Send us a text🌍 INAI • The Open AI HubThe Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news—free for all, updated every day.https://github.com/inai-sandy/inAI-wiki**Top Highlights:**- OpenAI released Sora 2 with photorealistic sound-synced video, sparking legal debates- OpenAI's $500B "Stargate" infrastructure initiative with Samsung, SK Hynix, Oracle, and Nvidia signals AI hardware race in Asia- White House tasked NIST with benchmarking US AI capabilities against global rivals- Google replaces Assistant with Gemini across Nest devices and Home app- Meta to use AI chat interactions for ad targeting across its apps**New Tools & Technologies:**- Tinker: API for synthetic data and distributed fine-tuning- Slack's context-aware APIs for AI agents in work conversations- LlamaAgents offers one-click document agent deployment- Hume Octave 2 and LiquidAI LFM2-Audio: multilingual text-audio models- Amazon FAR's OmniRetarget for robot motion retargeting- DeeptraceReward: 94% accurate deepfake detector**LLM Updates:**- Claude Sonnet 4.5 improves coding and creativity- Gemini 2.5 with Goedel-Prover V2 sets Putnam SOTA- DeepSeek V3.2 reduces token usage; QuestA (1.5B) advances small-model reasoning- GLM-4.6 and Qwen3-VL expand developer options- OpenAI Codex shows strong real-world results- Accessibility gains with Apriel-1.5-15B-Thinker, HunyuanImage 3.0, Dragon Hatchling**Research Highlights:**- MIT: ChatGPT reduces cognitive effort and memory retention- Self-evolving AI agents can unlearn safety protocols- FDA-cleared AI medical devices lack randomized trials- MENLO's 47-language benchmark for multilingual evaluation- DeepMind's AlphaEvolve discovers new complexity theory results**Industry & Policy:**- Stripe Open Issuance enables custom stablecoins and AI commerce- Microsoft and Databricks launch AI-first cybersecurity platforms**Tutorials & Guides:**- Advanced RAG strategies guide for production systems- vLLM deployment with Hugging Face Inference Endpoints- Anthropic's deep-dive on context design vs prompt engineering- Multi-agent system design explainer- Open-source multimodal tools survey on Hugging Face**Demos & Discussions:**- Veo 3's zero-shot reasoning over physical interactions- Kling 2.5 Turbo and Seedream 4K for rapid world-building- Lucid Origin shortens idea-to-video pipeline- Gemini-powered Ross Lovegrove collaboration; Moondream 3 previews- RLHI and prompt optimization show gains; reward design critical- New theories on optimizer behavior at stability edges- RL advancements in skill composition and humanoid manipulation- Only 11% of Python developers use coding agents regularly- Debates on GPU allocation priorities and content provenance- Community discussions on GRPO, RL startups, and open-source modelsSupport the show

Ver todos los episodios