13th January - AI News Daily - Apple Partners with Google Gemini to Power Next-Gen Siri and Intel...

Описание: Send us a text (https://www.buzzsprout.com/twilio/tex...)


🌍 INAI • The Open AI Hub


The 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 (https://github.com/inai-sandy/inAI-wiki/)


Top Highlights: Apple partners with Google Gemini for Siri and Apple Intelligence. Google launches Universal Commerce Protocol with Walmart and Shopify for in-search checkout. Governments escalate deepfake crackdowns—Southeast Asia bans Grok, UK criminalizes "nudification". OpenAI and SoftBank commit $1B to Texas AI data centers. Ramp's coding agent writes ~30% of merged code.


New Tools: Ramp's async agent delivers real productivity in enterprise repos. Apple GenCtrl API enables explicit creative model control. moPPIt generates protein binders in seconds. AnyDepth democratizes depth estimation for robotics and AR. Google Gemini API reads files from Cloud Storage up to 2GB. Breakthrough transcription converts hour-long video in 80 seconds.


LLM Updates: Zhipu GLM-4.7 offers 200K context and tops Code Arena. Claude 4.5 Opus leads SWE-fficiency benchmark. GPT-5.2 hits 99.4% on PutnamBench. Gemini Nano generates 1B+ images in 53 days. Devstral 2 opens for free use. Hugging Face adds GLM-4.7 and "chat with papers".


Research: DeepSeek Engram improves retrieval with hashed N-gram memory. Positional recalibration extends context without full retraining. Recursive Language Models handle million-token prompts. Stanford and NVIDIA show test-time training compresses live context into weights. NVIDIA NVFP4 enables stable 4-bit training. FineTranslations releases trillion-token English corpus.


Industry: OpenAI acquires Torch for ChatGPT Health data. Agent security risks emerge—ZombieAgent exploit and fake Grok malware.


Tutorials: Stanford CS224N adds agent modules. 12 fine-tuning methods compiled. Prompt caching cuts costs. Agent hygiene emphasizes traces. Portfolios trump credentials.


Showcases: Mortar prototypes games via AI co-creation. Gemini powers 3D geospatial storytelling. Claude builds MRI viewer and controls Blender. Kling 2.6 Motion Brush enables precise dances. OpenEnv offers $10K for RL agents. AI anime compresses history.


Discussions: AI op-ed sparks newsroom ethics debate. Likert scales criticized for LLM evals. Opinionated agents with clear defaults advocated. 2026 seen as breakthrough year for AI science. Concerns over surveillance and propaganda.


Support the show (https://www.buzzsprout.com/2507996/su...)


🌍 INAI • The Open AI Hub


The 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 (https://github.com/inai-sandy/inAI-wiki/)