 May 2026 22:00:22 GMT Robotics 1X https://www.a17news.com/news/replicate-streaming-any-model https://www.a17news.com/news/replicate-streaming-any-model Token-by-token streaming over HTTP for every model in the catalog, with no cold-start tax. Fri, 08 May 2026 22:00:22 GMT Tools Replicate https://www.a17news.com/news/reka-enterprise-multimodal https://www.a17news.com/news/reka-enterprise-multimodal Custom multimodal models trained on a customer's own video, image, and document corpora. Fri, 08 May 2026 20:00:22 GMT Startups Reka https://www.a17news.com/news/stanford-rag-dropouts https://www.a17news.com/news/stanford-rag-dropouts Randomly masking retrieved passages during training improves grounded-answer accuracy. Fri, 08 May 2026 18:00:22 GMT Research Stanford CRFM https://www.a17news.com/news/anthropic-jailbreak-classifier https://www.a17news.com/news/anthropic-jailbreak-classifier A second-generation safeguard layer that filters both inputs and outputs without significant capability loss. Fri, 08 May 2026 18:00:22 GMT Safety Anthropic https://www.a17news.com/news/amazon-anthropic-8b https://www.a17news.com/news/amazon-anthropic-8b AWS deepens its bet on Claude with another $4B and exclusive Trainium training commitment. Fri, 08 May 2026 17:00:22 GMT Big Players Reuters https://www.a17news.com/news/meta-llama-4-vision https://www.a17news.com/news/meta-llama-4-vision Native multimodal weights up to 90B, beating GPT-4V on three of four standard benchmarks. Fri, 08 May 2026 15:00:22 GMT Models Meta AI https://www.a17news.com/news/us-senate-ai-liability-bill https://www.a17news.com/news/us-senate-ai-liability-bill The bipartisan SAFE AI Act would make frontier model developers legally liable for harms caused by their models in high-risk domains. Fri, 08 May 2026 14:00:22 GMT Regulation Reuters https://www.a17news.com/news/pinecone-free-5m-tier https://www.a17news.com/news/pinecone-free-5m-tier A vastly expanded free tier as competition with pgvector and Turbopuffer intensifies. Fri, 08 May 2026 12:00:22 GMT Tools Pinecone https://www.a17news.com/news/mit-1bit-llm-reasoning https://www.a17news.com/news/mit-1bit-llm-reasoning Ternary-weight models maintain quality on GSM8K and MMLU at a fraction of memory and energy. Fri, 08 May 2026 09:00:22 GMT Research arXiv https://www.a17news.com/news/together-runpod-acquisition https://www.a17news.com/news/together-runpod-acquisition Consolidation in the GPU cloud space as Together expands beyond pure inference. Fri, 08 May 2026 07:00:22 GMT Startups Bloomberg https://www.a17news.com/news/xai-grok-4-code-interpreter https://www.a17news.com/news/xai-grok-4-code-interpreter Real-time tool use, deep web search, and a sandboxed Python runtime baked into the model. Fri, 08 May 2026 04:00:22 GMT Models xAI https://www.a17news.com/news/cerebras-wafer-scale-3 https://www.a17news.com/news/cerebras-wafer-scale-3 The third-generation WSE cuts training time for 70B-parameter models by 3x compared to a 512-GPU H100 cluster. Fri, 08 May 2026 02:00:22 GMT Hardware Cerebras