Artwork
iconShare
 
Manage episode 515503366 series 3670986
Content provided by Sandy. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by Sandy or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://staging.podcastplayer.com/legal.

Send us a text

🌍 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

Top Highlights: AMD signed a major chip deal with OpenAI worth potentially $100B with 10% equity. Anthropic gained access to ~1M Google TPUs. OpenAI's Sora 2 reached 1M downloads in 5 days, Android version coming. Salesforce launched MuleSoft Agent Fabric for enterprise AI governance. UK released AI scribing framework; OpenAI added UK data residency.

New Tools: Mistral AI's agent studio offers production observability. Meta open-sourced CTran for cross-vendor GPU support. Karpathy released Nanochat, an open chatbot pipeline. vLLM added TPU support and DeepSeek-OCR. Runway democratized video workflows. OpenRouter introduced :exacto for reliable tool-calling.

LLM Updates: Meta's ScaleRL predicts RL scaling. Lookahead improves model routing. Baseten achieved 650+ tokens/s on 120B models. MiniMax M2 excels at coding tasks. Scaf-GRPO boosted reasoning 44%.

Research: Anthropic's ImpossibleBench tests agent spec-following. Black-box model theft detection advances. Tahoe-x1 3B biotech model released. MedSAM gains clinical traction. New theory links graph algorithms to Transformers.

Industry: AMD-OpenAI deal faces TSMC bottlenecks. Anthropic-Google TPU expansion fuels browser competition. OpenAI backed Valthos ($30M); safety lawsuits mount. Illinois Bar issued AI ethics guidance.

Tutorials: Together AI published Nanochat deployment guide. PyTorch debugging insights shared. RL environment design explainer. Modular's Mojo GPU Puzzles (34 challenges). Stanford's full AI curriculum online.

Showcases: Vision Pro delivers 4K 120Hz wireless VR. Veo 3.1 frame editing impresses. Wan 2.2 + Glif achieve film-quality swaps. Higgsfield Popcorn enables precise video variants.

Discussions: Tensor Logic aims to unify reasoning paradigms. Coverage profiles link pretraining to performance. RAG critiques warn of shallow retrieval. Coding benchmark mismatches noted. Open-source defended as power check.

Support the show

🌍 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

  continue reading

147 episodes